**Female Specific Association of Low Insulin-Like Growth Factor 1 (IGF1) Levels with Increased Risk of Premature Mortality in Renal Transplant Recipients**

**Frank Klont 1,\*, Lyanne M. Kieneker 2, Antonio W. Gomes-Neto 2, Suzanne P. Stam 2, Nick H. T. ten Hacken 3, Ido P. Kema 4, André P. van Beek 5, Else van den Berg 2, Péter Horvatovich 1, Rainer Bischo**ff **<sup>1</sup> and Stephan J. L. Bakker <sup>2</sup>**


Received: 19 December 2019; Accepted: 17 January 2020; Published: 21 January 2020

**Abstract:** Associations between insulin-like growth factor 1 (IGF1) and mortality have been reported to be female specific in mice and in human nonagenarians. Intervention in the growth hormone (GH)-IGF1 axis may particularly benefit patients with high risk of losing muscle mass, including renal transplant recipients (RTR). We investigated whether a potential association of circulating IGF1 with all-cause mortality in stable RTR could be female specific and mediated by variation in muscle mass. To this end, plasma IGF1 levels were measured in 277 female and 343 male RTR by mass spectrometry, and their association with mortality was assessed by Cox regression. During a median follow-up time of 5.4 years, 56 female and 77 male RTR died. In females, IGF1 was inversely associated with risk (hazard ratio (HR) per 1-unit increment in log2-transformed (doubling of) IGF1 levels, 95% confidence interval (CI)) of mortality (0.40, 0.24–0.65; *p* < 0.001), independent of age and the estimated Glomerular filtration rate (eGFR). In equivalent analyses, no significant association was observed for males (0.85, 0.56–1.29; *p* = 0.44), for which it should be noted that in males, age was negatively and strongly associated with IGF1 levels. The association for females remained materially unchanged upon adjustment for potential confounders and was furthermore found to be mediated for 39% by 24 h urinary creatinine excretion. In conclusion, low IGF1 levels associate with an increased risk of all-cause mortality in female RTR, which may link to conditions of low muscle mass that are known to be associated with poor outcomes in transplantation patients. For males, the strongly negative association of age with IGF1 levels may explain why low IGF1 levels were not found to be associated with an increased risk of all-cause mortality.

**Keywords:** insulin-like growth factor 1; growth hormone; muscle mass; patient survival; physical activity; renal transplant recipients

#### **1. Introduction**

The peptide hormone insulin-like growth factor 1 (IGF1) is a key mediator of the biochemical/endocrine effects of growth hormone (GH) [1]. Synthesis of IGF1 is regulated by GH and mainly takes place in the liver after which IGF1 is secreted and transported to other tissues, where it acts as an endocrine hormone [2,3]. IGF1 provides a stable, integrated measure of the activity of the somatotropic axis thereby contrasting with GH secretion which is highly variable [3].

Reduced GH and IGF1 signaling extends lifespan in many laboratory models, including worms, yeast, and drosophila [4]. A specific role for IGF1 receptor signaling in mammalian longevity was first established in IGF1 receptor-(haplo) insufficient mice. These mice lived 33% longer than their wildtype littermates, yet this effect was restricted to females [5], which was subsequently confirmed in two follow-up studies in mice [6,7]. A similar link between IGF1 receptor-insufficiency and longevity has been proposed for humans following observations in several studies [8–10]. Moreover, IGF1 levels predict better survival in nonagenarians (i.e., people between the age of 90 and 99), and, notably, the corresponding association between IGF1 levels and longevity was found to be female specific [11]. It remains, however, unclear whether circulating levels of IGF1 are also associated with longevity in middle-aged subjects and whether such association is female specific.

Studying the association between IGF1 levels and longevity (survival) in specific patient groups appears to be interesting as well, for example, following the growing interest in ghrelin receptor agonists targeting the GH-IGF1 axis to potentially reverse the anorexia–cachexia syndrome in a variety of conditions, including renal insufficiency [12–15]. An important mechanism by which stimulation of the GH-IGF1 axis may improve long-term outcome is through stimulation of muscle mass accretion [15,16]. To this regard, a large and growing group of patients that might be worthwhile studying is that of renal transplant recipients (RTR), in which protein–energy wasting is always lurking [17–19]. In fact, it has been found that the risk of premature mortality in this population is 6–7 times higher compared to the general population [20], and this risk was particularly high in RTR with low muscle mass, as reflected by low 24 h urinary creatinine excretion [21,22]. Recent studies furthermore suggested that 24 h urinary creatinine excretion may be a noninvasive, easily accessible, inexpensive, and direct measurement of total body muscle mass [19], while this measure is often not included in clinical studies to complement the imaging technique armamentarium which is applied for evaluation of muscle mass in observational and clinical intervention studies [23–25].

In this study, we aimed to investigate (1) the nature of the association between circulating levels of IGF1 and mortality in RTR, (2) whether such (potential) association is female specific, and (3) furthermore whether such (potential) association could, in part or as a whole, be mediated by variation in muscle mass, as reflected by 24 h urinary creatinine excretion.

#### **2. Experimental Section**

#### *2.1. Study Population*

All RTR (aged ≥ 18 years) that were transplanted at the University Medical Center Groningen (UMCG) and that were one year or longer post-transplantation were approached for participation in this study during outpatient clinic visits between 2008 and 2010, as described previously [26]. The RTR included in this study had no known or apparent systemic diseases (e.g., malignancies, active infections) at inclusion. Written informed consent was obtained from 707 (87%) of the 817 RTR that were initially invited, and plasma IGF1 levels were measured in 620 RTR (76%). For this study, ethical approval has been granted by the UMCG's review board (METc 2008/186), and the study adheres to the Declaration of Helsinki. The study is registered as 'TransplantLines Food and Nutrition Biobank and Cohort Study (TxL-FN)' at ClinicalTrials.gov (NCT identifier 'NCT02811835 ).

#### *2.2. Data and Sample Collection*

Measurement of clinical parameters has been described in detail previously [26]. Physical activity was assessed with the Short QUestionnaire to ASsess Health-enhancing physical activity (SQUASH) as developed and validated by the Dutch National Institute of Public Health and Environment to assess daily life physical activity in the Dutch adult population [27]. Information on medical history and medication use was obtained from patient records. Diabetes was defined as the use of antidiabetic medication or fasting plasma glucose of at least 7.0 mmol/L. Twenty-four h urine was collected (per strict protocol) a day before the outpatient clinic visits while blood was drawn in the morning on the day of the outpatient clinic visit, yet after completion of the 24 h urine collection.

#### *2.3. Laboratory Procedures*

Blood and urine markers were measured by routine laboratory procedures with the exception of serum creatinine which was assessed using a modified version of the Jaffé method (MEGA AU 510; Merck Diagnostica, Darmstadt, Germany), and the urine total protein concentration which was obtained using the Biuret reaction (MEGA AU 510; Merck Diagnostica). Renal function was estimated with the 2012 Chronic Kidney Disease Epidemiology (CKD-EPI) Collaboration equation using both serum creatinine and cystatin C [28]. IGF1 was assessed in plasma samples (which had not undergone any previous freeze–thaw cycle) using a semi-automated mass spectrometric IGF1 assay [29] which was validated according to FDA guidelines [30]. The samples were analyzed in 13 analytical runs containing up to 81 clinical samples per run, as well as nine calibration samples and duplicate quality control (QC) samples at three concentrations (i.e., low, midrange, and high IGF1 levels). All runs met the acceptance criteria stipulated in the FDA guidelines thereby featuring 75% (though at least six) of the calibration samples with back-calculated levels within 15% (or 20% for the lowest level calibration sample) of their expected value, and at least 67% of the QC samples (though at least one replicate per QC level) yielding IGF1 levels within 15% of their respective nominal value (see Figures S1 and S2 in the Supplementary Material).

#### *2.4. Outcome Ascertainment*

All-cause mortality was the primary outcome of this study and was recorded until the end of September 2015. Up-to-date information on patient status was obtained on the basis of a continuous surveillance system of the outpatient program. In case the status of a patient was unknown, general practitioners or referring nephrologists were contacted. There was no loss to follow-up for the outcome. Specific causes of mortality were secondary outcomes of this study. This information was obtained by linking patient numbers to the database of the Dutch Central Bureau of Statistics (CBS) to retrieve causes of mortality reported by physicians. Infectious mortality was defined as mortality from infectious causes [31]. Cardiovascular mortality was defined as mortality caused by cardiovascular pathology, coded by ICD-10 codes I10-I52 [32]. Mortality due to malignancies was defined as mortality caused by malignant diseases. Miscellaneous causes of mortality were defined as other causes of death, not included in mortality from infectious causes, cardiovascular mortality, or mortality due to malignancies.

#### *2.5. Statistical Analyses*

Data analyses were performed using IBM SPSS Statistics for Windows (version 23.0.0.0; IBM Corp., Armonk, NY, USA) and STATA/SE (version 15.1; StataCorp, College Station, TX, USA). All p-values are two-tailed, and a p-value lower than 0.050 was considered statistically significant. Baseline characteristics are presented according to tertiles of plasma IGF1 levels for female and male RTR. Continuous data are presented as mean with SD for normally distributed variables and as median with interquartile range (IQR) for variables with skewed distributions, whereas categorical variables are presented as percentages. Differences in baseline characteristics across the tertiles were tested by one-way ANOVA, Kruskal–Wallis test, and linear-by-linear association χ<sup>2</sup> test for normally distributed

continuous, skewed continuous, and categorical variables, respectively. Multivariable linear regression was performed to assess associations between patients' characteristics and plasma IGF1 levels in female and male RTR. Models were included for analyses that were adjusted for age alone, for age and estimated glomerular filtration rate (eGFR), and for multiple variables selected following (automatic) stepwise backward elimination. The prospective associations of plasma IGF1 levels with all-cause mortality, as primary endpoint, and with cause-specific mortality, as secondary endpoint, were assessed by Cox proportional hazards regression. In order to verify the existence of effect modification by sex for the primary endpoint, we performed Cox regression analyses for the association of plasma IGF1 levels with all-cause mortality in which female and male RTR were grouped together, with additional inclusion of an interaction term of plasma IGF1 and sex in the Cox regression model. Hazard ratios (HR) and 95% confidence intervals (CIs) were calculated per 1 unit increment in log2-transformed IGF1 levels. Thereafter, we proceeded with sex-stratified prospective analyses for all-cause mortality as a primary endpoint. In addition to crude analyses, we performed analyses with adjustment for age and eGFR with and without additional physiological, lifestyle, routine clinical chemistry, transplantation, medication, and comorbidity related variables. Subsequently, we performed mediation analyses using the method as described by Preacher and Hayes, which allowed for testing the significance and magnitude of (potential) mediation [33]. In these analyses, mediation was assessed by computing bias-corrected confidence intervals upon running 2000 bootstrap samples. The proportion of mediation was obtained by dividing the indirect effect coefficient by the total effect coefficient, which were adjusted for age and eGFR. Mediation analyses were performed using IGF1 as a potential risk factor and 24 h urinary creatinine excretion, a marker of muscle mass, as potential mediator while also vice versa, because the observational nature of our study does not allow for drawing conclusions on cause–effect relationships. At last, we performed Cox-regression analyses for the association of plasma IGF1 levels with cause-specific mortality as secondary endpoints. Due to lower numbers of events, the exploratory nature of these analyses, and the generally accepted rule of thumb that allows for one variable to be included for each 7–10 events in Cox regression models [34], we performed sex-stratified crude analyses Cox regression analyses and sex-stratified age- and eGFR-adjusted analyses for the separate causes of mortality.

#### **3. Results**

#### *3.1. RTR Characteristics*

Baseline characteristics according to tertiles of plasma IGF1 levels for female and male RTR are shown in Table 1. At baseline, median IGF1 levels were 153 ng/mL (IQR: 118–196) in female and 168 ng/mL (IQR: 128–224) in male RTR (see Figure 1).

Female RTR who had higher IGF1 levels were more likely to have a larger waist circumference and a higher 24 h urinary creatinine excretion. In turn, the prevalence of diabetes mellitus as primary renal disease, the use of insulin therapy, the cumulative prednisolone dose, and the time between transplantation and baseline measurements were lower for these subjects.

For male RTR with higher levels of IGF1, subjects were more likely to be younger, to have a higher body weight and SQUASH score, and to have a lower waist circumference, prevalence of diabetes mellitus as primary renal disease, and cumulative prednisolone dose. Male RTR in the highest tertile of IGF1 levels were furthermore more likely to have received a graft from a living donor, to have undergone dialysis before transplantation, to have a shorter time between transplantation and baseline measurements, to use calcineurin inhibitors, whereas these subjects were less likely to use coumarin derivatives. Lastly, levels of serum creatinine, plasma albumin, plasma triglycerides, and 24 h urinary creatinine excretion were more likely to be higher whereas plasma aspartate transaminase (AST), gamma-glutamyltransferase (GGT), and high sensitivity C-reactive protein (hs-CRP) were more likely to be lower for these subjects.





1 lipoprotein; SQUASH: Short

dose of prednisolone

 until inclusion and the dose of prednisolone

methylprednisolone

 dose to its prednisolone

 dose equivalent).

QUestionnaire

 to ASsess

Health-enhancing

 physical activity [27]. 4 The cumulative dose of prednisolone

> or

methylprednisolone

 required for treatment of acute rejection (a conversion factor of 1.25 was used to convert

 was calculated as the sum of the maintenance

**Figure 1.** Association between insulin-like growth factor 1 (IGF1) levels and age for female and male renal transplant recipients (RTR).

#### *3.2. Association of Plasma IGF1 with Selected Variables in RTR*

Associations between plasma IGF1 levels and variables of interest adjusted for age alone, for age and eGFR, and for multiple variables which were selected following stepwise backward elimination are shown in Table 2.

For female RTR, the analyses adjusted for age featured positive and significant associations between plasma IGF1 and body weight, 24 h urinary creatinine excretion, and calcineurin inhibitor use. Significant inverse associations with plasma IGF1, independent of age, were observed for the prevalence of diabetes mellitus as primary renal disease, the time between transplantation and baseline measurements, the cumulative prednisolone dose, and GGT. After further adjustment for eGFR, the magnitude, direction, and significance of all associations generally remained the same. The final stepwise backward model featured an adjusted R<sup>2</sup> of 0.14 and revealed significant positive associations between plasma IGF1 and both 24 h urinary creatinine excretion and calcineurin inhibitor use, but also significant inverse associations with the prevalence of diabetes mellitus as primary renal disease, GGT, HDL cholesterol, and hs-CRP.

For male RTR, analyses adjusted for age showed significant positive associations between plasma IGF1 and both albumin and calcineurin inhibitor use. Significant inverse associations were observed between plasma IGF1 and eGFR, the prevalence of diabetes mellitus as primary renal disease, the cumulative prednisolone dose, AST, and GGT, and the time between transplantation and baseline measurements. Further adjustment for eGFR did not lead to major changes in the magnitude, direction, or significance of these associations. Lastly, an adjusted R<sup>2</sup> of 0.28 was obtained for the final stepwise backward model which featured significant positive associations between plasma IGF1 and both albumin and calcineurin inhibitor use. Significant inverse associations were furthermore revealed between plasma IGF1 and age, eGFR, the prevalence of diabetes mellitus as primary renal disease, and GGT.



Variables showing *p*-values below 0.10 for the trend of tertiles of IGF1 in at least one of the sexes (see Table 1), with the exception of highly correlated variables (e.g., BMI, waist circumference, serum creatinine), as well as body weight and body height were included for multivariable linear regression analysis. Abbreviations: AST: Aspartate transaminase; eGFR: Estimated glomerular filtration rate; GGT: Gamma-glutamyltransferase; HDL: High-density lipoprotein; hs-CRP: High sensitivity C-reactive protein; SQUASH: Short QUestionnaire to ASsess Health-enhancing physical activity [27]. 2 The cumulative dose of prednisolone was calculated as the sum of the maintenance dose of prednisolone until inclusion and the dose of prednisolone or methylprednisolone required for treatment of acute rejection (a conversion factor of 1.25 was used to convert the methylprednisolone dose to its prednisolone dose equivalent).

1

#### *3.3. Association of Plasma IGF1 with All-Cause Mortality in RTR*

Median follow-up was 5.4 years (IQR: 4.8–6.0 years) for female and 5.4 years (IQR: 4.8–6.3 years) for male RTR. During this prospective follow-up, 56 female and 77 male RTR died. We first investigated whether the association of plasma IGF1 levels with all-cause mortality was modified by sex. In these analyses, with data of female and male RTR combined, we found that higher plasma IGF1 levels were associated with a significantly decreased risk (HR per log2 increment of plasma IGF1, 95% CI) of all-cause mortality (0.61, 0.47–0.80; *p* < 0.001). Furthermore, inclusion of a product-term of (log2-transformed plasma) IGF1 levels and sex in the basic multivariable model (i.e., with adjustment for age and eGFR) revealed the existence of significant effect modification by sex (*p* for interaction = 0.02). After finding this significant interaction by sex, we proceeded with sex-stratified analyses of the association of (log2-transformed) plasma IGF1 levels with all-cause mortality. For female RTR, the crude analyses showed that higher plasma IGF1 levels were associated with a significantly decreased risk of all-cause mortality (0.42, 0.26–0.66; *p* < 0.001; see Figure 2 and Table 3), while a nonsignificant trend towards a decreased risk was observed for male RTR (0.74, 0.52–1.04; *p* = 0.09; see Table 3 and Figure 2).

**Figure 2.** Kaplan–Meier curves for all-cause mortality according to tertiles of plasma insulin-like growth factor 1 (IGF1) in (**a**) female and (**b**) male renal transplant recipients (RTR). For female RTR, IGF1 levels of the tertiles 1, 2, and 3 are below 131 ng/mL, range between 131 and 181 ng/mL, and are above 181 ng/mL, respectively. For male RTR, IGF1 levels of the tertiles 1, 2, and 3 are below 141 ng/mL, range between 141 and 202 ng/mL, and are above 202 ng/mL, respectively.

In the model with adjustment for age and eGFR, the significant inverse association of IGF1 with all-cause mortality remained in female RTR (0.40, 0.24–0.65; *p* < 0.001) and the association in male RTR remained insignificant (0.85, 0.56–1.29; *p* = 0.44). Further adjustment for potential confounders, which was assessed based on seven different multivariable models, did not substantially affect the associations between plasma IGF1 and mortality for both female and male subjects (see Table 3). Lastly, mediation analysis (according to the procedures of Preacher and Hayes [33]) was carried out for the female subjects and revealed 24 h urinary creatinine excretion as significant mediator (*p*-value for indirect effect < 0.05) accounting for 39% on the association between plasma IGF1 and all-cause mortality (see Table 4). Since the observational nature of our study does not allow for drawing conclusions regarding cause–effect relationships, we also performed alternative mediation analyses with 24 h urinary creatinine excretion as potential risk factors and plasma IGF1 levels as potential mediators. In these analyses, we found that plasma IGF1 levels as significant mediators (*p*-value for indirect effect

< 0.05) accounted for 9% on the association between 24 h urinary creatinine excretion and all-cause mortality (see Supplemental Table S5).

**Table 3.** Association between log2-transformed plasma IGF1 levels and the risk of all-cause mortality in female and male RTR 1.


<sup>1</sup> Hazard ratios (HR) per 1 unit increment in log2-transformed plasma IGF1 levels and corresponding 95% confidence intervals (CI) were derived from Cox proportional hazards models. <sup>2</sup> Multivariable model adjusted for age and estimated glomerular filtration rate (eGFR). <sup>3</sup> Multivariable model adjusted for age, eGFR, body length, body weight, waist circumference, systolic blood pressure, and diastolic blood pressure. <sup>4</sup> Multivariable model adjusted for age, eGFR, smoking status, alcohol consumption, and Short QUestionnaire to ASsess Health-enhancing physical activity (SQUASH) score [27]. <sup>5</sup> Multivariable model adjusted for age, eGFR, glucose, glycated hemoglobin (HbA1c), triglycerides, serum total cholesterol, high-density lipoprotein (HDL) cholesterol, and low-density lipoprotein (LDL) cholesterol. <sup>6</sup> Multivariable model adjusted for age, eGFR, serum creatinine, and urine total protein. <sup>7</sup> Multivariable model adjusted for age, eGFR, primary renal disease, graft rejection, dialysis before transplantation, time between transplantation and baseline visit, and donor status. <sup>8</sup> Multivariable model adjusted for age, eGFR, aspartate transaminase (AST), gamma-glutamyltransferase (GGT), serum albumin, high sensitivity C-reactive protein (hs-CRP), follicle-stimulating hormone, and luteinizing hormone. <sup>9</sup> Multivariable model adjusted for age, eGFR, antidiabetics, antihypertensive drugs, coumarin derivatives, proliferation inhibitors, calcineurin inhibitors, insulin, and prednisolone.

**Table 4.** Mediation analysis of the relationship between plasma IGF1, 24 h urinary creatinine excretion, and all-cause mortality in female RTR.


<sup>1</sup> Coefficients and corresponding 95% confidence intervals (CI) of the indirect and total effects are standardized for the standard deviations of the potential mediator, plasma IGF1, and all-cause mortality. <sup>2</sup> Coefficients are adjusted for age and estimated glomerular filtration rate (eGFR). <sup>3</sup> 95% CIs for the indirect and total effects are bias-corrected confidence intervals after running 2000 bootstrap samples. <sup>4</sup> The size of (statistically significant) mediated effects is calculated by dividing the standardized indirect effect by the standardized total effect followed by multiplication by 100.

#### *3.4. Association of Plasma IGF1 with Cause-Specific Mortality in RTR*

Next, we performed sex-stratified analyses of the association of log2-transformed plasma IGF1 levels with mortality from specific causes of death, namely death from infectious diseases, cardiovascular mortality, death from malignancies, and other, miscellaneous causes of death. In females, we found that higher plasma IGF1 levels were strongly associated with a significantly decreased risk of infectious disease-related mortality (0.17, 0.07–0.38; *p* < 0.001; see Supplemental Table S1, Model 1). In females, we also found a borderline significant association of higher plasma IGF1 levels with cardiovascular mortality (0.43, 0.18–1.00; *p* = 0.05; see Supplemental Table S2, Model 1), but neither a significant association with cancer-related mortality (1.50, 0.45–4.93; *p* = 0.51; see Supplemental Table S3, Model 1), nor with mortality from miscellaneous causes (0.43, 0.10–1.78; *p* = 0.24; see Supplemental Table

S4, Model 1). In males, no significant associations with cause-specific mortality were encountered (see respective Supplemental Tables S1–S4).

#### **4. Discussion**

This study showed that low plasma IGF1 levels were independently associated with an increased risk of all-cause mortality in female RTR. Such association was less pronounced and insignificant in male RTR, which should be seen in the context of IGF1 levels being negatively and strongly associated with age in males, which may explain why low plasma IGF1 levels were not associated with mortality in males. Adjustment for potential confounders did not alter the association observed in women, and 39% of this association was found to be mediated by 24 h urinary creatinine excretion, a marker of muscle mass. In alternative analyses, we found that 9% of the association of urinary creatinine excretion with mortality in women was mediated by plasma IGF1 levels. In secondary analyses, in which the association of plasma IGF1 with cause-specific mortality was assessed, we found a particularly strong association of low plasma IGF1 levels with increased risk of mortality due to infectious causes in females.

To our knowledge, this is the first study that investigated the association between IGF1 and long-term outcomes in RTR, hence we were limited in comparing our study with existing literature. Studies addressing associations between IGF1 and outcomes in other clinical settings are available, yet such studies are scarce and generally do not assess female and male subjects separately. When attempting to compare our results to studies on IGF1 in which both sexes were analyzed separately, we found inconsistent evidence. For example, in a cross-sectional study of 5388 US adults, the magnitude of the (positive) association between high IGF1 levels and the risk of chronic kidney disease was found to be stronger for males than for females [36]. In addition, a study of 183 healthy nonagenarians (i.e., people between the age of 90 and 99) reported a significant association between low IGF1 levels and longer survival in female subjects which was not observed for males [11]. Recently, a prospective population-based study on 1618 elderly adults reported that men featured greater decreases in IGF1 and its most important binding protein (i.e., IGF binding protein 3) with age as compared to females [37]. A recently described cross-sectional study on 200 elderly subjects furthermore reported a (negative) association between IGF1 levels and co-existent frailty and low muscle mass in female subjects whereas such association was not found for male subjects [38]. The difference between females and males as we observed in our study therefore links to previous data but also connects to why gender-specific reference ranges for IGF1 are being employed in routine clinical practice [39–41]. It should, however, be noted that all these results were obtained using different analytical methods, and it is known that different methods may yield different analyte levels, particularly in the case of IGF1 [42,43]. Moreover, IGF1 predominantly circulates being bound to IGF binding proteins [44], and the efficiency of dissociating such complexes may vary between (immuno)assays from different vendors and thereby lead to biased, or at least to incomparable results [45].

With respect to the observed association between IGF1 and mortality in female RTR, several other findings which were put forward in our study should be taken into consideration. Firstly, the identification of 24 h urinary creatinine excretion as a strong mediator in this association represents an interesting finding of our study. The fact that 24 h urinary creatinine is a widely available and accepted marker reflecting muscle mass [46–48] and the recognition of IGF1 as a growth hormone involved in muscle growth [49,50] support the biological plausibility of a link between IGF1 and physical fitness. Low physical activity is, in fact, known to be a risk factor for morbidity and mortality in RTR [51–53], hence further studies on IGF1 in this context are warranted. Secondly, the significant association between IGF1 and the use of calcineurin inhibitors should be viewed in this context as well. The target of these drugs, calcineurin, has been described as a regulator of muscle mass, although it should be noted that much is still unknown about the underlying mechanisms [54–56]. Thirdly, the observed strongly significant association of higher plasma IGF1 levels with lower risk for infectious disease-related mortality may be interesting as well in this regard. At last, it should be noted that

evidence for a potential link between IGF1 and physical fitness is currently still circumstantial and that further research is needed to verify and explore our findings.

Important limitations of this study include the facts that it represents a single-center study and that it addresses a population consisting mainly of Caucasian participants. It is unknown whether our findings can be extrapolated to other populations, and repeating this study in other populations is therefore desirable. Moreover, there may be untested or residual confounding relevant for the observed association, as is often true for observational studies. Moreover, laboratory markers were analyzed only once at baseline, hence corresponding changes over time could not be addressed in the present study. With respect to the IGF1 measurements, it should be noted that measurements were carried out using biobanked samples which had been stored for several years at −80 ◦C. Sample stability parameters (e.g., freeze-thaw stability, benchtop stability) were addressed during validation of our IGF1 method thereby following the US Food and Drug Administration (FDA) guidelines on bioanalytical method validation [30]. Nonetheless, storage conditions comparable to those applying to the long-term stored plasma samples could not possibly be addressed during validation, as is often the case when targeting biobanked samples. We could, however, monitor the extent of IGF1 oxidation which represents a prominent feature of our mass spectrometric IGF1 assay [29] considering that protein oxidation is a (unwanted) chemical modification occurring during storage of proteins [57]; yet, no abnormalities in IGF1 oxidation were observed. In order to reduce the (potential) impact of corresponding pre-analytical variability on the quality of our data, we only included samples which had not undergone any previous freeze–thaw cycle and we verified that the samples had not been exposed to deviating storage conditions, for example caused by power outages or freezer malfunctions.

Strengths of this study are its prospective design, the relatively large cohort of well-characterized, stable RTR, the complete follow-up for all-cause mortality, the availability of detailed data on potential confounders, and the use of a mass spectrometric IGF1 assay which allowed for highly selective IGF1 quantification.

In conclusion, low plasma IGF1 levels were found to be associated with an increased risk of all-cause mortality in female RTR, and this association was not found (to be significant) for male RTR. The association in females was mediated for a substantial proportion by 24 h urinary creatinine excretion which hints at a possible link with conditions of low muscle mass (e.g., poor physical fitness, poor nutritional state). Secondary analyses pointed towards a particularly strong association of low plasma IGF1 levels with mortality from infectious causes. Further research is, however, needed to explore the existence and/or relevance of such a link, and also to investigate whether IGF1 can be useful as a (predictive) marker of mortality in female RTR possibly by reflecting physical fitness in this population.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/2/293/s1. Figure S1: Overview of calibration data for the thirteen runs carried out for quantification of insulin-like growth factor 1 (IGF1) in the clinical samples. Figure S2: Overview of the quality control data obtained during the thirteen runs carried out for quantification of insulin-like growth factor 1 (IGF1) in the clinical samples. Table S1: Association between log2-transformed plasma IGF1 levels and the risk of infectious disease-related mortality in female and male RTR. Table S2: Association between log2-transformed plasma IGF1 levels and the risk of cardiovascular mortality in female and male RTR. Table S3: Association between log2-transformed plasma IGF1 levels and the risk of cancer-related mortality in female and male RTR. Table S4: Association between log2-transformed plasma IGF1 levels and the risk of miscellaneous-cause mortality in female and male RTR. Table S5: Mediation analysis of the relationship between 24 h urinary creatinine excretion, plasma IGF1 levels, and all-cause mortality in female RTR.

**Author Contributions:** F.K., N.H.T.t.H., P.H., R.B., and S.J.L.B.; Methodology, F.K., L.M.K., A.W.G.-N., S.P.S., N.H.T.t.H., I.P.K., A.P.v.B., E.v.d.B., P.H., R.B., and S.J.L.B.; Formal analysis, F.K., L.M.K., P.H., and S.J.L.B.; Data curation, F.K., L.M.K., A.W.G.-N., S.P.S., E.v.d.B., and S.J.L.B.; Writing—original draft preparation, F.K., L.M.K., R.B., and S.J.L.B.; Writing—review and editing, F.K., L.M.K., A.W.G.-N., S.P.S., N.H.T.t.H., I.P.K., A.P.v.B., E.v.d.B., P.H., R.B., and S.J.L.B.; Supervision, R.B. and S.J.L.B.; Funding acquisition, N.H.T.t.H., R.B., and S.J.L.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Netherlands Organisation for Scientific Research NWO domain Applied and Engineering Sciences (Perspectief program P12-04, project 13541) and the Top Institute Food and Nutrition (program A-1003).

**Acknowledgments:** The authors gratefully acknowledge the Dutch Biomarker Development Center (BDC; http: //www.biomarkerdevelopmentcenter.nl/) for support of this work.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **Ischemia and Reperfusion Injury in Kidney Transplantation: Relevant Mechanisms in Injury and Repair**

**Gertrude J. Nieuwenhuijs-Moeke 1,\*, Søren E. Pischke 2, Stefan P. Berger 3, Jan Stephan F. Sanders 3, Robert A. Pol 4, Michel M. R. F. Struys 1,5, Rutger J. Ploeg 4,6 and Henri G. D. Leuvenink <sup>4</sup>**


Received: 19 December 2019; Accepted: 13 January 2020; Published: 17 January 2020

**Abstract:** Ischemia and reperfusion injury (IRI) is a complex pathophysiological phenomenon, inevitable in kidney transplantation and one of the most important mechanisms for non- or delayed function immediately after transplantation. Long term, it is associated with acute rejection and chronic graft dysfunction due to interstitial fibrosis and tubular atrophy. Recently, more insight has been gained in the underlying molecular pathways and signalling cascades involved, which opens the door to new therapeutic opportunities aiming to reduce IRI and improve graft survival. This review systemically discusses the specific molecular pathways involved in the pathophysiology of IRI and highlights new therapeutic strategies targeting these pathways.

**Keywords:** ischemia reperfusion injury; kidney transplantation; delayed graft function; innate immune system; adaptive immune system; apoptosis; necrosis; hypoxic inducible factor; endothelial dysfunction

#### **1. Introduction**

To date, 10% of the worldwide population suffers from chronic kidney disease (CKD). The prevalence of the disease will most likely grow over the next decade due to the increase in the elderly population and the growing incidence of diabetes and hypertension. In 2015, CKD was ranked 12th in the global list of causes of death [1]. The population of patients needing renal replacement therapy (RRT) worldwide was estimated to be approximately 4.902 million (95% CI 4.438–5.431 million) in a conservative model and 9.701 million (95% CI 8.544–11.021 million) in a high estimate model, illustrating the magnitude of the disease burden of end stage renal disease (ESRD) [2].

For patients with ESRD, transplantation is still the optimal treatment. Long-term survival with kidney transplantation is dramatically better than dialysis and transplantation provides a sustainably higher quality of life. Unfortunately, there is a worldwide shortage of suitable donor organs for (kidney)

transplantation. The number of renal transplantations performed worldwide in 2018 was 75.664 [3]. Due to the persistent shortage of donor kidneys, many transplant centres have established large living donor programmes and transplant teams are also now accepting increasing numbers of older and higher risk organs, retrieved from deceased donors. The use of these extended criteria donors (ECD) has affected outcomes after transplantation due to an often-suboptimal quality of the donor organ [4,5]. As we will face more complex donors in the future with a reduced viability such as unstable donation after brain death (DBD) donors, donation after circulatory death (DCD) donors, and ECD, the challenge in transplantation is to be able to use these donor sources, however, without compromising successful immediate function and long-term graft survival after transplantation. It is therefore imperative that the condition of every graft-to-be is optimised prior to or at the time of transplantation and that additional injury is minimized in order to achieve the best possible post-transplant function and avoid primary non function (PNF), delayed graft function (DGF), and rejection with chronic graft failure.

Ischemia and reperfusion injury (IRI) is inevitable in (kidney) transplantation and one of the most important mechanisms for non- or delayed function immediately after transplantation [6–8]. It is accompanied by a proinflammatory response and is associated with acute rejection due to an increased immunogenicity favouring T-cell mediated rejection as well as anti-body mediated rejection (ABMR) [9,10]. In addition, it may result in progressive interstitial fibrosis and is associated with chronic graft dysfunction due to interstitial fibrosis and tubular atrophy (IFTA) [11]. In the past decade more insight has been gained in the complex molecular pathophysiology of IRI. This may open a door to new therapeutic targets aiming to reduce IRI. The aim of this review is to systematically highlight these molecular mechanisms and to discuss potential therapeutic strategies specifically targeting these molecular pathways.

#### **2. Ischemia and Reperfusion Injury**

IRI consists of a complex pathophysiology involving activation of cell death programs, endothelial dysfunction, transcriptional reprogramming and activation of the innate and adaptive immune system [8]. Numerous pathways and signalling cascades are implicated (Figure 1) and it is while worthy to dissect the distinct effects of ischemia and reperfusion (I/R).

**Figure 1.** Schematic overview of the pathophysiological consequences of ischemia and reperfusion. I/R: ischemia/reperfusion; ATP: adenosine triphosphate; EndMT: endothelial to mesenchymal transition; ROS: reactive oxygen species; mPTP: mitochondrial permeability transition pore.

#### *2.1. Ischemia*

Due to a decrease in oxygen supply, cells will switch from an aerobic to an anaerobic metabolism, which results in a decrease in adenosine triphosphate (ATP) production and intracellular acidosis due to the formation of lactate. This causes destabilisation of lysosomal membranes with leakage of lysosomal enzymes, breakdown of the cytoskeleton and inhibition of membrane-bound Na+/K<sup>+</sup> ATPase activity [12–14]. This last process gives rise to an intracellular accumulation of Na<sup>+</sup> ions and water with as a consequence cellular oedema. Due to declined Ca2<sup>+</sup> excretion, there is an intracellular Ca2<sup>+</sup> accumulation, which causes activation of Ca2<sup>+</sup> dependant proteases like calpains. Due to the acidosis, these calpains stay inactive during the ischemic period but may damage the cell after normalisation of the pH during reperfusion. The remaining ATP is broken down to hypoxanthine, which will accumulate in the cell, since further metabolism into xanthine requires oxygen [15]. In the mitochondria, the Ca2<sup>+</sup> overload is responsible for generation of reactive oxygen species (ROS) [8]. This will lead to opening of the mitochondrial permeability transition pores (mPTP) after reperfusion. During the ischemic period, only small amounts of ROS are produced compared to the entire I/R due to the reduction of cytochromes, nitric oxide synthases, xanthine oxidase and reduced nicotinamide adenine dinucleotide phosphate (NADPH) oxidase activation [16–19].

#### *2.2. Reperfusion*

During reperfusion, oxygen levels increase, and the pH normalises which is harmful for the previously ischemic cells. The intracellular Ca2<sup>+</sup> level further increases, which activates the calpains causing injury to the cell structure and cell death [8]. Restoration of normoxemia leads to the production of large amounts of ROS, together with a reduction in the antioxidant capacity [20]. This burst of ROS production was thought to be due to a generalised dysregulation of the electron transport chain with electrons leaking out at non-specific sites [21]. Recently, however, Chouchani et al. [22] showed that this superoxide production is generated by reverse action of complex I of the electron transport chain driven by a pool of succinate, a metabolite of the citric acid cycle, accumulated during ischemia. This massive amount of mitochondrially produced ROS is responsible for the activation of various injurious pathways through carbonylation of proteins or lipid peroxidation. This may contribute to injury of the cell membranes, the cytoskeleton and DNA and may lead to a disruption of ATP generation and induction of mPTP [20]. Additionally, the combination of ROS, dysfunctioning of the mitochondrial machinery and increase in mitochondrial Ca2<sup>+</sup> load causes opening of the mPTP and release of substances like cytochrome C, succinate and mitochondrial DNA (mtDNA), which are able to induce cell death through apoptosis and necrosis and may act as danger/damage associated molecular patterns (DAMPs) entailing activation of the innate and subsequently the adaptive immune system [23–26].

Recent insights in the pathophysiological mitochondrial mechanisms and general understanding of the pivotal role of the mitochondria in IRI has led to various strategies targeting mitochondria with the aim to reduce IRI including limiting oxidative stress and mitochondrial ROS generation [20]. Both lipophilic cations and mitochondrial targeted proteins have been developed to deliver antioxidants to the mitochondria [27]. Triphenylphosphonium (TTP), a lipophilic cation, is rapidly taken up by mitochondria where it releases covalently bonded bioactive compounds. MitoQ, with its bioactive compound ubiqinone, is the most investigated of these molecules. In the mitochondria ubiqinone is reduced to ubiquinol, a powerful ROS scavenger. Administration of MitoQ in renal I/R models resulted in reduced markers of oxidative stress, reduced renal injury and improved function [28–30]. Regarding the mitochondrial targeted proteins, the Szeto-Schiller (SS) proteins are the best known. Exact mechanism of action is poorly understood but a possible explanation of action is through interaction with cardiolipin, an important component of the inner mitochondrial membrane. SS peptides have shown to reduce renal IRI in rodents [31], and its lead compound SS-31 (Elamipretide, Stealth BioTherapeutics-Alexion Pharmaceuticals) is currently being investigated in humans for its efficacy in reducing IRI post-angioplasty for renal artery stenosis. A pilot study administration of SS-31 before and during percutaneous transluminal renal angioplasty and stenting has shown to attenuate post-procedural hypoxia, increased renal blood flow and improved kidney function [32].

Another strategy to reduce ROS generation is reduction of succinate formation by inhibition of succinate dehydrogenase, preventing the accumulation of succinate, a driving force of reverse action of complex I. This has been shown to be effective in various in vivo models of IRI including the heart but has yet been unexplored in renal IRI [22,33].

#### **3. Pathophysiological Consequences of IRI**

#### *3.1. Cell Death: Necrosis, Apoptosis, Regulated Necrosis and Autophagy*

#### 3.1.1. Necrosis

I/R leads to the activation of cell death programs. Of these programs, necrosis is the most uncontrolled form. It is due to swelling of the cell and subsequent rupture of the cellular membrane [34]. This will lead to an uncontrolled release of cellular fragments into the extracellular space. These fragments act as DAMPs and are able to activate the innate and adaptive immune system, entailing infiltration of inflammatory cells into the tissue and release of different cytokines.

#### 3.1.2. Apoptosis

In contrast to the uncontrolled process of necrosis, apoptosis is a highly regulated and controlled process in which activation of the caspase signalling cascade results in a self-limiting programmed cell death. These caspases, a family of proteases, are essential in this process. There are two types of caspases: initiator caspases (2,8,9,10) and effector caspases (3,6,7) [35,36]. The initiator caspases are activated by binding to a specific activator protein complex (death-inducing signalling complex (DISC), apoptosome) [37]. The formed complexes then activate the effector caspases through proteolytic cleavage upon which these proteolytically degenerate various intracellular proteins. Apoptosis gives rise to apoptotic bodies, containing these intracellular protein fragments, via the process of membrane blebbing. The apoptotic bodies will undergo phagocytosis before they can spill their content into the extracellular space and therefore will generate a less immune stimulating impulse compared to necrosis. Apoptosis can be initiated through the intrinsic pathway (mitochondrial dependent pathway) in which the initiating signal comes from within the cell (e.g., damaged DNA, hypoxia, metabolic stress) or the extrinsic pathway (cell death receptor pathway) due to signals from out of the cell (tumor necrosis factor-α (TNF-α), first apoptosis signal (Fas)-ligand, FasL) (Figure 2) [37].

A protein family playing an important role in the regulation of apoptosis is the B-cell lymphoma 2 (BcL-2) family [38]. Members of this family can act as protectors (BcL-2, BcL-xL) inhibiting apoptosis, sensors (BH3 only proteins, Bad, Bim, Bid) inhibiting the protectors, or effectors (Bax, Bad) initiating apoptosis by enhancing the permeability of the mitochondrial membrane [39]. In case of intrinsic signalling, intracellular signals of cell stress will lead to an increase in the BH3 only proteins resulting in inhibition of the protectors and activation of the effectors. These effectors increase the permeability of the mitochondrial membrane resulting in leakage of pro-apoptotic proteins upon which a caspase activator complex, the apoptosome, is formed in the intracellular space [40–43]. The apoptosome cleaves procaspase-9 to its active form of caspase-9, which in turn is able to activate the effector caspase-3. In case of the extrinsic signalling, binding of TNF-α (TNF path) or the FasL, expressed on cytotoxic T lymphocytes, (Fas path) to receptors of the TNF receptor (TNFR) family will lead to the formation of a complex called the death-inducing signalling complex (DISC) [44–46]. The DISC, amongst others, consisting of a death effector domain and three procaspase-8 or -10 molecules, cleaves and activates the procaspases [47]. Activation of the initiator caspase-8 by both paths directly activates other members of the caspase signalling cascade such as the effector caspase-3 but also can lead to an increase in BH3-only proteins (Bim, Bid) and trigger the intrinsic pathway (Figure 2) [48].

**Figure 2.** Extrinsic and intrinsic apoptotic pathway. The intrinsic pathway is mediated by intracellular signals of cell stress leading to an increase in the BH3 only proteins (members of the B-cell lymphoma 2 (Bcl-2) family) resulting in an inhibition of the protectors and activation of the effectors. The effectors Bax and Bad increase the permeability of the mitochondrial membrane (MOMP: mitochondrial outer membrane permeabilisation) resulting in leakage of apoptotic proteins. One of these proteins, known as second mitochondria-derived activator of caspases (SMAC), binds to proteins that inhibit apoptosis (IAPs, by suppression of the caspase proteins) causing an inactivation of these IAPs. Another protein released from the mitochondria is cytochrome c, which binds to Apoptotic protease activating factor-1 (Apaf-1) and ATP. This complex binds to procaspase-9 creating a complex, the apoptosome. The apoptosome cleaves procaspase-9 to its active form of caspase-9, which in turn is able to activate the effector caspase-3. The extrinsic pathway is mediated through receptors of the tumor necrosis factor (TNF) receptor (TNFR) family either via the TNF path or the Fas (first apoptosis signal) path. In the TNF path binding of TNF-α to a trimeric complex of TNFR1 molecules induces activation of the intracellular death domain and the formation of the receptor-bound complex 1 made up of TNF receptor-associated death domain (TRADD), receptor-interacting protein kinase 1 (RIPK1), two ubiquitin ligases (TNFR-associated factor (TRAF)-2 and cellular inhibitors of apoptosis (clAP)1/2) and the linear ubiquitin assembly complex (LUBAC). This complex 1 can lead to a pro-survival pathway or to apoptosis. In case of apoptosis the TRADD dependant complex IIa (consisting of TRADD, Fas-associated death domain protein (FADD) and caspase-8) or the RIPK-1 dependant complex IIb also known as the ripoptosome (consisting FADD, RIPK1, RIPK3 and caspase-8) is formed. In the Fas path, presence of the Fas ligand (FasL, expressed on cytotoxic T lymphocytes) causes three Fas receptors (CD95) to trimirize. This clustering and binding to the FasL initiates binding of FADD. Three procaspase-8 or -10 molecules can then interact with the complex by their own death effector domains. The complex formed is the death-inducing signalling complex (DISC) which cleaves and activates procaspase-8 and 10. Activation of the initiator caspase-8 by both paths directly activates other members of the caspase signalling cascade such as the effector caspase-3 but also can lead to an increase in BH3-only proteins (Bim, Bid) and trigger the intrinsic pathway).

#### 3.1.3. Regulated Necrosis

Recently, new pathways of a more regulated form of necrosis have been described. These processes show features of apoptosis as well as necrosis. One of the best-known pathways of regulated necrosis is via TNFR-1 and is called necroptosis [46]. In the absence of active caspase-8, phosphorylation of receptor-interacting protein kinase 1 (RIPK1) and RIPK3 in complex IIb leads to formation of a complex called the necrosome. The necrosome recruits Mixed Kinase Domain-Like protein (MLKL), which is then phosphorylated by RIPK3 [46]. MLKL activates the necrosis phenotype by entering the bilipid membranes of organelles and the cellular membrane. This causes formation of pores in these membranes and leads to release of cellular contents, functioning as DAMPs, into the extracellular space [49]. As in necrosis the DAMPs are able to activate both the innate and adaptive immune system promoting proinflammatory responses that activate rejection pathways [50,51]. A recent study in a kidney transplant mouse model showed that RIPK3-deficient kidneys had better function and longer rejection-free survival [52]. Therefore RIPK3-inhibiting drugs might be of interest in the reduction of IRI in organ transplantation. Next to TNFR-1, other death receptors and toll like receptors (TLR) have also shown to be able to induce necroptosis [46]. Other forms of regulated necrosis include mitochondrial permeability transition (MPT)-associated death (involving opening of mPTP leading to necrosis instead of apoptosis), ferroptosis (involving iron and gluthation metabolism), parthanatos (also known as PARP-1 (Poly(ADP-ribose) polymerase-1) dependent cell death, involving the accumulation of PAR (poly(ADP-ribose)) and the nuclear translocation of apoptosis-inducing factor (AIF) from mitochondria) and pyroptosis (involving caspase-1 and -11 in mice and caspase-4 and -5 in humans) [53]. The role of pyroptosis in IRI in the kidney, however, is unclear.

#### 3.1.4. Autophagy

Cells can preserve their metabolic function and escape cellular death. This is due to autophagy of damaged cell parts. There are several pathways of autophagy, namely, macro-autophagy, micro-autophagy and chaperone-mediated autophagy—the last two are beyond the scope of this review. Macro-autophagy (hereafter called autophagy) involves formation of autophagosomes containing damaged cell parts or unused proteins. These double membrane autophagosomes travel through the cytoplasm to fuse with lysosomes (autolysosome) leading to degradation of the damaged cell parts. This process is continuously active at low basal levels, preserving cellular homeostasis, but stimulated upon stress through various signals like nutrient deprivation, ROS formation, hypoxia, free amino acids, etc. [54–56]. Cellular building blocks obtained from recycling of damaged cell parts by autophagy may serve as anti-stress responses and energy source promoting cell survival.

The first step in autophagy, the initiation, is regulated by two kinases: mammalian target of rapamycin complex 1 (mTOR, mTORC1) and adenosine monophosphate-activated protein kinase (AMPK) [54,57,58]. Together, they regulate the activity of the Unc-51 like autophagy activating kinase 1/2 (ULK1/2) complex [59,60]. Activation of mTOR leads to the phosphorylation of this complex and inhibition of autophagy (for instance, through the phosphatidylinositol 3-kinase (PI3K)/Protein kinase B (AKT) or the mitogen-activated protein kinase (MAPK)/extracellular signal–regulated kinase (Erk) 1/2 signalling pathway). On the other hand, activation of AMPK, upon intracellular AMP increase, activates autophagy [61]. This occurs by inhibition of the mTORC1 through dissociation of mTORC1 from ULK1/2 (indirect) or in a direct way by phosphorylation of ULK1/2 forming the ULK1/2-complex [62,63]. Next to the ULK1/2 complex, inducible beclin-1 complex (or class III PI3K complex) is involved in initiation of autophagy. This complex is activated by the ULK-1/2 complex and inhibited by Bcl-2 and Bcl-XL. The ULK1/2 and class III PI3K complexes join to form the phagopore and eventually the autophagosme which will fuse with a lysosome [64–69]. The content of this formed autolysosome is degenerated, and the components are released to be reused to synthesise new proteins or to function as an energy source for the cell (Figure 3) [70].

In renal IRI, autophagy is considered a doubled-edged sword. Upon I/R, it is mostly upregulated, but both protective and harmful effects are observed, proposing a dual role for autophagy in renal IRI [71,72]. Decuypere et al. [71] hypothesize that autophagy can switch roles depending on the severity of the ischemic injury. The exact mechanism behind this switch is unclear but may depend on the survival vs death properties of beclin1 and its interaction with the Bcl-2 family proteins [71,73]. Autophagy can be considered a protective mechanism in (oxidative) stress injured cells through restoring cellular homeostasis. Kidneys from older donors are at increased risk of DGF. The age-dependent decline in autophagy activity and age-dependant autophagic dysfunction may be one of the underlying mechanisms of this phenomenon [74]. Extensive oxidative stress (amount or duration), however, may have detrimental effects which eventually could trigger the switch to aggravation of the injury through autophagy dependant cell death. Excessive or prolonged ROS exposure may lead to the oxidative modification of macromolecules making them only partially degradable by the autolysosome [75]. Furthermore, an energy dependent process of autophagy could deprive the cell of necessary energy. In this light, excessive autophagy seen after prolonged cold ischemia time in particular in DCD donors seems to be one of the underlying mechanisms behind augmentation of reperfusion injury seen in these circumstances, thereby increasing the risk of DGF [71,76]. Based on this dual role of autophagy in renal IRI and transplantation the goal would be to restrict autophagy levels within a protective window. Upon severe ischemia (prolonged cold ischemia time (CIT)) autophagy inhibitors most likely outweigh the activators [71]. Continuing efforts have to be made to elucidate the mechanism of autophagic transition from protective to harmful function.

**Figure 3.** Pathways of macro-autophagy. Initiation of autophagy is regulated by mTORC1 (mammalian target of rapamycin complex 1) and AMPK (AMP-activated kinase). Together, they regulate the activity of the ULK1/2 complex consisting of ULK1/2 (Unc-51 like autophagy activating kinase), FIP200 (FAK family kinase interacting protein of 200 kDa) and the autophagy related proteins (ATG) ATG13 and ATG10. Activation of mTOR leads to the phosphorylation of this complex and inhibition of autophagy (for instance, through the phosphatidylinositol 3-kinase (PI3K)/ Protein kinase B (AKT) or the mitogen-activated protein kinase (MAPK)/ extracellular signal–regulated kinase (Erk) 1/2 signalling pathway) whereas activation of AMPK activates autophagy. AMPK, activated upon intracellular AMP increase, is able to activate autophagy by inhibition of the mTORC1 through dissociation of mTORC1 from ULK1/2 allowing ULK1/2 to be activated. AMPK, is also able to initiate autophagy in a direct way by phosphorylation of ULK1/2 forming the ULK1/2-complex.Another complex involved in the initiation is the autophagy inducible beclin-1 complex (or class III PI3K complex) which consists of Vps34 (phosphatidylinositol 3-kinase), beclin-1 (a BH3 only domain protein member of the Bcl-2 family), vps15 and ATG14. This complex is activated by the ULK-1 complex and inhibited by Bcl-2 and Bcl-XL. The ULK1/2 and class III PI3K complexes join to form the phagopore and eventually the autophagosme. This process is mediated by the ATG5-ATG12-ATG16 complex and the formation of phosphatidylethanolamine-conjugated Light Chain (LC) 3 (LC3-II) facilitating elongation of the bilipid membrane to form a closed autophagosme. The autophagosome fuses with a lysosome and the content of the autolysosome is degenerated and the components are released to be reused to synthesise new proteins or to function as an energy source for the cell. PDK-1: pyruvate dehydrogenase kinase-1.

The different cell death programs described above are induced in response to common stimuli. Several proteins in the autophagy and apoptosis pathway are shared resulting in an intimate crosstalk between apoptosis and autophagy. Regulation of these proteins determines cellular fate to cell survival or cell death. Caspase-mediated degradation of several autophagy regulation proteins limits autophagosome formation and therefore autophagy [77–79]. Apoptosis inhibitors Bcl-2 and Bcl-XL also inhibit autophagy by binding to Beclin-1 limiting its availability to form the classIII PI3K complex [80,81]. Inhibition of cisplatin induced autophagy enhanced caspase-3 activation and apoptosis in renal proximal tubular cells [82,83]. On the other hand, overexpression of ATG5 and beclin-1 prevented cisplatinum induced caspase activation and apoptosis [84]. Additionally, there is evidence that autophagy induction regulates necroptosis. Inhibition of autophagy has shown to prevent necroptosis and vice versa inhibition of necroptosis is able to supress autophagy [85,86].

#### 3.1.5. Targeting Cell Death Programs

Targeting pathways of cell death programs to reduce IRI seems very attractive, since it directly preserves cellular function. Secondly, dead cells releasing DAMPs elicit a strong immune response not only in the organ exposed to I/R but also in other organs of the individual, so called remote organ injury. Therefore, interfering with this process might be immunosuppressive and organ protective. The relative contribution of each of the cell death programs to IRI and outcome in transplantation, however, has to be elucidated.

Nydam et al. [87] showed in a syngeneic mouse transplant model that administration of the pan-caspase inhibitor Q-VD-OPh during graft retrieval and cold preservation resulted in decreased caspase-3 expression and activity, reduced apoptosis in renal tubular cells and improved renal function post-transplantation. The pro-apoptotic gene p53 is activated upon hypoxia, oxidative stress and DNA damage and is able to induce cell cycle arrest, which enables DNA-repair proteins to repair the sustained injury. However, in case of severe DNA damage it induces apoptosis by initiating the intracellular pathway.

Inhibition of P53 in proximal tubular cells has been shown to decrease apoptotic cell death and provide protection against IRI [88,89]. QPI-1002 is a synthetic small interfering ribonucleic acid (siRNA) designed to reversibly and temporarily inhibit p53. In pre-clinical models it has been shown that QPI undergoes rapid glomerular filtration and uptake by proximal tubular epithelial cells [89]. Administration of QPI-1002 has shown to be safe in humans. Two phase I dose escalating safety and pharmacokinetics studies in patients undergoing major cardiovascular surgery (NCT00554359, NCT00683553) has been executed without dose-limiting toxicities or safety issues. A phase I/II study has been executed to evaluate QPI-1002 for the prevention of DGF in recipients of kidneys from deceased donors (NCT00802347) in which treatment with QPI-1002 resulted in lower incidence and severity of DGF [90]. Recently, a phase 3 randomized, double-blind, placebo-controlled study in recipients (*n* = 594) of (older) DBD donor kidneys (>45 years) has been completed (NCT02610296, ReGIFT-study). Results have not been reported yet.

Various pharmacological substances like necrostatins (RIPK1 inhibitors, necroptosis), ferrostatins (ferroptosis), sanglifehrin A (MPT-associated death) and olaparib (parthanatos) and many others have been developed to target specific key molecules of the different programs of regulated necrosis and are currently tested in various animal and disease models (Figure 4) [91,92]. The question remains how safe it will be to inhibit non-apoptocic cell death pathways in patients, since these pathways also function as a backup system when apoptosis fails or is inhibited for instance, by caspase inhibitor expressing viruses. Of these molecules, RIPK1 inhibitors have now entered clinical trials and their safety is being tested in healthy volunteers [93,94].

**Figure 4.** Programs of regulated necrosis and their inhibitors. RIPK1: receptor-interacting protein kinase 1; RIPK3: receptor-interacting protein kinase 3; MLKL: Mixed Kinase Domain-Like protein; MPT: mitochondrial permeability transition; mPTP: mitochondrial permeability transition pore; RN: regulated necrosis; CsA: cyclosporin A; PARP1: poly (ADP-ribose) polymerase-1; AIF: apoptosis-inducing factor.

#### *3.2. Endothelial Dysfunction*

At a vascular level, I/R leads to swelling of the endothelial cells (ECs), loss of the glycocalyx and degradation of the cytoskeleton. As a consequence, intercellular contact of endothelial cells is lost, increasing vascular permeability and fluid loss to the interstitial space [95]. Furthermore, the endothelium will produce vasoactive substances like platelet-derived growth factor (PDGF) and Endothelin-1 (ET-1), causing vasoconstriction [96]. This vasoconstriction can be enhanced by a reduced nitric oxide (NO) production during reperfusion due to decreased endothelial nitric oxide synthase (eNOS) expression and increased sensitivity of the arterioles for vasoactive substances like angiotensin II, thromboxane A2 and prostaglandin H2 [97–99]. Eventually this can lead to the so called no reflow phenomenon characterized by the absence of adequate perfusion on microcirculatory level despite reperfusion.

The regenerative capacity of ECs in peritubular capillaries is limited and injury to the microcirculation may lead to permanent peritubular capillary rarefaction [100,101]. Chronic hypoxia in these regions may induce transcription of fibrogenic genes like transforming growth factor-β (TGF-β) and connective tissue growth factor (CTGF) together with an accumulation of α-smooth muscle actin (α-SMA) [101]. In the end, this may lead to development of IFTA, a process which has mainly been attributed to resident fibroblasts. More recently, however, the role of endothelial-to-mesenchymal transition (EndMT) in this process has been described [102,103]. During EndMT, ECs lose their endothelial phenotype (such as expression of specific endothelial markers like Von Willebrand factor (VWF)) and acquire the phenotype of multipotent mesenchymal cells (MSC). These cells show an increased expression of α-SMA, neuronal (N)-cadherin, vimentin and fibroblast-specific protein-1 and exhibit enhanced migratory potential and increased extracellular matrix production [104–106]. In a porcine I/R model Curci et al. [102] showed that 20%–30% of the total α-SMA+ cells emerging after IRI were also CD31+ suggesting a different origin compared to resident activated fibroblasts. Man et al. [107] showed that in kidney transplant recipients experiencing IFTA and allograft dysfunction, progression of EndMT plays an important role. EndMT is controlled by complex signalling pathways

and networks. In their porcine I/R model, Curci et al. [102] showed a critical role of complement in this process. Kidneys of pigs treated with recombinant C1 inhibitor (C1-INH) showed preserved EC density, significant reduction of α-SMA expression and limited collagen deposition 24 h after I/R compared to untreated pigs. The ECs in the treated pigs showed preserved physiological conformation and position tight to the basal layer of the vessels. The number of transitioning ECs was significantly lower in the treated animals. In an additional in vitro experiment activating ECs with the anaphylatoxin C3a, they showed that C3a induced down regulation of the expression of VWF whilst upregulating α-SMA, by activating the Akt pathway. Activation of the ECs with C5a showed a similar response [102]. Targeting signalling pathways in EndMT in kidney transplantation could be of interest to reduce IFTA and enhance long-term graft survival. More insight however has to be gained to the exact role of EndMT in renal transplantation and what suitable targets to aim for. Furthermore, since EndMT gives rise to multipotent MSC this placidity could be of interest to push these MSCs in the direction of regeneration rather than fibrosis.

An important feature of IRI is the chemotaxis of leukocytes, endothelial adhesion and transmigration of these cells into the interstitial compartment [108]. This process is initiated by increased expression of P-selectin on the endothelial cells and interaction of P-selectin with P-selectin glycoprotein 1 (PSGL-) expressed on the leukocytes. This interaction results in rolling of the leukocytes on the endothelium. Subsequently, firm adherence of the leucocytes to the endothelium is achieved by the interaction of the β2-integrins lymphocyte function-associated antigen 1(LFA-1) and macrophage-1 antigen (MAC-1 or complement receptor 3, CR3) on the leukocyte and the intracellular adhesion molecule 1 (ICAM-1) on the endothelial cells. Platelet endothelial cell adhesion molecule 1 (PECAM-1) thereafter facilitates transmigration into the interstitial space. Once activated, these leukocytes will release several toxic substances like ROS, proteases, elastases and different cytokines in the interstitial compartment which will result in further injury like increased vascular permeability, oedema, thrombosis and parenchymal cell death (Figure 5) [109].

**Figure 5.** Interaction of leukocytes and endothelial cells in the process of transmigration of leukocytes. The increased expression of P-selectin on the endothelial cells upon I/R facilitates interaction with P-selectin glycoprotein 1 (PSGL-) expressed on the leukocytes. This results in rolling of the leukocytes on the endothelium. Subsequently, firm adherence of the leucocytes to the endothelium is achieved by interaction of lymphocyte function-associated antigen 1(LFA-1) and macrophage-1 antigen (MAC-1 or complement receptor 3, CR3) on the leukocyte and the intracellular adhesion molecule 1 (ICAM-1) on the endothelial cells. Finally, platelet endothelial cell adhesion molecule 1 (PECAM-1) facilitates transmigration of the leukocytes into the interstitial space. Once activated, these leukocytes will release several toxic substances like ROS, proteases, elastases and different cytokines in the interstitial compartment resulting in further injury like increased vascular permeability, oedema, thrombosis and parenchymal cell death.

#### *3.3. Innate and Adaptive Immune Response*

IRI is accompanied by sterile inflammation in which the innate as well as the adaptive immune system are involved.

#### 3.3.1. Innate Immune Response

The innate, or non-specific, immune system is evolutionary the oldest part of the immune system. It acts on infection or injury with a fast, short-lasting and non-specific response in which different cells and systems are involved.

#### Toll-Like Receptor Signalling

In the innate immune response, the toll-like receptors (TLRs) play an important role [110]. TLRs are transmembrane proteins and members of the interleukin-1 receptor (IL-IR) superfamily. They function as pattern recognition receptors (PRR) and are present on the cellular membrane and in the cytosol of cells like leukocytes, endothelial cells and tubular cells [111]. The human TLR family contains 10 members, TLR1–TLR10 [112]—of which, TLR2 and TLR4 have shown to be upregulated in tubular epithelial cells upon ischemia [113–117]. Both are attributed an equal importance in initiating apoptosis in a genetic knock-out renal I/R mouse model [115]. TLR activation leads to the downstream recruitment of various adapter molecules (TNF receptor-associated factor 6 (TRAF6), Myeloid differentiation primary-response protein 88 (MyD88), toll-interleukin 1 receptor (TIR) domain containing adaptor protein (TIRAP), TIR-domain-containing adapter-inducing interferon-β (TRIF), TRIF-related adaptor molecule (TRAM)) activating different kinases (IL-1 receptor-associated kinase (IRAK)-1 (IRAK-1), IRAK-4, inhibitor of nuclear factor-κB kinase (IKK), TANK-binding Kinase-1 (TBK1)), leading to activation of transcription factors (nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), IFN-regulatory factor 3 (IRF3) resulting in transcription of proinflammatory genes and the subsequent inflammatory response [8,112].

TLR2 and TLR4 have polyvalent ligand binding activity and can be activated by exogeneous (e.g., lipopolysaccharide, LPS) and endogenous ligands comprising DAMPs released upon I/R. These DAMPs vary depending on type of injury and tissue involved. High-mobility group box-1 (HMGB-1), an intracellular protein involved in the organisation of DNA and the regulation of gene transcription, is one of the DAMPs linked to the pathogenesis of IRI [118–120]. From the nucleus, HMGB-1 can be released into the cytosol or extracellular space by passive leakage from injured cells or through active secretion by immune cells [121,122].

In IRI in the kidney, TLR4 plays an important role. Bergler et al. [123] showed that TLR4 is highly upregulated after renal IRI, and that high TLR4 expression is strongly correlated with graft dysfunction in an allogenic renal transplant model in rats. Furthermore, TLR4-deficient mice are protected against renal IRI and kidneys from donors with a TLR4-loss of function allele show less pro inflammatory cytokines in the kidney after transplantation and a higher percentage of immediate graft function [118,124]. Activation of TLR4 in renal IRI has various consequences on the graft. First of all it promotes the release of different proinflammatory mediators like IL-6, IL-1β and TNF-α, accompanied by an increased expression of macrophage inflammatory protein-2 (MIP-2) and monocyte chemo attractant protein-1 (MCP-1) involved in the recruitment of neutrophils and macrophages [124]. Second, TLR-4 activation leads to increased expression of adhesion molecules ICAM-1, vascular cell adhesion molecule 1 (VCAM-1) and E-selectin facilitating leukocyte migration and infiltration into the interstitial space. TLR-4 signalling seems mandatory for this increased expression. Chen et al. [125] showed that increased expression of adhesion molecules after renal IRI was absent in TLR4 knockout mice in vivo and the addition of HMGB-1 to isolated endothelial cells increased adhesion molecule expression on cells from wild-type but not from TLR4 knockout mice. Thirdly, activation of TLR4 on circulating immune cells of the innate immune system leads to activation of these cells. Neutrophils and macrophages are involved in an early stage after reperfusion. Neutrophils are regarded as the primary

mediators of injury and their activation leads to ROS release, secretion of different proteases and renal tissue injury [126]. Upon activation, macrophages release proteolytic enzymes and proinflammatory cytokines like TNF-α, IL-1β and interferon-γ (IFN-γ) [127]. In TLR-4 knockout mice subjected to IRI, neutrophil and macrophage infiltration was reduced [124]. Finally, the TLR4-facilitated immune response is linked to renal fibrosis. The upregulation of TLR4 upon I/R induces a strong inflammatory response accompanied by tubular necrosis, loss of brush border, formation of casts and tubular dilatation [124]. Such a robust inflammation is known to potentiate interstitial fibrosis [128].

Proposed endogenous ligands for TLR-4 in renal IRI include HMGB-1, extracellular matrix (ECM) components like biglycan, heparin sulphate and soluble hyaluronan, and heat shock proteins (Hsps) [129–134]. Upon ligand binding, activation of TLR4 leads to downstream signalling via the MyD88-dependent and MyD88 independent pathway (Figure 6). The MyD88-dependent pathway in which MyD88 and TIRAP or MyD88 adapter-like (Mal) recruits and activates members of the IRAK family is considered to be the dominant pathway [124,135]. Wang et al. [136] demonstrated that MyD88- and TRIF-deficient mice showed a significant reduction in interstitial fibrosis reflected by α-SMA and collagen I and II accumulation Furthermore, Administration of the MyD88 specific inhibitor TJ-M2010-2, a small molecular compound, inhibiting the homodimerisation of MyD88, in a renal I/R model in mice has shown to prolong the survival rate, preserve renal function and attenuate the inflammatory responses and apoptosis in the kidney. In the long term, inhibition of the TLR/MyD88 signalling pathway with TJ-M2010-2 attenuated renal fibrosis via inhibition of TGF-β-induced epithelial to mesenchymal transition [137]. Liu et al. [138] showed that pre-treatment with the synthetic TLR4 inhibitor eritoran (Eisai co., Ltd, Tokyo, Japan) in an renal I/R rat model resulted in reduced expression of TNF-α, IL-1β and MCP-1, attenuated monocyte infiltration in the kidney and improved renal outcome Altogether in view of the pivotal role of TLR4 in renal IRI, inhibition of TLR4 or upstream or downstream mediators could be an interesting target in reducing IRI and optimising graft survival.

Next to TLR4, TLR 2 is markedly upregulated upon ischemic injury in the kidney and its upregulation is associated with the initiation of an inflammatory response [139]. Kidneys of TLR2-/ mice subjected to I/R showed less tubular damage compared to TLR2+/+ mice. Reduced levels of MIP-2, MCP-1, and IL-6 and reduced levels of infiltrating leucocytes were seen [140]. The role of TLR2 in the development or progression of renal fibrosis, however, is less clear. Leemans et al. [139] showed that in a mouse model of obstructive nephropathy TLR2 does not play a significant role in renal progressive injury and fibrosis. In addition to this de Groot et al. [141] showed in human allograft biopsies that TLR2 expression 6, 12 and 24 months after transplantation is associated with superior graft outcome in the long run Currently, the humanized immune globuline (Ig) G4 (IgG4) monoclonal antibody against TLR2 OPN-305 (Tomaralimab, Opsona Therapeutics Ltd, Dublin, Ireland) has entered phase 2 trials (NCT01794663) with the aim to reduce delayed graft function in recipients of post-mortal donor kidneys. In the first part (A) of this study a single dose of 0.5 mg/kg administered 1h before reperfusion was associated with full inhibition of TLR2 and an 80% reduction of IL-6 [142]. Subsequently, this dose has been used in part B of the study, which has been completed but results have not been reported yet.

**Figure 6.** Toll-like receptor 4 signalling. Activation of toll-like receptor 4 (TLR4) by danger associated molecular patterns (DAMPs), like high mobility group box-1 (HMGB-1), heat shock proteins (hsp) and extracellular matrix (ECM) components, leads to downstream signalling via the MyD88 (Myeloid differentiation primary-response protein 88) dependent and MyD88 independent pathway. The MyD88-dependent pathway in which MyD88 and TIRAP (toll-interleukin 1 receptor (TIR) domain containing adaptor protein) or MyD88 adapter-like (Mal) recruits and activates members of the IL-1 receptor-associated kinase (IRAK) family is considered to be the dominant pathway. IRAK activation leads to recruitment of TRAF6 (TNF receptor-associated factor 6) and subsequently activation of transforming growth factor beta-activated kinase 1 (TAK1). Activation of TAK1 then leads to the activation of inhibitor of nuclear factor-κB kinase (IKK), which results in the release of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) from its inhibitor, promoting translocation to the nucleus. The MyD88 independent pathway is mediated by the adapter molecules TIR-domain-containing adapter-inducing interferon-β (TRIF)/TRIF-related adaptor molecule (TRAM) and downstream signalling leads to activation of 2 inhibitor of nuclear factor-κB kinase (IKK) homologs IKKε and TANK-binding Kinase-1 (TBK1), which possibly form a complex together and activate transcription factors NF-κB and IFN-regulatory factor 3 (IRF3). From here, proinflammatory gene transcription is initiated. TLR4 signalling is inhibited by Eritoran and TJ-M2010-2.

#### Complement System

The complement system is the second crucial player in the innate immune response in IRI. The system consists of soluble proteins, regulatory proteins and membrane-bound receptors and comprises three pathways. DAMPs released upon I/R are able to activate all three pathways via binding to C1q (classical pathway), C3 (alternative pathway) or PRRs of the lectin pathway (LP).

Recently, the LP has been pointed out as the primary route of renal complement activation after I/R [143]. Activation of the LP can take place through various PRRs like collectins (manose binding lectin (MBL) and collectin-11) [144] and ficolins (ficolin 1-3) [145]. Upon binding of the collectin–mannan-binding lectin serine protease (MASP) complex to carbohydrate-bearing ligands (for instance, mannose or fructose expressed on stressed cells) the MASPs are activated to cleave complement component (C) 4 (C4) and C2. LP activation is critically dependant on the action of MASP-2 [146,147]. In an isograft transplantation model in wild-type and MASP-2-deficient mice, Asgari et al. [147] showed that renal function was preserved with MASP-2 deficiency After complex-ligand interaction,

LP proceeds with cleavage of C4 and C2, mediated by MASP-2, leading to the synthesis of the classical pathway C3 convertase. Recently, a C4 independent bypass in the LP pathway was also demonstrated [122]. This could explain why C4-deficient mice are not protected against renal I/R and cellular mediated rejection [148,149]. One of the PRRs assigned an important role in the LP is collectin-11 (CL-11), a soluble C-type lectin containing a carbohydrate recognition domain and MASP binding domain [150]. In renal tissue, tubular cells are the main source of CL-11 and expression increases after IRI [151]. CL-11 has been appointed an important role in complement activation in the kidney. It has been shown that CL-11 engages l-fucose at sites of ischemic stress and inflammation initiating the LP [147]. In a renal I/R model, CL-11-deficient mice showed no post-ischemic and complement mediated injury supporting the importance of CL-11 in triggering renal complement activation.

All activating routes converge and lead to the formation of the C3 convertase (C4b2b, C3bBbP). C3 convertase cleaves and activates additional C3, creating C3a and C3b. C3b together with C4b2b forms the C5 convertase, which will cleave C5 into C5a and C5b. C5b together with C6–9 will then form the Membrane Attack Complex (MAC, C5b-9). The formed complement effectors will lead to opsonisation (C3b), chemotaxis of neutrophils and macrophages (C3a, C5a) [143]. The formed MAC inserted into the cellular membrane is associated with a proinflammatory response via noncanonical NF-KB signalling (Figure 7) [152,153].

Next to inducing inflammation and cell death, the complement system is able to modulate antigen presentation and T cell priming via C3a and C5a and is therefore playing a role in donor antigen sensitisation and rejection [154]. Antigen-presenting cells (APC) express C3 and C5 along with complement receptors C3aR and C5aR1. Upon complement activation in the extracellular space, C3a and C5a increase the presentation of alloantigens and expression of co-stimulatory molecules on the APC enhancing APC priming of T cells [143]. Furthermore, C3a and C5a promote T-cell differentiation of CD4+ and CD8+ T-cells. CD8+ cells mediate vascular and cellular T-cell mediated rejection. Upon activation, CD4+ T-cells can stimulate further CD8+ T-cell differentiation, they can proliferate and differentiate to memory and effector CD4+ cells which can activate macrophages, recruit leukocytes and stimulate inflammation and finally CD4+ cells stimulate B-cell differentiation and in the end antibody production [143]. The B-cells response can also be enhanced in a direct manner via C3b and C3d on the APC and the complement receptor 2 (CR2) on the B-cell. Activation of the B-cell by binding to the donor alloantigen induces class switching of the donor specific antibody from IgM to IgG. Subsequently, ABMR occurs when IgG donor specific antibodies (DSA) recognizes antigens in the kidney graft and engage with C1q, C1r and C1s to activate the classical pathway [143]. Under normal physiological circumstances, formation of the complement effectors is controlled by proteins (soluble or surface bound) that mediate break down of the C3 and C5 convertases. After I/R this balance shifts to uncontrolled complement activation predisposing the graft to complement mediated injury and rejection [155].

Many interventions on the level of C3, C5, and regulatory proteins in I/R injury and especially kidney transplantation have been evaluated in pre- and clinical studies [156]. Eculizumab (Soliris®, Alexion Pharmaceuticals, New Haven, CT, USA) is to date the best studied complement inhibitor in kidney transplantation. Therapeutic inhibition of C5 with the use of eculizumab, an anti-human C5 micro antibody, showed potential in the prevention and/or treatment in AMBR [157–159] and has been investigated as such in several phase 2/3 clinical trials (NCT01567085, NCT01106027, NCT01399593). All studies report a safety profile of the drug that is consistent with that reported for eculizumab's approved indications like atypical haemolytic uremic syndrome. Results of these trials suggest a potential role of eculizumab in the prevention and treatment of ABMR in patients with DSA [160,161]. Next to ABMR, eculizumab has been investigated for the prevention of DGF (NCT01919346, NCT02145182). Again, the safety profile was good but pre-treatment with eculizumab had no effect on the incidence of DGF. Groups in these studies, however, were rather small [162]. Another anti-C5 antibody Tesidolumab (LFG-316, MorphoSys, Novartis) has currently entered phase 1 studies (NCT02878616).

**Figure 7.** Routes of the complement system with its inhibitors currently studied in kidney transplantation. Damps released upon I/R are able to activate all three pathways via binding to C1q (classical pathway), C3 (alternative pathway) or pattern recognition receptors (PRRs) of the lectin path. All activating routes converge and lead to the formation of the complement component (C) 3 (C3) convertase (C4b2b, C3bBbP). C3 convertase cleaves and activates additional C3, creating C3a and C3b. C3b together with C4b2b forms the C5 convertase, which will cleave C5 into C5a and C5b. C5b together with C6–9 will then form the Membrane Attack Complex (MAC, C5b-9). The formed complement effectors will lead to opsonisation (C3b), chemotaxis of neutrophils and macrophages (C3a, C5a). The formed MAC inserted into the cellular membrane is associated with a proinflammatory response via noncanonical NF-KB signalling. C1-inhibitors (C1-INH), Cinryze®and Berinert®target complement initiation and APT070 complement amplification. Eculizumab and Tesidolumab inhibit complement activation at the level of C5.

In addition to targeting terminal complement pathways, therapeutics targeting complement initiation (C1) and amplification (C3, convertases) have been developed. C1 esterase inhibitors (C1-INH) should not be considered complement-specific inhibitors, since these broad protease inhibitors and their functions extend beyond the classical pathway and even beyond the complement system [163]. The C1INH Cinryze®(Shire US Inc., Lexington, MA, USA) is recently being evaluated for treatment of ABMR (NCT02547220). The study was terminated May 2019 following a pre-scheduled interim analysis, it was determined that the study met the pre-specified criteria for futility. Cinryze®is still listed to be tested as a pre-treatment to reduce IRI and DGF (NCT02435732). Another C1INH, Berinert®(CSL Behring, King of Prussia, PA, USA), has been evaluated in a phase 1/2, double-blind, placebo-controlled study assessing its safety and efficacy for prevention of delayed graft function in recipients of deceased donor kidneys [164]. Although the primary outcome measure (DGF) was not met, treatment with Berinert®was associated with significantly fewer dialysis sessions 2 to 4 weeks post-transplantation. In addition, a better renal function was seen at 1 year compared with the placebo treated group. No significant adverse events were noted in this study [164]. Finally, Mirococept (APT070) a membrane-localising C3 convertase inhibitor is currently being evaluated in a double-blind randomised controlled investigation its efficacy for preventing IRI deceased donor kidneys (EMPIRIKAL-trial, ISRCTN49958194) [165].

#### Translation to the Adaptive Immune System

The link between the innate and adaptive immune response is made by dendritic cells (DCs, Figure 8). DCs are APCs and play an essential role in the pathogenesis of IRI. Immature DCs can be activated by DAMPs via TLRs and the complement system. After maturation, they are able to activate the adaptive immune system in a direct manner by antigen presentation to B- and T-cells or indirectly via cytokine signalling [8,166]. This process can already start in the donor in which in case of a DBD donor, DCs are activated by oxidative stress or C5a and present donor antigens to T-cells of the recipient [167]. Furthermore, it is thought that DCs (subtype CDC11c+ and F4/80+) play an important role in the early pathophysiology of IRI by secretion of TNF-α, Chemokine (C-C motif) ligand 5 (CCL5), IL-6 and MCP-1within the first 24h after IRI [168]. Further, at a later stage, DCs contribute to allograft dysfunction. Batal et al. [169] looked at kidney transplant biopsies performed > 15 days after transplantation and found that a high DC density was independently associated with poor graft survival. Additionally, they found that high DC density was correlated with an increased T-cell proliferation and poor patient outcome in patients with high total inflammation scores of biopsies, including inflammation in areas of tubular atrophy. In these patients, DC density could predict allograft loss. When looking at the origin of the DCs they showed that initially donor DC predominated but found that in late biopsies the majority of DCs were of recipient origin. These data suggest a potential rationale to target DCs influx in the kidney to improve long-term allograft survival.

**Figure 8.** Interaction of the innate and adaptive immune system in the pathophysiology of ischemia and reperfusion injury. DAMPs released upon I/R are able to activate the innate immune system by binding to PRRs like complement receptors and TLRs. Activation of these receptors will lead to production of pro-inflammatory cytokines and chemokines and chemotaxis, opsonisation and activation of leucocytes like macrophages, neutrophils and natural killer (NK) cells. Additionally, immature dendritic cells can be activated, which, after maturation, are able to activate the adaptive immune system in a direct manner by antigen presentation to B- and T-cells or indirectly via cytokine signalling. Treg: regulatory T-cell.

#### 3.3.2. Adaptive Immune Response

In contrast to the non-specific nature of the innate immune response, the role of the adaptive immune system is to recognize alloantigens and to react with an alloantigen-specific response, simultaneously generating immunological memory. Involved cells are B- and T-cells.

#### T-Cells

Activation of T-cells occurs through binding of the T-cell receptor (TCR) on the surface of the T-cell, to the major histocompatibility complex (MHC, in case of humans the human leucocyte antigen (HLA) system) on the APC. This can be in a direct way when the TCR binds to unprocessed allogenic MHC on the APC of the donor or in an indirect manner when MHC proteins of the donor have been taken up by APC of the recipient, processed and presented by the MHC of the recipient [170]. In case of IRI, CD4+ T helper (Th) cells as well as CD8+ cytotoxic T-cells are found in the kidney and are important mediators of IRI [171–174]. T-cell-deficient mice showed attenuated renal IRI and adoptive T-cell transfer experiments in athymic mice resulted in acute kidney injury (AKI) [175–177].

The TCR on CD4+ T-cells can only bind to MHC class 2 molecules (HLA DP, DQ, DR). Upon activation, these CD4+ T-cells become cytokine producing effector cells harming the graft through cytokine mediated inflammation [170]. The effector CD4+ Th cells can differentiate into three major subtypes Type 1 (Th1), Type 2 (Th2) and Th17 cells depending on the cytokines they produce and the transcription factors they express. This differentiation process, referred to as polarisation, starts with induction in lymphoid tissue. Cytokines produced by APCs (DCs and macrophages), NK cells, basophils and mast cells act on T-cells stimulated by the antigen and co-stimulators. This induces transcription of cytokine genes characteristic for the particular subset. Upon continued activation, genetic modifications occur, keeping the characteristic cytokine genes in a transcriptionally active state (commitment). The cytokines produced by the subset promote development of this subset and inhibit differentiation toward other subsets (amplification) [170]. The main effector cytokine of Th1 cells is IFN-γ and the key Th1 transcription factors are signal transducer and activator of transcription (STAT) 4 (STAT-4) and the T-box transcription factor T-bet. Main effector cells are macrophages, B-cells,

CD8+ T-cells and CD4+ T-cells (amplification). IFN-γ secreted by Th1 cells will activate macrophages leading to secretion of inflammatory cytokines (TNF, IL-1 and IL-2), an increased production of toxic substances like ROS, NO and lysosomal enzymes and finally stimulation of expression of costimulatory molecules enhancing the efficiency of the macrophage as APC [170]. The main effector cytokines of Th2 are IL-4, IL-5 and IL-13 and key transcription factors are GATA binding protein 3 (GATA-3) and STAT-6. IL-4 act on B-cells to stimulate production of IgE antibodies which can lead to mast cell degranulation upon binding of IgE with mast cells. IL-5 activates eosinophils, inducing defence against helminthic infections. IL-4 and IL-13 are involved in alternative macrophage activation promoting development of M2 macrophages which have anti-inflammatory effects and may promote tissue repair and fibrosis [170]. Signature cytokines of Th17 are IL-17 and IL-22. Differentiation into this subtype is mediated by IL-6 and TGF-β leading to activation of transcription factors STAT-3 and retinoic acid-related orphan receptor γ*t* (RORγ*t*) respectively. IL-17 act on leukocytes and tissue cells and stimulates production of several chemokines and cytokines (TNF-α, IL-1β, IL-6) that recruit neutrophils and to a lesser extend monocytes to generate an inflammatory response. IL-22 produced in epithelial cells is primarily involved in maintaining the barrier function of epithelia [170]. Th17 T-cell most likely play a significant role in IRI-induced inflammation. STAT-3 KO mice are protected from renal IRI via downregulation of Th17 activity [178]. The differentiated T-cells can convert from one subtype to another by changes in activation circumstances [179]. It is suggested that Th1/Th2 ratio plays an important role in the pathogenesis of IRI [180,181]. Yokota et al. [181] demonstrated that STAT-6-deficient mice with a defective Th2 phenotype have enhanced renal I/R injury whereas STAT-4-deficient mice have mild improved function In addition, Loverre et al. [182] showed that kidney transplant recipients experiencing DGF predominantly expressed Th1 phenotype within the graft In literature both Th1 and Th17 cells are associated with T-cell mediated rejection [183–188].

The TCR on CD8+ T-cells can only bind to MHC class 1 molecules (HLA A, B, C) presented on APCs. Upon activation in lymphoid tissue, they differentiate into cytotoxic T-cells (CTLs) or memory cells. This differentiation is facilitated by CD4+ Th1 cells by secreting cytokines that act directly on the CD8+ cells [170]. The main cytokines involved are IL-2 (proliferation, differentiation CTL/memory cell), IL-12/IFN (differentiation CTL), IL-15 (memory cell survival), IL-21 (memory cell induction). The CTLs are able to kill cells which present the allogenic class 1 MHC of the donor in the graft. This through binding on the target cell and release of granule content into the immune synapse. These granules contain perforin and granzymes. Perforin induces the uptake of granzymes into the target cell. These granzymes are capable of activating caspases and inducing apoptosis. The killing of the target cell can also be Fas/Fas-L mediated in which the CTL expose the Fas ligand on the membrane which will bind to the Fas receptor on the target cell inducing apoptosis. Only CTLs that are activated in the direct way (by donor MHC on donor APC) are able to kill graft cells [170]. Like CD4+ Th cells, CTL secrete inflammatory cytokines, (predominantly IFN-γ) that attribute to inflammation and injury of the graft. The role of CD8+ cells in early phase of renal IRI is unclear, in a mouse model CD4+ deficient mouse was protected from IRI but CD8+ deficient mouse was not [176].

Ko et al. [189] showed that already 6 h after renal IRI, transcriptional activity occurs in T-cells and that these gene expression changes persist up to 4 weeks after the event. Genes involved in immune cell trafficking and cellular movement were most upregulated in the early phase (6 h, 3 days). On day 10 this was shifted to genes related to cellular development products involved in immune responses and on day 28 to genes involved in cellular and humoral immune response involved in antigen presentation. In addition, they found that the CC motif chemokine receptor 5 (CCR5) was one of the most upregulated genes at all time points, which was confirmed at a protein level. Subsequently, the addition of CCR5 antibody attenuated IRI and led to decreased T-cell activation [189].

#### B-Cells

Next to alloreactive CD4+ and CD8+ T-cells, antibodies (immune globulins, Ig) against the graft contribute to rejection. Most of these Igs are produced by Th dependant alloreactive B-cells. The naive B-cell recognizes allogenic MHC-molecules, processes these MHC-molecules and presents them to Th cells that were activated previously by the same alloantigen presented by APCs. The produced Igs (IgM/IgG) are then able to induce complement activation, and activation of neutrophils, NK cells and macrophages. The T-cells are responsible for T-cell mediated rejection and B-cells together with complement activation for ABMR [170].

#### Regulatory T-Cells

The T-cells which most likely play a protective role in renal IRI are regulatory T-cells (Tregs), a subset of CD4+ T-cells whose function is to supress the innate as well as the adaptive immune response and maintain self-tolerance. Tregs can be discriminated from other T-cells by expression of FoxP3 amongst other proteins like CD25. FoxP3 is probably the most important transcription factor for Treg differentiation. The mechanism of action of Tregs is production of immune suppressive cytokines IL-10 and TGF-β, reduction of APC is to stimulate T-cells (possibly by binding to B7 proteins on the APC) and finally consumption of IL-2, an important growth factor for other T-cells [170]. TGF-β inhibits various immune cells amongst which: proliferation and effector functions of T-cells, macrophages, neutrophils and endothelial cells. It regulates differentiation of FoxP3+ Tregs and promotes polarisation towards Th17 cells. Furthermore, TGF-β promotes tissue repair by the ability to stimulate collagen synthesis and matrix modifying enzyme by macrophages and fibroblasts. IL-10 inhibits the production of IL-12 by activated macrophages and DCs, therefore inhibiting these cells and their IFN-γ production. It also inhibits T-cell activation by inhibiting the expression of co-stimulators and MHC-II molecules on DCs and macrophages [170].

Tregs play a potentially promising role in the reduction of IRI and graft tolerance [190–193]. Currently, several clinical trials are running evaluating the safety and effeciacy of FoxP3 cellular therapy in kidney transplantation (NCT02091232, NCT03284242, NCT01446484) [194,195]. However, all that glitters is not gold, since recent studies have shown that human FoxP3+ T-cells show great variations in gene expression phenotype and function [196–199]. Furthermore, recently a subset of FoxP3+ Tregs mimicking Th cells was discovered that secreted pro-inflammatory cytokines [200]. Also, the effect of different immune suppressive agents on the Treg phenotype needs to be elucidated, since these drugs might influence Treg phenotype [200,201]. Altogether, more insight in function and biology is needed before this therapy finds its way to clinical settings.

#### *3.4. Transcriptional Reprogramming*

Finally, cells can protect themselves from hypoxia and ischemia and maintain homeostasis via an evolutionary conserved mechanism with the use of oxygen sensors and activation of specific transcription factors. These so called hypoxic inducible factors (HIFs) regulate various genes involved in the metabolic cell cycle, angiogenesis, erythropoiesis, energy conservation and cell survival and are therefore able to induce a protective cell response to hypoxia [202].

HIFs are heterodimeric transcription factors consisting of an α and β subunit. There are two types of α subunits, HIF-1α and HIF-2α, which have common, but also subunit-specific target genes. In the kidney, HIF-1αis predominantly localized in tubular and glomerular cells, whereas HIF- 2αcan be found in glomerular cells, peritubular endothelial cells and fibroblasts [203–205]. In aerobic circumstances, HIFs are inactive. Oxygen-sensing prolylhydroxylase (PHD) hydroxylates the amino acid proline on the HIF-1α/HIF-2α subunit. This induces a conformational change enabling von Hippel–Lindau tumour suppressor protein (pVHL) to bind with the α-subunit, leading to degradation of the HIF-α subunit. Ischemia/hypoxia will lead to inhibition of the oxygen-dependent PHD, which enables nuclear translocation of the α subunit, binding of the α and β subunit and formation of HIF. In the nucleus HIF binds with the hypoxia response promotor element (HRE) leading to the transcription of various genes like glycolysis enzymes Glut-1 and aldolase (enabling ATP production under hypoxic circumstances), NF-κB, TLRs, adenosine receptors, vascular endothelial growth factor (VGEF), CD73 and erythropoietin. Activation of HIF can also occur in normoxemic circumstances, for instance, by ROS, LPS, various

cytokines and TCR-CD28 stimulation. Transcriptional reprogramming is a consequence of I/R that should be considered a protective mechanism (Figure 9) [206].

**Figure 9.** Intracellular stabilisation and activation of hypoxic inducible factor. Under normoxemic conditions, proline on the hypoxic inducible factor (HIF) α (HIFα) subunit is rapidly hydroxylated by oxygen-sensing prolyl hydroxylase (PHD). This induces a conformational change enabling von Hippel–Lindau tumour suppressor protein (pVHL) to bind with the α-subunit, leading to degradation of the HIF-α subunit. Ischemia (or other signals like lipopolysaccharide (LPS), various cytokines, etc.) will lead to inhibition of the oxygen-dependent PHD, enabling nuclear translocation of the α subunit, binding of the α and β subunit and formation of HIF. In the nucleus, HIF binds with the hypoxia response promotor element (HRE) leading to the transcription of various genes. VGEF: vascular endothelial growth factor.

Conde et al. [207] showed in various models and human post-transplantation biopsies that HIF-1α is induced in a biphasic manner namely during the hypoxic as well as the reperfusion phase. They pointed out the PI3K/Akt mTOR pathway to be responsible for this HIF-1α accumulation during the normoxemic reperfusion phase. In their study, this second increase (e.g., during reperfusion) seemed crucial for tubular cell survival and recovery. During the hypoxic phase, an increase in HIF-1 resulted predominantly in the upregulation of PHD3 and VGEF mRNA, which remained elevated during oxygenation. EPO mRNA was upregulated upon reperfusion. EPO and VGEF have been suggested to be involved in proximal tubular regeneration [208–210]. Their human post-transplantation biopsies revealed HIF-1α expression in proximal tubular cells without ischemic damage or features of regeneration suggesting a protective role for HIF-1α during I/R [207]. Oda et al. [211] had similar findings in patients receiving a DBD/DCD donor kidney. Their analysis of 46 post-transplant biopsies, gained 1h after reperfusion, showed that expression levels of PI3K, Akt, mTOR and HIF-1α were significantly higher in patients without DGF compared to patients experiencing DGF (76% of the patients). The expression levels of HIF-1α and donor type (DCD) were independently associated with DGF HIF-2α expression in renal endothelial cells is suggested in several studies to be protective against renal IRI via protection and preservation of the vasculature endothelium by upregulation of angiogenic factors like VGEF and their receptors Tie2 and VGEFreceptor-2 (FLK-1) [212–215]. Increased production of HIF in myeloid and lymphoid cells influences the innate and adaptive immune response. T-cell activation and proliferation is reduced under hypoxic conditions [216]. A study of Zhang et al. [217] revealed a hypoxia/HIF-2α/adenosine2A receptor axis to be responsible in reduction of NK T-cells activation and renal IRI upon I/R. HIF-1α induces a shift from Th1 to Th2 cells (decrease

Th1/Th2 ratio) accompanied by a decrease in excretion of inflammatory cytokines. Furthermore, HIF-1α promotes transcription of FoxP3 and therefore generation activation of Tregs.

Various PHD inhibitors have been developed and tested in animal I/R models. In a rat model, Wang et al. [218] showed that use of the PHD-1 inhibitor acetate prior to the ischemic event was able to stabilize HIF in a dose-dependent manner and was associated with improved renal outcome. In addition, in an allogenic renal transplant model in rats, the use of the PHD inhibitor FD-4497 pre-donation was associated with increased HIF expression and improved graft outcome and reduced mortality of recipients [219]. Hence, activation and/or upregulation of HIF could be an interesting approach to reduce renal IRI and improve renal transplant outcome. Several PHD inhibitors are currently being tested in clinical trials in order to treat anaemia in patients with chronic kidney disease but have not been tested in the field of transplantation yet.

#### **4. Summary**

The past decade's research in kidney transplant recipients has focussed on post-transplant patient management, with a predominant emphasis on immunosuppression. However, the biggest 'hit' to the donor organ is encountered during the process of donation and reperfusion at time of transplantation, i.e., ischemia and reperfusion injury. An important initiating step in IRI is the uncontrolled ROS formation during reperfusion and dysfunction of the mitochondrial machinery leading to the opening of mPTP and the release of DAMPs in the intra- and extracellular space. From here, several injury cascades are activated, including activation of cell death programs like apoptosis and (regulated) necrosis, endothelial dysfunction implicating increased vasoconstriction upon reperfusion, loss of specific phenotype of endothelial cells and transmigration of leucocytes into the interstitial space. Activation of the innate and subsequently the adaptive immune system will take place through binding of DAMPs to the toll-like receptors and activation of the complement system, leading to further injury of the graft, increased immunogenicity favouring T-cell and antibody mediated rejection and the initiation of fibrosis associated with chronic graft dysfunction. Currently, several novel agents targeting various pathways are tested and, although most are still in the preclinical phase, some have already entered clinical trials. Intervention early in this cascade of events (e.g., on a mitochondrial level), seems very attractive, since mitochondrial dysfunction plays a pivotal role in the initiation of IRI. Due to the complexity of the pathophysiological mechanisms, however, it may be predicted that a multiple treatment strategy using a combination of agents given at various time points during the donation, preservation and transplantation process will most likely be the best strategy to reduce IRI.

**Author Contributions:** G.J.N.-M.: participated in conceptualisation, investigation, visualisation, and writing—original draft preparation. S.E.P.: participated in conceptualisation, investigation, and writing—original draft preparation. S.P.B.: writing—original draft preparation. J.S.F.S.: writing—original draft preparation. R.A.P.: writing—original draft preparation. M.M.R.F.S.: writing—original draft preparation. R.J.P.: participated in conceptualisation, investigation, and writing—original draft preparation. H.G.D.L.: participated in conceptualisation, investigation, and writing—original draft preparation. All authors have read and agreed to the published version of the manuscript.

**Funding:** S.E.P. received funding from the Norwegian Research Council (274352).

**Conflicts of Interest:** The authors declare no conflict of interest. Parts of this review are adapted from chapter 2 of the PhD thesis: Perioperative renal protective strategies in kidney transplantation, Gertrude J. Nieuwenhuijs-Moeke, 2019.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Urinary Excretion of** *N***1-methyl-2-pyridone-5-carboxamide and** *N***1-methylnicotinamide in Renal Transplant Recipients and Donors**


Received: 18 December 2019; Accepted: 4 February 2020; Published: 6 February 2020

**Abstract:** *N*1-methylnicotinamide (*N*1-MN) and *N*1-methyl-2-pyridone-5-carboxamide (2Py) are successive end products of NAD<sup>+</sup> catabolism. *N*1-MN excretion in 24-h urine is the established biomarker of niacin nutritional status, and recently shown to be reduced in renal transplant recipients (RTR). However, it is unclear whether 2Py excretion is increased in this population, and, if so, whether a shift in excretion of *N*1-MN to 2Py can be attributed to kidney function. Hence, we assessed the 24-h urinary excretion of 2Py and *N*1-MN in RTR and kidney donors before and after kidney donation, and investigated associations of the urinary ratio of 2Py to *N*1-MN (2Py/*N*1-MN) with kidney function, and independent determinants of urinary 2Py/*N*1-MN in RTR. The urinary excretion of 2Py and *N*1-MN was measured in a cross-sectional cohort of 660 RTR and 275 healthy kidney donors with liquid chromatography-tandem mass spectrometry (LC-MS/MS). Linear regression analyses were used to investigate associations and determinants of urinary 2Py/*N*1-MN. Median 2Py excretion was 178.1 (130.3–242.8) μmol/day in RTR, compared to 155.6 (119.6–217.6) μmol/day in kidney donors (*p* < 0.001). In kidney donors, urinary 2Py/*N*1-MN increased significantly after kidney donation (4.0 <sup>±</sup> 1.4 to 5.2 ± 1.5, respectively; *p* < 0.001). Smoking, alcohol consumption, diabetes, high-density lipoprotein (HDL), high-sensitivity C-reactive protein (hs-CRP) and estimated glomerular filtration rate (eGFR) were identified as independent determinants of urinary 2Py/*N*1-MN in RTR. In conclusion, the 24-h urinary excretion of 2Py is higher in RTR than in kidney donors, and urinary 2Py/*N*1-MN increases after kidney donation. As our data furthermore reveal strong associations of urinary 2Py/*N*1-MN with kidney function, interpretation of both *N*1-MN and 2Py excretion may be recommended for assessment of niacin nutritional status in conditions of impaired kidney function.

**Keywords:** *N*1-methyl-2-pyridone-5-carboxamide; *N*1-methylnicotinamide; urinary excretion; renal transplantation; kidney function; biomarker; niacin status; tryptophan; vitamin B3

**Carolien P. J. Deen 1,2,3,\*, Anna van der Veen 2, António W. Gomes-Neto 1,4, Johanna M. Geleijnse 5, Karin J. Borgonjen-van den Berg 5, M. Rebecca Heiner-Fokkema 2, Ido P. Kema <sup>2</sup> and Stephan J. L. Bakker 1,3,4**

#### **1. Introduction**

Niacin, or vitamin B3, is the precursor of the nicotinamide nucleotide coenzyme NAD<sup>+</sup>. An adequate niacin status is vital to provide reducing equivalents for energy metabolism, and substrates of NAD<sup>+</sup> consuming enzymes, including adenosine diphosphate (ADP)-ribosyl transferases and deacetylases, that transfer ADP-ribose moieties from NAD<sup>+</sup> and NADP<sup>+</sup> [1,2].

Niacin nutritional status is most commonly assessed by the 24-h urinary excretion of *N*1-methylnicotinamide (*N*1-MN) as a breakdown product of NAD<sup>+</sup>, and recommended as such by authorities, including the WHO and the European Food Safety Authority (EFSA) [3,4]. However, *N*1-methyl-2-pyridone-5-carboxamide (2Py) is the end product of NAD<sup>+</sup> catabolism, after aldehyde oxidase (AOX1)-dependent oxidation of *N*1-MN (Figure 1) [5,6]. Although the 24-h urinary excretion of *N*1-MN has shown the most sensitive response to oral test doses of niacin equivalents [3,7], excretion of 2Py, whether or not combined with that of *N*1-MN, has also been implicated for the assessment of niacin status [8–11].

**Figure 1.** Schematic overview of NAD<sup>+</sup> catabolism. 2Py is the end product of NAD<sup>+</sup> catabolism after AOX1-dependent oxidation of *N*1-MN, framed by the dotted line. AOX1, aldehyde oxidase; *N*1-MN, *N*1-methylnicotinamide; 2Py, *N*1-methyl-2-pyridone-5-carboxamide.

In a recent study, we found that *N*1-MN excretion is lower in renal transplant recipients (RTR) than in healthy controls [12]. As this discrepancy could not be explained by lower dietary intake of niacin equivalents, enhanced enzymatic conversion of *N*1-MN to 2Py by AOX1 might be present in this population due to the suggested contribution of AOX1 to *N*1-MN clearance with lower kidney function [13,14]. It is unclear whether 2Py excretion is increased in RTR, and if so, whether a shift in excretion of *N*1-MN to 2Py can be attributed to kidney function.

Hence, to evaluate the applicability of *N*1-MN excretion as a biomarker of niacin nutritional status in conditions of impaired kidney function, we measured the 24-h urinary excretion of 2Py and *N*1-MN in RTR and kidney donors before and after kidney donation, allowing us to (1) compare the 24-h urinary excretion of 2Py in RTR and kidney donors, (2) investigate the effect of kidney donation on the excretion of 2Py and *N*1-MN in kidney donors, (3) assess whether the urinary ratio of 2Py to *N*1-MN (2Py/*N*1-MN) is associated with kidney function, and (4) identify determinants of urinary 2Py/*N*1-MN in RTR.

#### **2. Materials and Methods**

#### *2.1. Study Population*

This cross-sectional study was based on a well-characterized, single-center cohort of 707 RTR (aged ≥18 years) who visited the outpatient clinic of the University Medical Center Groningen, Groningen, the Netherlands, between 2008 and 2011, with a functioning graft for at least 1 year and no history of alcohol and/or drug abuse [15–17]. As a control group, 367 healthy kidney donors were included who participated in a screening program before kidney donation, and of whom biomaterial was collected before and, after declared eligible, 3 months after kidney donation. Exclusion of subjects with missing biomaterial or niacin supplementation use left 660 RTR and 275 kidney donors, of which 85 underwent donor nephrectomy during the inclusion period, eligible for statistical analyses. Signed informed consent was obtained from all participating subjects and the study protocol was approved by the institutional review board (METc 2008/186) adhering to the Declaration of Helsinki. This study included the same cohort of 660 RTR and 275 kidney donors for data collection as reported previously [12].

#### *2.2. Data Collection and Measurements*

Participants were instructed to collect a 24-h urine sample on the day before their morning visit to the outpatient clinic, and to fast overnight for 8 to 12 h. Urine samples were collected under oil, and chlorhexidine was added as an antiseptic agent. Fasting blood samples were drawn after completion of the urine collection. Laboratory measurements were performed directly with spectrophotometric-based routine clinical laboratory methods (Roche Diagnostics, Rotkreuz, Switzerland). Body composition and hemodynamic parameters were measured according to a previously described, strict protocol [15]. Diabetes was diagnosed if fasting plasma glucose was ≥7.0 mmol/L or antidiabetic medication was used. Proteinuria was diagnosed if total urinary protein excretion was ≥0.5 g/day as measured by a biuret reaction-based assay (MEGA AU510; Merck Diagnostica, Darmstadt, Germany).

Dietary intake was assessed with a validated semi-quantitative food frequency questionnaire (FFQ) [18,19]. The self-administered questionnaire was filled out at home and inquired about 177 food items over the last month. During the outpatient clinic visit, the FFQ was checked for completeness by a trained researcher and inconsistent answers were verified with the participant. The FFQ was validated for RTR as previously reported [16]. Dietary data were converted into daily nutrient intake using the Dutch Food Composition Table of 2006 [20]. Intake of niacin equivalents was calculated by adding up niacin and one-sixtieth of tryptophan intake. Subjects who were using niacin supplementation were excluded. Smoking behavior was assessed with a separate questionnaire [21]. Data on medication and vitamin supplements use, and medical history were obtained from medical records [21].

The estimated glomerular filtration rate (eGFR) was calculated by the combined creatinine and cystatin C-based Chronic Kidney Disease Epidemiology Collaboration equation [22], which has shown to be the most accurate equation in RTR [23]. The glomerular filtration rate (GFR) was measured by infusion of 125I-Iothalamate as described previously [24].

#### *2.3. Assessment of 2Py and N1-MN Excretion*

Measurement of 2Py and *N*1-MN concentrations was performed with a validated liquid chromatography (Luna HILIC column; Phenomenex, Torrance, CA, USA) isotope dilution-tandem mass spectrometry (Quattro Premier; Waters, Milford, MA, USA) (LC-MS/MS) method, as described previously [25], with the addition of *N*1-methyl-2-pyridone-5-carboxamide-d3 in acetonitrile as an internal standard. The 24-h urinary excretion of 2Py and *N*1-MN (μmol/day) was obtained by multiplying concentrations (μmol/L) by total urine volume calculated from weight (L/day).

#### *2.4. Statistical Analysis*

Data are presented as the mean ± SD, median (IQR) and absolute number (percentage) for normally distributed, skewed and nominal data, respectively. Assumptions for normality were checked by visual judgments of the corresponding frequency distribution and Q-Q plot.

Baseline characteristics of RTR and the total cohort of kidney donors were compared by means of *t*, Mann–Whitney, and Chi-Square tests, of which age, sex, body surface area, *N*1-MN excretion and eGFR have been reported previously [12]. Crude associations of 2Py and *N*1-MN excretion with intake of niacin equivalents were investigated with linear regression analyses. Characteristics of kidney donors before and after kidney donation were compared by means of paired samples *t* and Wilcoxon signed rank tests.

Linear regression analyses were used to investigate associations of urinary 2Py/*N*1-MN with kidney function in RTR and kidney donors, with additional adjustments for age and sex. Effect modification between either age or sex and kidney function with urinary 2Py/*N*1-MN was assessed by including the corresponding cross product term in the linear regression model.

Linear regression analyses were furthermore employed to investigate cross-sectional associations of urinary 2Py/*N*1-MN with baseline variables in RTR. Variables were 2-base log-transformed when assumptions of normality and homogeneity of variance of the residuals, based on visual judgement of P-P and scatter plots, respectively, were not met. Multivariable linear regression analyses were used to identify determinants of urinary 2Py/*N*1-MN, by entering terms with *p*-value <0.1 in univariable analysis, and eliminating the least significant term stepwise until the remaining terms contributed significantly to the model.

For all statistical analyses, a two-sided *p*-value of less than 0.05 was considered to indicate statistical significance and SPSS Statistics version 23.0 (IBM, Armonk, NY, USA) was used as software.

#### **3. Results**

#### *3.1. Excretion of 2Py in Kidney Donors and RTR*

The total cohort consisted of 660 stable RTR (57% male; mean age 53.0 ± 12.7 years), included at a median time of 5.6 (2.0–12.0) years after transplantation, and 275 healthy kidney donors (41% male; mean age 53.3 ± 10.7 years) (Table 1). Differences in the 24-h urinary excretion of 2Py and *N*1-MN and kidney function are shown in Table 1. 2Py excretion was higher in RTR than in kidney donors (178.1 (130.3–242.8) versus 155.6 (119.6–217.6) μmol/day, respectively; *p* < 0.001), while *N*1-MN excretion was lower in RTR than in kidney donors (22.0 (15.8–31.8) versus 41.4 (31.6–57.2) μmol/day, respectively; *p* < 0.001). Kidney function was significantly lower in RTR than in kidney donors (eGFR: 45.8 <sup>±</sup> 18.7 versus 91.0 <sup>±</sup> 14.2 mL/min/1.73 m2, respectively; *p* < 0.001 and GFR: 52.4 <sup>±</sup> 17.4 versus 82.3 <sup>±</sup> 29.7 mL/min/1.73 m2, respectively; *p* < 0.001). Urinary 2Py/*N*1-MN was significantly higher in RTR than in kidney donors (8.7 ± 3.8 versus 4.0 ± 1.4, respectively; *p* < 0.001), while the sum of 2Py and *N*1-MN excretion was similar (198.3 (155.9–269.4) versus 203.7 (149.4–274.7) μmol/day, respectively; *p* = 0.98). The urinary fraction of 2Py was higher in RTR than in kidney donors (89.1% (86.4%–91.3%) versus 79.0% (75.6%–82.1%), respectively; *p* < 0.001), and that of *N*1-MN was lower (10.9% (8.7%–13.6%) versus 21.0% (17.9%–24.4%), respectively; *p* < 0.001). The 24-h urinary excretion of 2Py, *N*1-MN and the sum of 2Py and *N*1-MN, but not urinary 2Py/*N*1-MN, were directly associated with intake of niacin equivalents (Table S1).

#### *3.2. Excretion of 2Py and N1-MN before and after Kidney Donation in Kidney Donors*

The 24-h urinary excretion of 2Py and *N*1-MN and kidney function in 85 kidney donors before and after kidney donation are shown in Table 1 and Figure 2. At a median time of 1.64 (1.61–1.87) months after kidney donation, 2Py excretion did not change significantly (152.8 (124.4–215.1) to 161.7 (116.6–227.8) μmol/day, respectively; *p* = 0.31), while *N*1-MN decreased (40.9 (31.0–58.2) to 32.5 (23.4–44.0) μmol/day, respectively; *p* < 0.001). Kidney function decreased significantly after kidney donation (eGFR: 92.8 <sup>±</sup> 13.9 to 60.1 <sup>±</sup> 12.1 mL/min/1.73 m2, respectively; *p* < 0.001 and GFR: 103.7 <sup>±</sup> 16.7 to 65.3 <sup>±</sup> 10.4 mL/min/1.73 m2, respectively; *p* < 0.001). Urinary 2Py/*N*1-MN increased after kidney donation (4.0 ± 1.4 to 5.2 ± 1.5, respectively; *p* < 0.001), while the sum of 2Py and *N*1-MN excretion did not change (198.3 (162.3–270.8) to 189.7 (141.9–271.6) μmol/day, respectively; *p* = 0.90). The urinary fraction of 2Py increased after kidney donation (78.3% (75.5%–81.8%) to 83.5% (80.0%–86.0%), respectively; *p* < 0.001), and that of *N*1-MN decreased (21.7% (18.2%–24.5%) to 16.5% (14.0%–20.0%), respectively; *p* < 0.001).

#### *3.3. Associations of Urinary 2Py*/*N1-MN with Kidney Function*

Urinary 2Py/*N*1-MN was associated with kidney function in RTR (eGFR: <sup>β</sup> <sup>=</sup> <sup>−</sup>0.40; *<sup>p</sup>* <sup>&</sup>lt; 0.001 and GFR: β = −0.39; *p* < 0.001) and the total cohort of kidney donors (eGFR: β = −0.17; *p* = 0.03 and GFR: β = −0.20; *p* = 0.003), but not in the pre- (eGFR: β = −0.01; *p* = 0.94 and GFR: β = −0.02; *p* = 0.89) and post-donation subgroups of kidney donors (eGFR: β = −0.11; *p* = 0.42 and GFR: β = 0.15; *p* = 0.27), with adjustment for age and sex (Table 2). No significant interaction between either age or sex with kidney function was found in the association with urinary 2Py/*N*1-MN in RTR and kidney donors.



Datapresented (IQR)(percentage)normallyrespectively. *p*-valuebetween RTR and the total cohort of kidney donors was tested by *t* and Mann–Whitney tests for normally and skewed distributed continuous variables, respectively. 3 *p*-value for difference between kidney donors before and after kidney donation was tested by paired samples *t* and Wilcoxon signed rank tests for normally and skewed distributed continuous variables, respectively. 4 The urinary fraction of 2Py or *N*1-MN (percentage) was calculated by dividing 2Py or *N*1-MN excretion by the sum of 2Py and *N*1-MN excretion, respectively, and multiplying by 100. BMI, body mass index; eGFR, estimated glomerular filtration rate; GFR, glomerular filtration rate; *N*1-MN, *N*1-methylnicotinamide; RTR, renal transplant recipients; 2Py, *N*1-methyl-2-pyridone-5-carboxamide; 2Py/*N*1-MN, ratio of 2Py to *N*1-MN.

1

**Figure 2.** Box plots of (**a**) urinary 2Py/*N*1-MN, (**b**) eGFR and (**c**) GFR in kidney donors before (*n* = 85) and after kidney donation (*n* = 85) and RTR (*n* = 660), respectively. Boxes, bars and whiskers represent IQRs, medians and values <1.5 × IQR, respectively, whereas outliers (1.5–3 × IQR) are indicated by circles and extreme outliers (>3 × IQR) by asterisks. *p*-value for difference between kidney donors before and after kidney donation was tested by paired samples *t* and Wilcoxon signed rank tests for normally and skewed distributed continuous variables, respectively. *p*-value for difference between RTR and kidney donors before donation was tested by *t* and Mann–Whitney tests for normally and skewed distributed continuous variables, respectively. eGFR, estimated glomerular filtration rate; GFR, glomerular filtration rate; *N*1-MN, *N*1-methylnicotinamide; RTR, renal transplant recipients; 2Py, *N*1-methyl-2-pyridone-5-carboxamide; 2Py/*N*1-MN, ratio of 2Py to *N*1-MN.


**Table 2.** Associations of urinary 2Py/*N*1-MN with kidney function in RTR and kidney donors before and after kidney donation 1.

<sup>1</sup> Linear regression analyses were performed to investigate associations of urinary 2Py/*N*1-MN with kidney function, with adjustment for age and sex. eGFR, estimated glomerular filtration rate; GFR, glomerular filtration rate; *N*1-MN, *N*1-methylnicotinamide; RTR, renal transplant recipients; 2Py, *N*1-methyl-2-pyridone-5-carboxamide; 2Py/*N*1-MN, ratio of 2Py to *N*1-MN.

#### *3.4. Characteristics and Associations with Urinary 2Py*/*N1-MN in RTR*

Characteristics of the RTR cohort are shown in Table 3. Urinary 2Py/*N*1-MN was positively associated with body surface area, body mass index (BMI), glucose homeostasis parameters, triglycerides, mean arterial pressure, high-sensitivity C-reactive protein (hs-CRP), proteinuria, and use of antidiabetics, antihypertensives, acetylsalicylic acid, proton pump inhibitors and tacrolimus. Inverse associations were found between urinary 2Py/*N*1-MN and smoking, alcohol consumption, energy intake, vitamin B6 intake, high-density lipoprotein (HDL) and eGFR.


**Table 3.** Associations of urinary 2Py/*N*1-MN with characteristics in 660 RTR 1,2.

<sup>1</sup> Data are presented as mean ± SD, median (IQR) and absolute number (percentage) for normally distributed, skewed and nominal data, respectively. <sup>2</sup> Linear regression analyses were performed to investigate associations of urinary 2Py/*N*1-MN with baseline variables, of which standardized β and *p*-value are presented. BMI, body mass index; eGFR, estimated glomerular filtration rate; HbA1c, hemoglobin A1c; HDL, high-density lipoprotein; hs-CRP, high-sensitivity C-reactive protein; LDL, low-density lipoprotein; *N*1-MN, *N*1-methylnicotinamide; RTR, renal transplant recipients; 2Py, *N*1-methyl-2-pyridone-5-carboxamide; 2Py/*N*1-MN, ratio of 2Py to *N*1-MN.

#### *3.5. Determinants of Urinary 2Py*/*N1-MN in RTR*

Stepwise multivariable linear regression analyses with backward elimination revealed smoking, alcohol consumption, diabetes, HDL, hs-CRP and eGFR as independent determinants of urinary 2Py/*N*1-MN in RTR (Table 4). In the final model, urinary 2Py/*N*1-MN was positively associated with diabetes (β = 0.10; *p* = 0.01) and hs-CRP (β = 0.10; *p* = 0.009), and inversely associated with smoking (β = −0.13; *p* = 0.001), alcohol consumption (β = −0.12; *p* = 0.002), HDL (β = −0.12; *p* = 0.002) and eGFR (β = −0.38; *p* < 0.001).



<sup>1</sup> Stepwise multivariable linear regression with backward elimination was performed to identify determinants of urinary 2Py/*N*1-MN, of which standardized β and *p*-value are presented. BMI, body mass index; eGFR, estimated glomerular filtration rate; HDL, high-density lipoprotein; hs-CRP, high-sensitivity C-reactive protein; LDL, low-density lipoprotein; *N*1-MN, *N*1-methylnicotinamide; RTR, renal transplant recipients; 2Py, *N*1-methyl-2-pyridone-5-carboxamide; 2Py/*N*1-MN, ratio of 2Py to *N*1-MN.

#### **4. Discussion**

This study aimed to investigate the 24-h urinary excretion of both 2Py and *N*1-MN as major catabolic products of NAD<sup>+</sup> with regard to kidney function. We assessed 2Py and *N*1-MN excretion in RTR and healthy kidney donors as a model of renal disease, and in kidney donors before and after unilateral nephrectomy as a model of isolated renal function impairment. In RTR, 2Py excretion was significantly higher compared to that in kidney donors. Urinary 2Py/*N*1-MN increased significantly in kidney donors after donation. In both RTR and kidney donors, urinary 2Py/*N*1-MN was associated with kidney function. Kidney function was furthermore revealed as the strongest determinant of urinary 2Py/*N*1-MN in RTR.

NAD<sup>+</sup> is formed either de novo from tryptophan via the kynurenine pathway, or via salvage pathways from preformed nicotinamide, nicotinic acid and nicotinamide riboside [26], commonly known as niacin, or vitamin B3. NAD<sup>+</sup> catabolism proceeds via nicotinamide and its downstream metabolites *N*1-MN and 2Py, respectively (Figure 1), and these products are found in both plasma and urine [27]. *N*1-MN itself exhibits anti-inflammatory properties, and is produced by muscle in response to hypoxia and depletion of energy stores, besides its primary production in the liver [28]. Whereas nicotinamide is reabsorbed by renal tubules and only small amounts appear in urine, *N*1-MN and 2Py account for 20%–35% and 45%–60%, respectively, of all urinary NAD<sup>+</sup> metabolites [29]. The WHO and the EFSA recommend the 24-h urinary excretion of *N*1-MN for laboratory assessment of niacin nutritional status accordingly [3,4]. In a previous study, we found that *N*1-MN excretion is clearly reduced in RTR compared to healthy kidney donors [12]. The fact that this is paralleled by a significant elevation of 2Py excretion in the present study, raises speculation that enhanced enzymatic conversion of *N*1-MN to 2Py by AOX1 may be present in RTR. Furthermore, the opposing shifts of 2Py and *N*1-MN excretion in kidney donors after donation, may imply a putative isolated effect of renal function impairment on urinary 2Py/*N*1-MN.

Regarding kidney function, urinary 2Py/*N*1-MN was positively associated with kidney function in both RTR and the total cohort of kidney donors. Renal clearance of *N*1-MN is affected by lower kidney function [13,14], being freely filtered at the glomerulus and tubular secreted, with negligible and saturable tubular reabsorption [30,31]. 2Py has previously been classified as a uremic retention product by the European Uremic Toxin Working Group [32,33], though specific mechanisms of its renal clearance have yet not been characterized. Whereas plasma concentrations of 2Py are reported to increase progressively with chronic kidney disease stages [34], those of *N*1-MN are suggested to be less sensitive to kidney function because of the contribution of AOX1 to *N*1-MN clearance [13,14]. In view of this, we can speculate upon slower excretion of *N*1-MN, hence prolonged exposure to 2Py-forming AOX1 that is related to kidney function, rather than retention of 2Py primarily. This speculation is supported by the fact that kidney function appeared to have only a minor effect on the daily excretion of the sum of 2Py and *N*1-MN in all groups. The presence of a significant association of urinary 2Py/*N*1-MN with kidney function in the total cohort of kidney donors, but not in the pre- and post-donation subgroups, is most likely due to smaller effect sizes in the latter subgroups of kidney donors being declared eligible after pre-donation screening.

The identification of eGFR as the strongest independent determinant of urinary 2Py/*N*1-MN in RTR further supports the notion of an isolated effect of kidney function. Other identified determinant factors include those that are known to affect the enzymatic activity of the aforementioned 2Py-forming AOX1 and most likely contribute as such. In fact, urinary 2Py/*N*1-MN has been used as an index to estimate in vivo AOX1 levels and activity [35], being regulated by a wide variety of endogenous and exogenous factors [36]. Smoking and alcohol consumption are well-known factors [37,38] that showed an inverse association with urinary 2Py/*N*1-MN in RTR. Diabetes and inflammatory mediators [37], including hs-CRP [39,40], have also been implicated in AOX1 activity, as well as HDL-cholesterol-levels via interaction of AOX1 with the ATP-binding cassette transporter A1 (ABCA1) which is a regulator of HDL metabolism [41,42]. Surprisingly, medication use did not appear to affect urinary 2Py/*N*1-MN in RTR, despite the significant function of AOX1 in metabolizing xenobiotics. Importantly, the fact that

urinary 2Py/*N*1-MN has multiple determinants in addition to eGFR, precludes its use as a biomarker of kidney function.

Excessive poly (ADP-ribose) polymerase (PARP) activation induced by stressors such as inflammation, oxidative stress and DNA damage that are predominant in RTR [43,44], has also been implicated in higher production of 2Py from NAD<sup>+</sup> degradation [45,46]. One would, however, expect that this would be reflected by an overall increase of NAD<sup>+</sup> catabolites, which is opposed by the two-fold reduction of *N*1-MN excretion in our RTR population.

In general, higher urinary output of NAD<sup>+</sup> metabolites indicates higher niacin nutritional status, being excreted after the pool of pyridine nucleotide coenzymes is filled [47]. Acute stress may alter this output, but not steady state conditions, in which elimination and production rates are equal [48]. However, the ratio of metabolites is subject to factors that affect not only the activity of 2Py-forming AOX1, but according to our data also kidney function. In a previous study, we found *N*1-MN excretion to be lower in RTR independent of dietary intake of niacin equivalents, as well as to be positively associated with kidney function [12]. According to the present study, the latter association remains when taking into account 2Py excretion, by means of urinary 2Py/*N*1-MN. Therefore, although urinary excretion of *N*1-MN is the most common and recommended index [3,4], our findings suggest that this index might be of limited value in conditions of kidney function impairment and future studies may confirm whether 2Py excretion should at least be additionally interpreted for evaluation of niacin nutritional status.

The speculative presence of slower excretion, hence prolonged exposure of *N*1-MN to 2Py-forming AOX1 with kidney function impairment has not been confirmed in previous studies. In fact, this speculation indicates straight substrate conversion kinetics, which is unlikely to fully account for the previously reported, increased serum concentrations of 2Py in patients with chronic renal failure [46]. More specifically, Rutkowski et al. suggested high serum concentrations of 2Py in chronic renal failure to be a result of kidney function impairment, based on the fact that serum concentrations of 2Py were approximately 20-fold higher in patients with advanced renal failure than in healthy subjects (15.5 ± 5.8 μmol/L versus 0.83 ± 0.18 μmol/L), with only a transient drop after dialysis, and a permanent reduction after kidney transplantation [46]. Accordingly, given its accumulation, along with a deterioration of kidney function, and its toxic properties due to significant inhibition of PARP activity, 2Py has been identified as a uremic toxin [32,45,46]. As we only measured urinary excretion of 2Py, it cannot be ruled out whether increased urinary excretion of 2Py is solely the consequence of increased serum concentrations of 2Py, due to decreased renal clearance, rather than conversion kinetics of *N*1-MN to 2Py by AOX1, and future studies are warranted to address this matter.

Strengths of this study include the large sample size of a specific patient group and the availability of healthy kidney donors before donation as a control group, and after donation as a model of isolated renal function impairment. Moreover, the extensive characterization of RTR allowed us to control for other factors that could affect 2Py/*N*1-MN in 24-h urine, and to comprehensively identify determinants of urinary 2Py/*N*1-MN. The ratio of metabolites in 24-h urine provides a measure to demonstrate changes in metabolism related to renal function, while being the least sensitive to 24-h urine collection errors. Limitations of this study are its observational nature, which prohibits causal inferences, as well as final conclusions on underlying mechanisms of increased urinary 2Py/*N*1-MN in RTR and kidney donors after kidney donation, and associations with kidney function. Therefore, it remains to be determined whether the association of urinary 2Py/*N*1-MN with kidney function is a causal relation. The observational design of this study did neither allow us to rule out increased serum concentrations of 2Py due to decreased renal clearance, or higher production of 2Py from NAD<sup>+</sup> degradation due to PARP activation by means of an experimental design. Conclusions are yet additionally supported by the presence of direct associations of the 24-h urinary excretion of 2Py, *N*1-MN, and the sum of 2Py and *N*1-MN, but not urinary 2Py/*N*1-MN, with niacin nutritional intake (Table S1). Future studies are strongly encouraged to elaborate on serum concentrations of 2Py and *N*1-MN along with their urinary excretion. The present study is confined to the urinary excretion of the major NAD<sup>+</sup>

metabolites, comprising the most common and recommended indices of niacin nutritional status according to existing literature and authorities, including the WHO and the EFSA [3,4], respectively. Other indices, including serum or erythrocyte concentrations of niacin and its metabolites [49], are considered inferior as urinary concentrations have shown the most sensitive response to oral test doses of niacin equivalents [3,7]. Given the aforementioned limitations, this study should be conceived as a descriptive report that precludes final conclusions on the applicability of *N*1-MN excretion as a biomarker of niacin nutritional status in conditions of impaired kidney function. Finally, although niacin deficiency is considered to be uncommon in the developed world, it might be prevalent in subpopulations, including RTR [12]. Still, it should be emphasized that assessment of niacin nutritional status might not be feasible in the developing world given the costs.

#### **5. Conclusions**

The 24-h urinary excretion of 2Py is higher in RTR than in kidney donors, and urinary 2Py/*N*1-MN clearly increases after kidney donation. Urinary 2Py/*N*1-MN is associated with kidney function in both RTR and kidney donors, and kidney function is identified as the strongest determinant of urinary 2Py/*N*1-MN in RTR. Therefore, interpretation of both *N*1-MN and 2Py excretion, rather than *N*1-MN alone, may be recommended for assessment of niacin nutritional status in conditions of impaired kidney function.

**Supplementary Materials:** The following is available online at http://www.mdpi.com/2077-0383/9/2/437/s1, Table S1: Associations of 2Py and *N*1-MN excretion with niacin equivalents intake in RTR and kidney donors.

**Author Contributions:** Conceptualization, I.P.K. and S.J.L.B.; Formal Analysis, C.P.J.D. and S.J.L.B.; Investigation, C.P.J.D. and A.v.d.V.; Resources, J.M.G. and K.J.B.-v.d.B.; Data Curation, A.W.G.-N and S.J.L.B.; Writing—Original Draft Preparation, C.P.J.D.; Writing—Review & Editing, A.v.d.V., A.W.G.-N., K.J.B.-v.d.B, M.R.H.-F. and S.J.L.B.; Supervision, M.R.H.-F., S.J.L.B.; Funding Acquisition, S.J.L.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by TiFN, grant numbers A-1003 and 16NH01.

**Acknowledgments:** Supported by FrieslandCampina and Danone Nutricia Research. The cohort on which the study was based is registered at clinicaltrials.gov as "TransplantLines Food and Nutrition Biobank and Cohort Study (TxL-FN)" with number NCT02811835.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Management of Immunosuppression in Kidney Transplant Recipients Who Develop Malignancy**

**Danwen Yang 1, Natanong Thamcharoen 2,\* and Francesca Cardarelli <sup>1</sup>**


Received: 17 November 2019; Accepted: 9 December 2019; Published: 11 December 2019

**Abstract:** The risk of cancer increases after transplantation. However, the consensus on immunosuppression (IS) adjustment after diagnosis of malignancy is lacking. Our study aims to assess the impact of IS adjustment on mortality of post-kidney transplant patients and allograft outcomes. We retrospectively reviewed the data in our center of 110 subjects. Our results showed IS dose adjustment was not statistically associated with mortality risk (HR 1.94, 95%CI 0.85–4.41, *p* = 0.12), and chemotherapy was the only factor that was significantly related to mortality (HR 2.3, 95%CI 1.21–4.35, *p* = 0.01). IS reduction was not statistically associated with worsening graft function (OR 3.8, 95%CI 0.77–18.71, *p* = 0.10), nor with graft survival (SHR 4.46, 95%CI 0.58–34.48, *p* = 0.15) after variables adjustment. Creatinine at cancer diagnosis and history of rejection were both negatively associated with graft survival (SHR 1.72, 95%CI 1.28–2.30, *p* < 0.01 and SHR 3.44, 95%CI 1.25–9.49, *p* = 0.02). Reduction of both mycophenolate and calcineurin inhibitors was associated with worsening graft function and lower graft survival in subgroup analysis (OR 6.14, 95%CI 1.14–33.15, *p* = 0.04; HR 17.97, 95%CI 1.81–178.78, *p* = 0.01). In summary, cancer causes high mortality and morbidity in kidney transplant recipients; the importance of cancer screening should be emphasized.

**Keywords:** malignancy; cancer; kidney transplant; immunosuppression; graft failure; survival

#### **1. Introduction**

The number of solid organ transplants has increased in the past decade, with 21,167 kidney transplants performed in the United States in 2018. Multiple studies have shown that there is an increased risk of malignancy in transplant recipients [1]. The overall cancer incidence rate is 90 per 1000 patients at 10 years after transplant, which is twice as high as in the general population, while the dialysis population has a 1.35 standardized cancer incident ratio compared to the general population [2]. Nonmelanoma skin cancer is even more frequent, with an incidence rate 14 times higher in transplant recipients compared to the general population.

The burden of malignancy in kidney transplant patients is very high, and the mortality risk in kidney transplant recipients diagnosed with cancer is also greater than nontransplant patients. The median survival of kidney transplant patients with cancer is significantly lower than kidney transplant patients without cancer (2.1 years vs. 8.3 years). Malignancy is currently the second most common cause of death in kidney transplant patients after cardiovascular disease [3].

Despite the surging incidence of cancer in kidney transplant recipients, there is very limited data of how immunosuppression (IS) should be managed after malignancy diagnosis. In current practice, the consensus is that IS dose should be decreased in renal transplant patients with newly diagnosed malignancy, since there is evidence supporting that IS is associated with an increased risk of malignancy and can promote tumor growth [4]. However, specific recommendations regarding how to adjust

IS after diagnosis of malignancy in kidney transplant patients are lacking, and management varies depending on institutions, and even by provider in the same practice. The regimen adjustment ranges from no dose reduction, dose reduction, or cessation of one or more immunosuppressive medications, to class switch. The aim of our study was to assess the impact of changes of IS on patient survival and graft function by retrospectively reviewing data on patients who were diagnosed with malignancy after kidney transplantation in our center.

#### **2. Methods**

This is a retrospective data analysis, in which we identified subjects by manual search of medical records of patients who had kidney transplantations and cancer diagnosis from January 1990 to December 2018 at Beth Israel Deaconess Medical Center, Boston, MA, USA. Data on immunosuppressive regimen, creatinine at cancer diagnosis and one year after diagnosis were extracted from medical records. Time from transplant to cancer diagnosis, patient and graft survival data were calculated from actual dates. Study data were collected and managed using REDCap (Research Electronic Data Capture) electronic data capture tools hosted at Harvard Catalyst—Beth Israel Deaconess Medical Center [5]. Data were collected by chart review following HIPAA guidelines. Institutional Review Board approval was obtained for data collection and analysis, with a waiver for individual consent.

We included all adult patients (18 years or older) who were diagnosed with malignancy after renal transplantation, as seen in Figure 1. Nonmelanoma skin cancer patients who did not require chemotherapy or radiation for cancer treatment were excluded from the analysis. The primary outcome in this study was patient survival. Secondary outcome included graft failure (defined as renal replacement therapy requirement) and worsening renal function (defined as glomerular filtration rate (GFR) reduction of more than 30% or developed graft failure at one year after cancer diagnosis). The mortality and graft function information were obtained from medical records. Our primary variable of interest was dose reduction defined by any types of IS dose reduction. Variables considered to have potential confounding effect were included in the multivariable models, specifically we included demographics of the subjects (i.e., age, race, gender), creatinine at cancer diagnosis, history of rejection, cancer type, donor type, history of chemotherapy, and history of radiation therapy. Races were divided into black and nonblack, which includes Asian, Hispanic, and others. Cancer types were differentiated as solid organ malignancy and hematologic malignancy. Missing data and loss to follow-up were excluded from the analysis. For survival analysis, loss to follow-up cases were censored.

We stratified the population based on whether individuals had IS dose reduction. Means and standard deviations were used to summarize continuous variables with normal distribution. Median (interquartile range) was used for skewed continuous variables. Categorical variables were summarized as percentage. We used t-test to assess the differences in continuous variables that were normally distributed. We tested the difference in categorical variables with Fisher's exact test. Wilcoxon rank sum test was used to test skewed continuous variables. We used logistic regression to assess variables for worsening graft function at one year after cancer diagnosis. For graft failure outcome, competing risk survival analysis (Fine and Gray model) was used to assess cumulative graft failure incidence, and the covariable effect on graft failure was reported as subdistribution hazard ratio. Death without graft failure was considered as a competing outcome. The Cox proportional hazards model was used to assess risk factors for mortality, and the data was censored by last follow-up date. Patient survival was analyzed using the Kaplan–Meier method with significance tested using the log-rank test.

For subgroup analysis, patients were divided into groups according to type of IS reduction (mycophenolate mofetil (MMF), calcineurin inhibitor (CNI), and reduction of both MMF and CNI). We compared each group to the group without any IS changes to assess the risk of worsening graft function and graft failure between these groups. Propensity score adjustment was utilized for subgroup analysis given the small number of subjects in each group. Propensity score of each subject was calculated based on significant factors derived from initial analysis of worsening graft function and graft failure. Then, we performed regression analysis for worsening graft function outcome and Cox regression model for graft failure outcome. Propensity score was applied to the model for adjustment.

All multivariable models ware built based on clinical risk factors and statistically significant variables from univariable analyses. *p* < 0.05 was considered to be significant. The data collected were analyzed using the Stata software version 15.0 (Stata Corp., College Station, TX, USA).

**Figure 1.** Summary of the study.

#### **3. Results**

#### *3.1. Baseline Characteristics of the Study Population*

One hundred and ten subjects who underwent kidney transplantation and developed malignancy were included in our analysis, as seen in Figure 1. Patients' demographics are shown in Table 1.

The mean age at cancer diagnosis was 60.2 years. Male gender contributed to 65.5% of subjects. The ethnicities of subjects were 77.3% non-Hispanic White, and 11.8% non-Hispanic Black. Our study population underwent transplantation during 1971–2018 (1971–1999 in 24 patients and 2000–2019 in 86 patients). Of the study population, 73.6% underwent IS regimen changes (dose reduction or class switch), 26.4% patients had no changes in their IS regimen. Among patients with IS reduction, 26 patients had reduction of both mycophenolate mofetil (MMF) and calcineurin inhibitor (CNI), 19 patients had reduction of CNI only, while 25 patients had reduction of MMF only.

The IS regimen of our patient population is presented in Figure 2. Degree of dose reduction for each IS was showed as median of percent reduced from precancer diagnosis dose in Table 2.


**Table 1.** Baseline characteristic of subjects (N = 110).

**Figure 2.** Types of immunosuppression (IS) used by subjects in the study. MMF = mycophenolate mofetil, AZA = azathioprine.



Medians of percent dose reduction were 60% for tacrolimus, 100% (completely discontinued) for cyclosporine, and MMF or mycophenolic acid. Solid organ malignancies represented 79.1% of the cases; the remainders were hematological cancers. Number of subjects for each type of malignancy are shown in Figure 3.

**Figure 3.** Number of patients in each type of cancer; PTLD = Post-transplant Lymphoproliferative Disorders, GU = Genitourinary, GYN = gynecology, GI = gastrointestinal. Other cancers are head/neck, Kaposi sarcoma, other sarcoma, brain, and unknown origin.

Deceased donor kidney transplant constituted 51.8% of the transplants, and the remainders were from living donors. Mean baseline creatinine at time of cancer diagnosis was 1.6 mg/dL (interquartile range 1.1–1.8 mg/dL). Median time of cancer diagnosis was 6.76 years after transplantation (interquartile range 2.7–11.7 years).

#### *3.2. Mortality*

The mortality rate was very high, at 46.4 % (51/110), with median survival time of 1.8 years after cancer diagnosis (interquartile range 0.7–5.6 years). Thirty patients died within one year of cancer diagnosis. Analysis of mortality in the transplantation eras before and after 2000 was performed by chi-square test, mortality rate between both eras was not statistically significant, *p* = 0.65. Of 51 patients who died, malignancy was the cause of death in 27 patients. Infection was the cause of death in four patients. Eighteen patients had no cause of death recorded. Other causes of death were cardiovascular disease and unknown cause. Kaplan–Meier curve and log-rank test revealed that IS dose reduction significantly increased mortality, *p* = 0.01, as seen in Figure 4.

**Figure 4.** Kaplan–Meier curve and log-rank test of IS dose management and mortality risk.

We performed univariate Cox regression analysis to assess relationship of each variable to mortality, as shown in Table 3. According to our univariate regression analysis model, older age, male gender, IS dose reduction, and chemotherapy were associated with higher mortality. However, in the multivariate model, only chemotherapy remained significant (HR 2.3, 95%CI 1.21–4.35, *p* = 0.01). When we excluded patients who died within six months of cancer diagnosis, the results did not change.


**Table 3.** Effect of immunosuppression dose reduction on patients' mortality. Multivariable analysis was adjusted for age, IS dose reduction, chemotherapy history, and gender. Nonblack race = White, Asian, Hispanic, and other races. \* = Statistically significant, *p* < 0.05.

We also checked the interaction between chemotherapy and dose reduction; the *p* value of 0.36, indicates no strong interaction between those two variables. The spearman correlation coefficient between chemotherapy and dose reduction was 0.28.

#### *3.3. Worsening Graft Function*

There were 100 patients who had post-cancer diagnosis creatinine at one year available. Twenty percent of patients (20/100) developed worsening graft function. In univariate logistic regression, creatinine at cancer diagnosis and female gender were associated with worsening renal function. Those variables remained significant in the multivariable analysis after adjusting for creatinine at cancer diagnosis, IS dose reduction, age, and gender. Interestingly, cancer type, chemotherapy, and donor type were not associated with worsening graft function at one year. The result is shown in Table 4.

**Table 4.** Impact of immunosuppression dose reduction on worsening GFR > 30% at one year after cancer diagnosis. Multivariable analysis was adjusted for age, creatinine at cancer diagnosis, IS dose reduction, and gender. Nonblack race = White, Asian, Hispanic, and other race. \* = Statistically significant, *p* < 0.05.


It is important to note that the direction and magnitude of the estimates for IS dose reduction suggest a potentially strong effect on worsening graft function and mortality outcome, but our study did not have enough power to detect this, given the small number of patients.

#### *3.4. Graft Failure*

In our study, the graft failure rate was 16.4% (18/110). Median graft survival after cancer diagnosis in patients with graft failure was 2.97 years (interquartile range 0.56–4.22 years). Causes of graft failure were acute kidney injury in five patients, "chronic allograft nephropathy" in five patients, and acute rejection in five patients. BK nephropathy, multiple myeloma, and unknown cause contributed to the remaining patients.

As shown in Table 5, in competing risk survival model, creatinine at cancer diagnosis, history of rejection and hematologic cancer were associated with increased risk of graft failure in univariable analysis. After adjusting for age at cancer diagnosis, creatinine at cancer diagnosis, IS dose reduction, malignancy type, and history of rejection, our result showed that creatinine at cancer diagnosis and

history of rejection have remained statistically significant with SHR 1.72, 95% CI 1.28–2.30, *p* < 0.01 and SHR 3.44, 95% CI 1.25–9.49, *p* = 0.02, respectively.


**Table 5.** Impact of immunosuppression dose reduction on graft survival. Multivariable analysis was adjusted for age, creatinine at cancer diagnosis, history of rejection, IS dose reduction, and cancer type. Nonblack race = White, Asian, Hispanic, and other races. \* = Statistically significant, *p* < 0.05.

IS was reduced in all the patients who had graft failure, except for one patient who did not have his IS adjusted, as he was only on low dose tacrolimus monotherapy due to BK viremia. PTLD diagnosis contributed to five out of 18 cases of graft failure.

#### *3.5. Subgroup Analysis*

#### 3.5.1. Worsening Graft Function

We performed subgroup analysis in patients who had IS reduction, defined by reduction of CNI (19 patients), reduction of MMF (25 patients), and reduction of both (29 patients), compared to 29 patients who had no IS change at all to analyze their impact on worsening graft function at one year. After adjusting for gender, age at cancer diagnosis, creatinine at cancer diagnosis using propensity score, reduction of two types of IS was a significant factor for worsening graft function at one year in logistic regression, OR 6.14, 95% CI 1.14–33.15, *p* = 0.04, as seen in Table 6.

**Table 6.** Impact of each type of IS reduction compared to no dose reduction on worsening GFR > 30% at one year after cancer diagnosis Adjusted for gender, age at cancer diagnosis, and creatinine at cancer diagnosis. \* = Statistically significant, *p* < 0.05.


#### 3.5.2. Graft Failure

Subgroup analysis was also performed to assess the impact of different IS reduction regimens on graft failure. The patient groups are the same as subgroup analysis in worsening graft function. In the Cox model adjusted for age at cancer diagnosis, creatinine at cancer diagnosis, history of rejection, and cancer type using propensity score, reduction of both CNI and MMF was associated with graft failure, HR 17.97, 95%CI 1.81–178.78, *p* = 0.01, as seen in Table 7.

**Table 7.** Impact of each type of IS reduction compared to no dose reduction on graft survival. Adjusted for age at cancer diagnosis, creatinine at cancer diagnosis, history of rejection and cancer type. \* = Statistically significant, *p* < 0.05.


#### **4. Discussion**

Although there is increasing evidence of high morbidity and mortality of kidney transplant patients diagnosed with malignancy, specific recommendation on how to adjust IS is lacking. A randomized trial comparing low cyclosporine dose to regular dose found no difference in graft survival or function, although the low-dose regimen was associated with fewer malignant disorders and more frequent rejections [6]. Another randomized controlled trial in 489 kidney transplant patients with 20-year follow-up showed that azathioprine and cyclosporine-based regimens were associated with similar overall long-term cancer risks. In addition, gender, previous antithymocyte globulin (ATG) exposure, and graft failure showed no association with development of malignancy, excluding skin cell carcinoma [7]. One retrospective observational study in heart transplant patients showed that everolimus treatment was associated with lower malignancy risk than MMF [8]. Previous studies showed that sirolimus was associated with reduction in the risk of malignancy and nonmelanoma skin cancer in kidney transplant recipients; however, it was associated with increased mortality risk [9].

KDIGO guidelines published in 2010 recommend considering a reduction of IS for kidney transplant recipients with malignancy (2C recommendation). Important factors to consider (not graded) include the stage of cancer at diagnosis, malignancies which are likely to be exacerbated by IS, available therapies, and whether IS interferes with ability to administer standard chemotherapy [10]. The likelihood of cancer being exacerbated by IS can be assessed using standardized incidence ratio (SIR), which compares the malignancy risk in kidney transplant patients to that in the general population. Cancers with SIR > 3, such as Kaposi's sarcoma, PTLD, and ano-genital cancer, are mostly associated with viral infections, e.g., Human Herpesvirus 8 (HHV8), Epstein–Barr virus (EBV), human papillomavirus (HPV). It has been shown that the incidence of Kaposi's sarcoma, non-Hodgkin's lymphoma, HPV related ano-genital cancer, and melanoma were significantly elevated in patients with functioning transplant graft, but not after transplant failure, when patients were back on dialysis, suggesting that IS has significant effect on these types of cancer. As a consequence, IS adjustment should be strongly considered in these types of malignancy [11,12].

Our study showed that mortality rate in kidney transplant patients with diagnosis of malignancy was high (46.4%), with median survival time of 1.8 years after cancer diagnosis (interquartile range 0.7–5.6 years). Mortality rate was not significantly different between patients who had transplantation before and after year 2000. Interestingly, in our study, malignancy was the main cause of death in subjects whose cause of death was recorded, while the leading cause of death in kidney transplant recipients in general is cardiovascular disease. This data suggests that malignancy contributes to major of mortality in kidney transplant recipients with cancer diagnosis. In addition, more than half

of deceased subjects died within two years of their cancer diagnosis, possibly reflecting advanced cancer at presentation and/or aggressive disease in transplant patients. Our data emphasizes that the appropriate cancer screening could reduce mortality and its importance should be particularly stressed in transplant recipients.

The possible causes of increased mortality risk in this population have been attributed to reduction of immune surveillance in the setting of IS and limited use of certain chemotherapy regimens due to reduced renal function. Notably, kidney transplant recipients and patients with HIV share a similar pattern of increased risk of cancer. Consequently, the increased risk of malignancy after kidney transplantation is thought to be caused by viral infection along with chronic IS use [2].

The significant variable between dose reduction and no reduction groups was whether patients required chemotherapy, suggesting that physicians are more inclined to reduce IS when the cancer is more advanced. The type of cancer (hematologic or solid organ malignancy) did not appear to affect the decision of changing the IS. According to Kaplan–Meier analysis, mortality was significantly higher in the dose reduction group, which is likely confounded by the fact that patients with more advanced stage malignancy tended to have their IS adjusted. Our result is comparable to a previous study in a different center [13]. For multivariate analysis, our study demonstrated that chemotherapy is the only variable associated with mortality, which could be similarly explained by the severity of disease.

As expected, patients with baseline poor kidney function had higher risk of graft failure. The degree of IS dose reduction was significant in majority of patients (IS dose was reduced by at least 50% to completely stopped) putting patients at higher risk of acute allograft rejection. Interestingly, our data showed a novel and important factor in subgroup analysis, reduction of both CNI and MMF put patients at higher risk of graft failure. As a consequence, we recommend that providers should carefully weigh the risks and benefits before drastically changing IS in transplant recipients after cancer diagnosis. A multidisciplinary approach is necessary, focusing on the individual patient's wishes and goals in terms of survival, quality of life, and factor in the possibility of graft failure and return to dialysis. Patients with renal allograft failure returning to dialysis seem to have inferior quality of life and higher rate of depression compared to wait-listed transplant naive patients [14].

Based on our cohort, patients with PTLD had the highest mortality (seven out of 17 patients). Graft failure incidence in patients diagnosed with PTLD was also the highest compared to any other malignancy, as five out of 18 patients who had graft failure were diagnosed with PTLD.

Our study has many limitations. First, it is an uncontrolled retrospective study; therefore, the direct and independent effect of IS changes on mortality could not be clearly determined. Second, our database is from a single center, which has a relatively small number of subjects and heterogeneous cancer types, which might contribute a major confounder. Third, despite adjusting for chemotherapy and radiation therapy, cancer staging was not included in our analysis due to lack of record and heterogeneity of cancer diagnosis. While some chemotherapy regimens could have been a cause graft failure, we did not include this data in our analysis. Lastly, we disregarded the effect of sirolimus and steroid adjustment since both drugs are not part of the standard immunosuppressive regimen at our transplant center.

#### **5. Conclusions**

Our study shows no difference in mortality and graft survival outcomes between reduction and no reduction of IS in kidney transplant recipients diagnosed with cancer. However, it is important to note that the direction and magnitude of the estimates for IS dose reduction suggest a potentially strong effect on worsening graft function and mortality outcome, but a lacking power, caused by the small group of subjects, prevented us to detect the differences. The mortality rate in this population is high and malignancy is usually aggressive; therefore, kidney transplant patients would benefit from early detection of disease by routine cancer screening. The data from our study reveals a novel finding: the risk of graft failure appears remarkably higher after adjusting two immunosuppressive medications. Most importantly, providers should have an extensive discussion with patients regarding the risk

and benefit of IS adjustment, chances of prolonging survival from cancer treatment, and worsening quality of life in case patients develop kidney allograft failure requiring dialysis. As a future direction, a prospective study might be the key to define the temporal effect of IS adjustment on patient's survival, malignancy, and allograft outcomes in kidney transplant recipients.

**Author Contributions:** All authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Conceptualization, Supervision, F.C.; Methodology, Data Curation, Data Analysis, Draft Preparation, D.Y. and N.T. D.Y. and N.T. equally have contributed to this manuscript as first authors.

**Funding:** This research received no external funding.

**Acknowledgments:** We acknowledge the statistic consultation from Harvard Catalyst. REDCap is a secure, web-based application designed to support data capture for research studies, providing (1) an intuitive interface for validated data entry; (2) audit trails for tracking data manipulation and export procedures; (3) automated export procedures for seamless data downloads to common statistical packages; and (4) procedures for importing data from external sources.

**Conflicts of Interest:** F.C. has been part of the Natera scientific board. D.Y. and N.T. declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Beliefs of UK Transplant Recipients about Living Kidney Donation and Transplantation: Findings from a Multicentre Questionnaire-Based Case–Control Study**

### **Pippa K. Bailey 1,2,\*, Fergus J. Caskey 1,2, Stephanie MacNeill 1, Charles Tomson 3, Frank J. M. F. Dor <sup>4</sup> and Yoav Ben-Shlomo <sup>1</sup>**


Received: 8 November 2019; Accepted: 19 December 2019; Published: 21 December 2019

**Abstract:** Differing beliefs about the acceptability of living-donor kidney transplants (LDKTs) have been proposed as explaining age, ethnic and socioeconomic disparities in their uptake. We investigated whether certain patient groups hold beliefs incompatible with LDKTs. This questionnaire-based case–control study was based at 14 hospitals in the United Kingdom. Participants were adults transplanted between 1 April 2013 and 31 March 2017. LDKT recipients were compared to deceased-donor kidney transplant (DDKT) recipients. Beliefs were determined by the direction and strength of agreement with ten statements. Multivariable logistic regression was used to investigate the association between beliefs and LDKT versus DDKT. Sex, age, ethnicity, religion, and education were investigated as predictors of beliefs. A total of 1240 questionnaires were returned (40% response). DDKT and LDKT recipients responded in the same direction for 9/10 statements. A greater strength of agreement with statements concerning the 'positive psychosocial effects' of living kidney donation predicted having an LDKT over a DDKT. Older age, Black, Asian and Minority Ethnic (BAME) group ethnicity, and having a religion other than Christianity were associated with greater degree of uncertainty regarding a number of statements, but there was no evidence that individuals in these groups hold strong beliefs against living kidney donation and transplantation. Interventions should address uncertainty, to increase LDKT activity in these groups.

**Keywords:** living kidney donation; living-donor kidney transplantation; beliefs; inequity

#### **1. Introduction**

Living-donor kidney transplantation offers the best treatment in terms of life-expectancy and quality of life [1–6] for most people with kidney failure. The healthcare costs associated with living-donor kidney transplants (LDKTs) are less than for dialysis or deceased-donor kidney transplants (DDKTs) [7,8]. The medium-term risks of donating a kidney are small [9–12] and the quality of life of donors returns to pre-donation levels after donation [13,14].

Only 20% of those listed on the UK national transplant waiting list receive an LDKT each year [15]. Certain individuals with renal disease appear to be disadvantaged: people from Black and Asian ethnic groups in the UK are less likely to receive an LDKT when compared to White people with kidney

disease [16,17]. Socioeconomic deprivation is also associated with reduced access to living-donor kidney transplantation [16,17]. Older people with kidney disease are less likely to receive an LDKT when compared to younger patients [17], and women are less likely to receive an LDKT when compared to men [18,19]. Ensuring equity in living-donor kidney transplantation has been highlighted as a UK and international research priority by patients and clinicians [20–22]. Differing beliefs in the acceptability of living kidney donation and transplantation have been proposed as a possible explanation for the observed differences in access [17,23,24].

In this questionnaire-based case–control study, we compared the beliefs of LDKT and DDKT recipients about the acceptability of living kidney donation and transplantation. We investigated whether beliefs about living-donor kidney transplantation were associated with an individual's sex, age, ethnicity, religion and education. We aimed to identify groups with beliefs against living-donor kidney transplantation, that may explain the observed disparities in the uptake of LDKTs.

#### **2. Experimental Section**

#### *2.1. Participants*

The study was based at 14 UK hospitals (listed in Supplementary Methods). We wanted to investigate beliefs about living-donor kidney transplantation specifically, and not kidney transplantation in general. Therefore, we did not invite people with Chronic Kidney Disease or those on dialysis to participate, as some of these individuals may have held beliefs against transplantation in general, as opposed to living-donor kidney transplantation specifically. We obtained from each site an anonymised list of all individuals who received kidney transplants between 1 April 2013 and 31 March 2017, stratified by LDKT and DDKT status. Individuals aged <18 years at the time of transplantation, and individuals lacking mental capacity according to the Mental Capacity Act 2005, were excluded. We performed stratified random sampling using Stata 15 [25] to select, on average, 110 LDKTs and 110 DDKTs from each site, weighted by the number of transplants performed annually at each study site. Sex and 5-year age group strata-matched sampling was used to ensure a similar sample distribution by age and sex. The case–control study was designed to detect a 7-point difference in a continuous measure of patient activation (analysis of this variable not presented here) between LDKT cases and DDKT controls with 90% power, assuming a 5% significance level. The calculation indicated that 170 patients would be needed, and that, therefore, a total of 944 would be needed to allow analyses stratified by Index of Multiple Deprivation rank quintile and allow for 10% missing data. This sample size allows for the detection of a far smaller difference (0.16 Standard Deviation) for a dichotomous exposure or between 6–8% for a categorical outcome [26].

#### *2.2. Questionnaire Content and Survey Tools*

Paper questionnaires were mailed by post to participants by research collaborators at the study sites. Questionnaires were accompanied by a patient information sheet, an invitation letter and a return postage paid envelope. A website-address was provided so that participants could complete the questionnaire online if preferred. Non-responders were sent a second questionnaire after 4–6 weeks. Anonymised data were extracted from returned paper questionnaires at the University of Bristol and uploaded onto a secure REDCap database [27].

Transplant beliefs were assessed using questions developed by Stothers et al. [28,29]. In development, the questions were reviewed by three expert focus groups, then evaluated in a pilot study to test content reliability and validity [28]. Test–retest analysis was reported as demonstrating excellent internal consistency, and there was no evidence of 'skew' or 'halo' effects (an overall perception/feeling of satisfaction that influences all responses rather than allowing a thoughtful consideration of each individual question) [28]. Participants were asked to read ten statements describing a belief regarding living-donor kidney transplantation (Box 1). These included statements regarding the acceptability of receiving a donated kidney, coercion or pressure on family to donate,

rewards for the donor, required closeness of relationship, the subsequent effect on relationship, beliefs about recipients asking family to donate, donation from offspring to parents, and the risks of donation. Participants were asked to tick one of the following options: (i) Strongly disagree, (ii) Disagree, (iii) Agree, (iv) Strongly agree, (v) Don't know.

#### **Box 1.** Belief statements.


Questionnaires assessed participant demographics as indicated in Box 2.

**Box 2.** Participant demographic data collected.

• Sex

	- -10–19 years; 20–29 years; 30–39 years; 40–49 years; 50–59 years; 60–69 years; 70–79 years; 80–89 years
	- -No religion; Christian; Muslim; Jewish; Hindu; Sikh; Buddhism
	- - No formal education; Primary school; Secondary school; Vocational/Technical; University—undergraduate; University—postgraduate; Other
	- -White;
	- -Asian/Asian British;
	- -Black/African/Caribbean/Black British;
	- - Mixed/Multiple (White and Black Caribbean, White and Black African, Any other Mixed/Multiple ethnic background);
	- -Other (Arab, Any other ethnic group)

#### *2.3. Statistical Analysis*

We compared demographic characteristics between DDKT and LDKT recipients using chi2 tests. The proportion of DDKT and LDKT recipients selecting each level of agreement with a belief statement was calculated and initially compared using chi<sup>2</sup> tests. We used multivariable logistic regression to look at the association of transplant type (LDKT versus DDKT) with a recipient's agreement with a belief statement. For the multivariable logistic regression, the response options were coded 1–4 (1 = strongly disagree, 2 = disagree, 3 = agree, 4 = strongly agree) with 'Don't know' coded as missing. For each belief statement we ran an unadjusted model and one adjusted for potential confounders. We specified, a priori, potential confounders including sex, age, education level, ethnicity and religion. We used robust standard errors to account for clustering within renal centres. Statistical analyses were performed in Stata 15 [25].

Basic descriptive statistical tests (chi<sup>2</sup> tests) then were performed to look for differences in response (agreement = strongly agreed and agreed; disagreement = strongly disagreed and disagreed; and don't know) across different patient demographic groups. For these analyses, age was dichotomised into age <60 years and age ≥60 years, ethnicity into White, Black, Asian and Minority Ethnic (BAME) groups, education into university education or no university education, and religion divided into three categories: no religion, Christianity, or other religion. Small numbers of respondents from certain ethnic groups and from religions other than Christianity or none limited subgroup analysis. Small numbers and single participant responders in some groups risked identification: we were therefore required to combine Islam, Hinduism, Judaism, Buddhism, and Sikhism as 'religions other than Christianity' for analysis.

#### *2.4. Ethical Approval and Consent*

We received NHS Research Ethics Committee (REC) (REC reference 17/LO/1602) and Health Research Authority (HRA) approval. A consent form formed the first page of the questionnaire. The study was funded by a Kidney Research UK Project Grant (RP\_028\_20170302). The clinical and research activities being reported are consistent with the Principles of the Declaration of Istanbul as outlined in the 'Declaration of Istanbul on Organ Trafficking and Transplant Tourism'.

#### **3. Results**

A total of 1240 questionnaires were returned from 3103 patients (40% response). Participant characteristics are reported in Table 1.

LDKT recipients were more likely to respond than DDKT recipients (46% vs. 34%) and women were more likely to respond than men (43% vs. 37%) (Table S1). However, the study participants were a population representative sample (Table S2). Overall, the proportion of missing data was small (<3% for belief questions and <10% for all demographic variables) (Supplementary Missing data).

#### *3.1. Comparison of LDKT and DDKT Recipients*

DDKT recipients expressed greater uncertainty than LDKT recipients regarding all belief statements, with a greater proportion of DDKT than LDKT recipients selecting 'Don't know' for every question (Table 2).

The direction of belief for DDKT and LDKT recipients was the same for nine statements (Table 2). The majority of both DDKT and LDKT recipients agreed with the statements: (1) It is morally acceptable to take a kidney from a healthy person; (3) Donating a kidney is a rewarding experience for live donors; (5) A living-donor kidney transplant may strengthen the relationship between the donor and recipient; (8) It is acceptable for a parent to receive a kidney from his/her child (over 18 years old); (9) Decisions about donation should be made by the donor alone. The recipient should not ask for a kidney. The majority of both DDKT and LDKT recipients disagreed that: (4) Donating a kidney to someone requires an extremely close personal relationship; (10) Since the donor operation is not risk free, someone who needs a kidney transplant should wait for a kidney from someone who has died. For these seven statements, DDKT and LDKT recipients who indicated that they had a belief (rather than did not know) reported the same direction of belief but for all questions a greater proportion of LDKT recipients indicated a stronger belief than DDKTs.

No difference between DDKT and LDKT recipients was found with either direction or strength of belief with respect to Statement (3)—'Asking someone to donate makes the recipient seem selfish'. Statement (6)—'Approaching a potential donor who then says no will change the relationship between

the two people'—was associated with the greatest uncertainty for all participants; 36% of DDKT recipients and 34% of LDKT recipients selecting 'Don't know' for this question.

DDKT and LDKT recipients differed in the direction of their belief with respect to only one statement. For statement (2)—'Donors often agree to donate due to feelings of guilt or family pressure'—the majority of LDKT recipients disagreed whilst DDKT recipients were split between disagreement, agreement and not knowing (Table 2).

#### *3.2. Predictors of Case–Control Status*

The strength of agreement with seven belief statements predicted case–control status, even after adjustment for potential confounders (Table 3). A greater level of agreement with statements 1, 3, 5, and 8 predicted being an LDKT over a DDKT recipient. These statements concern the 'acceptability' of living donation and transplantation, and its 'positive effects' ('rewarding experience' and 'strengthening relationship'). A greater level of disagreement with statements 2, 6 and 10 predicted being an LDKT over a DDKT recipient. These statements relate to beliefs about individuals experiencing 'pressure to donate' and the 'risks/negative impacts of living donation'.

#### *3.3. Participant Characteristics and Beliefs (Table S3a–e)*

#### 3.3.1. Sex

For only one of the ten statements, responses from women and men differed. The majority of women and men agreed with Statement 8—'It is acceptable for a parent to receive a kidney from his/her child (over 18 years old)'—but a greater proportion of women disagreed compared to men (14% versus 8%, chi<sup>2</sup> *p*-value < 0.001 across all categories of agreement).

#### 3.3.2. Age

For four of the ten statements, older respondents indicated greater uncertainty by selecting 'Don't know' rather than indicating a direction of belief. Individuals aged ≥60 years were more likely than individuals aged <60 years to answer 'Don't know' for statement (2)—'Donors often agree to donate due to feelings of guilt or family pressure' (36% versus 24%, chi2 *p*-value < 0.001 across all categories of agreement), statement (5)—'A living-donor kidney transplant may strengthen the relationship between the donor and recipient' (23% versus 16%, chi<sup>2</sup> *p*-value 0.02 across all categories of agreement), statement (6)—'Approaching a potential donor who then says no will change the relationships between the two people' (41% versus 31%, chi<sup>2</sup> *p*-value < 0.001 across all categories of agreement), and statement (7)—'Asking someone to donate makes the recipient seem selfish' (32% versus 18%, chi<sup>2</sup> *p*-value < 0.001 across all categories of agreement).

For one statement, statement (9)—'Decisions about donation should be made by the donor alone. The recipient should not ask for a kidney'—the direction of belief differed with age. People aged <sup>≥</sup>60 years were much more likely to agree compared to people aged <60 years (73% versus 57%, chi<sup>2</sup> *p*-value < 0.001 across all categories of agreement).



*J. Clin. Med.* **2020**, *9*, 31

a

nearest whole number. c Hindu, Jewish, Sikh, Buddhist and Other combined due to single participant responders in some groups risking identification.



DDKT = deceased-donor kidney transplant; LDKT = living-donor kidney transplant.

#### *J. Clin. Med.* **2020**, *9*, 31



Adjusted for sex, 10-year age-group, ethnicity (White and Black, Asian and Minority Ethnic (BAME) groups), religion (No religion, Christian, Other), university education (university education or no university education.

a

#### 3.3.3. Education

For two of the ten statements, a greater proportion of those who did not go to university disagreed with the statement compared to those who did: statement (5)—'A living-donor kidney transplant may strengthen the relationship between the donor and recipient' (13% vs. 7%, chi<sup>2</sup> *p* = 0.008), and statement (6)—'Approaching a potential donor who then says no will change the relationship between the two people' (49% versus 42%, chi2 *p*-value 0.03). For statement (9)—'Decisions about donation should be made by the donor alone. The recipient should not ask for a kidney'—individuals without a university degree were more likely to agree than those with (66% versus 58%, chi2 *p*-value 0.04).

Individuals without a university degree indicated greater uncertainty with respect to statement (7)—'Asking someone to donate makes the recipient seem selfish'—with a higher proportion selecting 'Don't know' compared to those with a university degree (26% versus 18%, chi2 *p*-value 0.01).

#### 3.3.4. Ethnicity

The majority of both white and non-white individuals agreed with statement (1) regarding the moral acceptability of taking a living-donor transplant (89% and 81%), but of the remainder, non-white individuals were more likely to select 'Don't know' than white individuals (13% versus 6%, chi2 *p* value = 0.002). Statement (10)—'Since the donor operation is not risk free, someone who needs a kidney transplant should wait for a kidney from someone who has died'—generated the greatest ethnic difference in opinion: non-white individuals were less likely to say they disagreed with this statement (69% versus 85%) and more likely to indicate that they did not know (21% versus 9%, chi<sup>2</sup> *p* < 0.001).

#### 3.3.5. Religion

For statement (1)—'It is morally acceptable to take a kidney from a healthy person'—a greater proportion of people from the 'Other religions' group selected 'Don't know' (13%) compared to those of no religion (5%) and Christians (7%) (Chi2 *p* = 0.01). Similarly, for statement (3)—'Donating a kidney is a rewarding experience for the live donors'—individuals from the group comprising religions other than Christianity were less likely to agree, and more likely to select 'Don't know' (24%) compared to those of no religion (19%) and Christians (11%) (Chi2 *p* < 0.001). For statement (10)—'Since the donor operation is not risk free, someone who needs a kidney transplant should wait for a kidney from someone who has died'—a smaller proportion of people in the 'Other religions' group said that they disagreed with this statement (65%) compared to people of no religion (89%) or Christians (89%), and a greater proportion selected 'Don't know' (24%) compared to Christians (10%) and people with no religion (8%) (chi2 *p* < 0.001).

For statement (6)—'Approaching a potential donor who then says no will change the relationship between the two people'—a slightly greater proportion of Christians (49%) disagreed with the statement compared those of 'Other religions' (43%) or none (42%) (chi2 *p* = 0.008).

#### **4. Discussion**

In this questionnaire-based case–control study, we compared the beliefs of LDKT and DDKT recipients about the acceptability of living kidney donation and transplantation. We found no evidence that DDKT recipients hold strong beliefs against living-donor kidney transplantation. Rather, DDKT recipients hold similar beliefs to LDKT recipients, but report less conviction and greater uncertainty. We did not investigate the source of beliefs in this questionnaire, but it would be interesting to investigate whether the greater uncertainty in the DDKT respondents influences or reflects the beliefs of family members and potential donors. Uncertainty may reflect differing or conflicting beliefs within a family regarding the acceptability of living-donor kidney transplantation.

We aimed to identify groups with beliefs against living-donor kidney transplantation that may explain observed sex, age, ethnic and socioeconomic disparities in the uptake of LDKTs. Overall, we did not find any evidence of significant difference in the direction of belief with sex, age, ethnicity, religion or education. This suggests that inequality in LDKT uptake associated with sex, age, ethnic, or socioeconomic position is not explained by disproportionately high numbers of individuals in these groups holding beliefs that are incompatible with living-donor kidney transplantation.

BAME group ethnicity and having a religious affiliation other than Christianity were both associated with greater uncertainty regarding a number of belief statements. BAME individuals were particularly uncertain as to whether one should wait for a DDKT, given that living kidney donation is not risk free. Uncertainty regarding organ donation and transplantation has previously been reported in qualitative research amongst certain ethnic and religious groups, attributed specifically to uncertainty regarding religious edicts [31,32]. One qualitative study from the Netherlands identified a lack of awareness about the 'official' position of an individual's religion regarding living organ donation within communities, and confusion due to differing interpretations of religious texts [32]. Research from the USA has shown that, amongst church-attending African-American individuals without kidney disease, 37% disagreed with living donation [33], and members of the clergy were more likely to express reservations about living donation than deceased donation (33.3% versus 16.7%) [33]. These studies suggest that faith leaders might play an important educational role, that their opinion might be influential, and that clarity over the position of the religion on living-donation needs to be made explicit [32–34]. To this end, during the preparation of this manuscript, a new fatwa clarifying Islamic approval of living and deceased organ donation and transplantation was published in the UK [35].

Older people reported greater uncertainty in their beliefs about the impact of donation on the family, and whether asking is selfish on the recipient's part. Older people have been reported as being unhappy to accept an organ from a younger living donor [36,37], in part due to parents believing they should protect their children from harm [36,37]. This belief regarding the acceptability of living-donor kidney transplantation might be influenced by clinicians: research from the USA has suggested that eligible older people with kidney disease are less likely to be encouraged to seek a transplant by their nephrologists [38].

Our findings suggest that the majority of DDKT recipients believe living kidney donation and living-donor kidney transplantation are acceptable, appropriate and justifiable. The majority of demographic groups believe that there are benefits from LDKTs to both the donor and the recipient. Given these beliefs, it suggests that there is capacity to increase LDKT activity in the UK. There should be no assumption that people of certain groups (BAME or older people) have strong beliefs against an LDKT—but rather, any uncertainty should be taken as an opportunity to engage in discussion. Attitudes towards living kidney donation are often open to change and, accordingly, can be influenced [39]. Conversations with religious leaders may help to overcome specific uncertainties regarding a particular religion's position on living donation [34,35].

The belief statements in this study were first developed and used in a Canadian population [29]. LDKT recipients and wait-listed patients surveyed in Canada were found to have the same direction of response as LDKT recipients and DDKT recipients in the UK for all statements except for Statements (4) and (10). For Statement (4)—'Donating a kidney to someone requires an extremely close personal relationship'—69% Canadian LDKT recipients agreed or strongly agreed with this statement, compared to 26% of UK LDKT recipients. For statement (10)—'Since the donor operation is not risk free, someone who needs a kidney transplant should wait for a kidney from someone who has died'—a greater proportion of UK DDKT recipients disagreed with this statement when compared to Canadian wait-listed patients (72% versus 52%). These differences may reflect transplant practice and beliefs changing over time, since the Canadian study was undertaken over 15 years earlier. However, these differences may in part explain why the UK's LDKT activity is greater than Canada's [40], and this requires further investigation.

In our study, statement (10)—'Since the donor operation is not risk free, someone who needs a kidney transplant should wait for a kidney from someone who has died'—generated the most difference in opinion; therefore, how beliefs will change with the UK's move to an opt-out deceased-donation law in 2020 will need to be investigated.

This was a large, multicentre study. To our knowledge, this is the first quantitative study to investigate beliefs about living-donor kidney transplantation amongst transplant recipients. The questionnaire was evaluated in cognitive interviews prior to use, validated and then piloted [26]. The proportion of missing data was small. However, the study has limitations: (i) Although our response rate was reasonable for an unincentivized postal survey, and compares to the response rate of other postal surveys in the UK [41,42] and that of previous a previous European transplant survey [43], there is a risk of self-selection bias. We have reported in our results that our population appeared population representative (Table S2). In addition, we compared our findings to those from the Access to Transplantation and Transplant Outcome Measures (ATTOM) study (which had 72% participation), and found the same effect sizes between socioeconomic position and likelihood of an LDKT (see Table S4) providing further evidence our sample is fairly representative of the total population of such patients. (ii) A total of 14% of participants were from BAME groups—this is not a surprising finding as in the UK between 2013 and 2017 BAME individuals comprised 17% of LDKT recipients and 27% of LDKT kidney transplant recipients [44], but this did prevent the analysis of individual ethnic groups (e.g., Asian, Black, Chinese).

The questionnaire was administered to LDKT and DDKT transplant recipients, both of whom have experienced transplantation; thus in the analyses examining the relationship between beliefs and transplant type, one might expect responses to be subject to a range of cognitive biases, including justifying their decision, and endowment effects. However, evidence against a significant endowment effect on the direction of belief includes the finding that the majority of DDKT recipients expressed positive beliefs about living donation and transplantation. Were there significant endowment effects, we would not have expected the majority of DDKT recipients to express positive beliefs about LDKTs. Cognitive biases do not explain the differences in beliefs between different demographic groups.

#### **5. Conclusions**

The majority of both DDKT and LDKT recipients across all demographic groups reported holding positive beliefs about living donation and transplantation. This encouraging finding suggests that, at least on the part of the transplant candidate, beliefs that are incompatible with LDKT are not a major barrier to living-donor transplantation in the UK, and that there is capacity to increase LDKT activity.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/1/31/s1, Supplementary Methods; Table S1. Responders and non-responders, Table S2. Responders compared to national denominator population, Supplementary Missing Data, Table S3a. Difference in beliefs with participant sex, Table S3b. Difference in beliefs with participant age, Table S3c. Difference in beliefs with participant education, Table S3d. Difference in beliefs with participant ethnicity, Table S3e. Difference in beliefs with participant religion, Table S4. Comparison with participants in ATTOM study.

**Author Contributions:** Conceptualization, P.K.B., Y.B.-S., C.T.; Methodology, P.K.B., Y.B.-S., S.M.; Software, P.K.B.; Formal analysis, P.K.B., S.M.; Investigation, P.K.B.; Data curation, P.K.B.; Writing—original draft preparation, P.K.B.; Writing—review and editing, P.K.B., S.M., F.J.M.F.D., Y.B.-S., C.T., F.J.C.; Supervision, Y.B.-S., C.T., F.J.C.; Project administration, P.K.B.; Funding acquisition, P.K.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This report is independent research arising from a Kidney Research UK Project Grant (Reference RP\_028\_20170302). Neither Kidney Research UK nor the University of Bristol had any role in study design, data collection, analysis, interpretation, manuscript preparation of the decision to submit the report for publication. YB-S is the equity theme lead for the NIHR Collaboration for Leadership in Applied Health Research and Care West (CLAHRC West) at University Hospitals Bristol NHS Foundation Trust. CLAHRC West is part of the NIHR and is a partnership between University Hospitals Bristol NHS Foundation Trust and the University of Bristol. The views expressed in this publication are those of the authors and not necessarily those of the funder.

**Acknowledgments:** The authors would like to thank all the study participants, the participating centre research nurses and coordinators (Hugh Murtagh, Nina Bleakley, Mary Dutton, Kulli Kuningas, Cecilio Bing Andujar, Ann-Marie O'Sullivan, Nicola Johnson, Kieron Clark, Thomas Walters, Mary Quashie-Akponeware, Jane Turner, Gillian Curry, Hannah Beer, Lynn.D Langhorne, Sarah Brand, Maria Weetman, Molly Campbell, Megan Bennett, Sharirose Abat, and Agyapong Kwame Ansu) and the local collaborators who facilitated the study (Sarah Heap, Mysore Phanish, Shafi Malik, Aisling Courtney, Adnan Sharif, Nicholas Torpey, Refik Gökmen, Michael Picton, Linda Bisset, Edward Sharples, and Simon Curran).

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **Intraoperative Fluid Restriction is Associated with Functional Delayed Graft Function in Living Donor Kidney Transplantation: A Retrospective Cohort Analysis**

**Gertrude J Nieuwenhuijs-Moeke 1,\*, Tobias M Huijink 2, Robert A Pol 2, Mostafa El Moumni 2, Johannes GM Burgerhof 3, Michel MRF Struys 1,4 and Stefan P Berger <sup>5</sup>**


Received: 8 August 2019; Accepted: 23 September 2019; Published: 2 October 2019

**Abstract:** Background: In 2016 we observed a marked increase in functional delayed graft function (fDGF) in our living donor kidney transplantation (LDKT) recipients from 8.5% in 2014 and 8.8% in 2015 to 23.0% in 2016. This increase coincided with the introduction of a goal-directed fluid therapy (GDFT) protocol in our kidney transplant recipients. Hereupon, we changed our intraoperative fluid regimen to a fixed amount of 50 mL/kg body weight (BW) and questioned whether the intraoperative fluid regimen was related to this increase in fDGF. Methods: a retrospective cohort analysis of all donors and recipients in our LDKT program between January 2014–February 2017 (*n* = 275 pairs). Results: Univariate analysis detected various risk factors for fDGF. Dialysis dependent recipients were more likely to develop fDGF compared to pre-emptively transplanted patients (*p* < 0.001). Recipients developing fDGF received less intraoperative fluid (36 (25.9–50.0) mL/kg BW vs. 47 (37.3–55.6) mL/kg BW (*p* = 0.007)). The GDFT protocol resulted in a reduction of intraoperative fluid administration on average by 850 mL in total volume and 21% in mL/kg BW compared to our old protocol (*p* < 0.001). In the unadjusted analysis, a higher intraoperative fluid volume in mL/kg BW was associated with a lower risk for the developing fDGF (OR 0.967, CI (0.941–0.993)). After adjustment for the confounders, prior dialysis and the use of intraoperative noradrenaline, the relationship of fDGF with fluid volume was still apparent (OR 0.970, CI (0.943–0.998)). Conclusion: Implementation of a GDFT protocol led to reduced intraoperative fluid administration in the LDKT recipients. This intraoperative fluid restriction was associated with the development of fDGF.

**Keywords:** fluid management; kidney transplantation; delayed graft function; goal-directed fluid therapy

#### **1. Introduction**

During the procedure of organ donation and transplantation a number of potentially harmful processes will inevitably occur, affecting the viability of the kidney graft. Both donor and recipient are subjected to anesthesia and surgery, which will produce a sequence of systemic and local changes, including a significant proinflammatory and procoagulatory response [1]. The donor organ is by definition, exposed to a number of phases of injury from the moment the donor suffers from cerebral injury (in case of brain death) until the kidney is reconnected to the circulation in the recipient. These phases include a profound systemic and local proinflammatory and procoagulatory response during donor management and retrieval, associated with hypoxia and ischemia of the kidney. In addition, prolonged warm ischemia in the deceased circulatory death (DCD) donor will affect the viability of the donor kidney. These combined effects on the graft-to-be result in a cascade of renal damage that will reveal itself at the time of transplantation, when the donor kidney is reperfused in the recipient and has been named an ischemia-reperfusion injury (IRI) [2]. Typically, IRI will clinically manifest as immediate nonfunction of the transplant with the need for dialysis treatment until the graft recovers from the insult and starts eventually to function. This 'secondary' recovery is called delayed graft function.

DGF, a form of acute kidney injury post-transplantation, is an uncommon complication after living donor kidney transplantation (LDKT), most likely due to very short ischemia times and healthy living donors. Incidences reported vary between 1%–8% [3,4]. In transplantation with kidneys from deceased brain death (DBD) donors, however, the incidence of DGF increases to 15%–25% and may rise up to 72% in transplantation with kidneys from deceased DCD donors [5,6]. DGF is a risk factor for acute rejection (AR) and the combination of DGF and AR reduces graft and patient survival [7–9]. Also in the absence of AR, DGF has been shown to be an independent risk factor for long term graft loss. Reported risk factors for DGF are: deceased donor, longer ischemia times, donor and recipient older age, female donor, male recipient, history of dialysis, higher body mass index (BMI), hypertension in the donor, diabetes in the recipient, retransplantation, higher panel-reactive antibody levels, and higher human leukocyte antigens (HLA) mismatch [3,5,7,10]. This variety of risk factors underscores the complex pathological mechanisms underlying DGF.

Regarding the intraoperative period, several studies suggest that an adequate/supranormal fluid state is associated with a reduced risk of DGF [5,7,11–14]. These studies, however, are mainly retrospective and often comprise a variety of donor types with variable incidences of DGF hampering an adequate analysis. Central venous pressure (CVP)-guided fluid therapy has been suggested until recently [11,12], but CVP does not correlate well with intravascular fluid state and its use to guide fluid therapy is currently discouraged [15]. Blood pressure and heart rate are also affected by several variables, unrelated to the circulatory state of the patient, like pain, temperature, anesthetics, and analgesics, making them less suitable as an indicator of the intravascular volume [16,17].

Recently, goal directed fluid therapy (GDFT) has been shown to improve patient outcomes after major (abdominal) surgery [18–20]. During 2015, our department implemented a GDFT approach in kidney transplant recipients to replace our standard intraoperative fluid regimen of four to five liters (L) of balanced crystalloids. In the first half year of 2016 a marked increase in DGF and functional (f)DGF in our LDKT population was noticed. During 2014 and 2015, respectively, 8.5% and 8.8% of the patients experienced fDGF. From January to June 2016 the incidence of functional delayed graft function (fDGF) rose to 23.0%, which was a significant increase compared to 2014 and 2015 (*p* = 0.039 and *p* = 0.021, respectively). Since the incidence of fDGF in this population has been stable over the past two decades and no protocol changes were implemented with the exception of the GDFT protocol, we questioned whether this increase in fDGF was due to the altered fluid regimen. To our surprise, a retrospective analysis revealed that the implementation of GDFT protocol had resulted in a reduced intraoperative fluid administration which seemed associated with the increase in fDGF. Based on these results, we promptly changed the intraoperative fluid protocol in September 2016 to a fixed amount of 50 mL/kg BW with a lower limit of 2500 mL and upper limit of 6000 mL (50 kg–120 kg), unless patients comorbidity determined otherwise. After six months the incidence of fDGF was back to baseline at 8.2%.

Since we were interested in whether the amount of fluid administered intraoperatively was indeed an independent factor predicting fDGF in this LDKT population, we performed a retrospective cohort analysis of all donors and recipients in our living donor program between January 2014–February 2017.

#### **2. Materials and Methods**

#### *2.1. Study Design and Population*

This retrospective cohort analysis comprised all consecutive donor and recipient pairs of the LDKT program of the University Medical Centre of Groningen (UMCG) between January 2014 and February 2017. The Institutional Review Board approved the study (METc 201600968), which was conducted in adherence to the Declaration of Helsinki. Due to the observational and retrospective character of the analysis, the requirement for informed consent was waived.

#### *2.2. Definition of DGF*

Twenty-two definitions of DGF were identified in literature based on dialysis, serum creatinine levels, urine output or a combination of these 3 [21], Most commonly used was dialysis requirement the first week after transplantation (also used in this analysis for DGF). This dialysis-based definition, however, is criticized for its subjectivity since there are center- or physician-specific thresholds for the use of dialysis after transplantation [22]. Furthermore, since approximately half of our LDKT population was transplanted preemptively, this dialysis-based definition was unsuitable for this analysis. Another definition, referred to as functional (f)DGF, is failure of serum creatinine level to decrease spontaneously by at least 10% daily on 3 consecutive days during the first postoperative week, discounting creatinine decreases due to dialysis. Moore and colleagues showed that fDGF is independently associated with reduced death-censored graft survival in contrast to DGF based on the dialysis definition and suggested a superiority of this definition over the dialysis-based definition [23]. To prevent misclassification in patients with excellent early graft function, failure of creatinine to decrease on postoperative day three was not classified as fDGF if optimal graft function had already been achieved by day 2. In this analysis, we compared patients undergoing LDKT with fDGF and without fDGF (nofDGF).

#### *2.3. Intra- and Postoperative Management and Surgical Procedure*

Anesthetic management was according to local protocol. Propofol was used for induction of anesthesia and either propofol or sevoflurane were used for maintenance of anesthesia. Sufentanil or remifentanil were used to control nociception and rocuronium or cis-atracurium for muscle relaxation. Until the implementation of the GDFT protocol, donors and recipients were given 4–5 L of balanced crystalloids throughout the procedure unless their comorbidity determined otherwise. During 2015, a GDFT protocol was gradually implemented in the recipients (not in the donors). For a detailed description of this protocol, see below. From September 2016 fluid protocol in recipients was changed to a fixed amount of 50 mL/kg BW intraoperatively. Timeline of fluid management in recipients is given in Figure 1. Fluid management in donors was not actively changed during our observation period. Regarding the type of fluid, predominantly Ringers' lactate (RL) was used. If hyponatremia occurred RL was replaced by 0.9% saline. Colloids were not given and administration of blood products was according to our local transfusion protocol with thresholds based upon patients comorbidity. Regarding hemodynamics, the goal was to keep the blood pressure within 80% range of the baseline blood pressure of the patient. As baseline, we used blood pressure measured at the preoperative visit. If hypotension occurred, the first step was to adjust depth of anesthesia or analgesia. If that was insufficient or not possible, patients received one or more doses of ephedrine or phenylephrine or a continuous infusion of noradrenaline was started. Kidney donation was performed using a hand-assisted laparoscopic approach. Thereafter the kidney was flushed and perfused with cold University of Wisconsin solution (ViaSpan®, BMS, Bruxelles, Belgium or CoStorSol®, Bridge to Life, Elkhorn, WI, USA) and stored on

ice. Transplantation was performed according to local, standardized protocol. Postoperative fluid management comprised 1 L NaCl 0.45%-Glucose 2.5% per 24 h, complemented with the volume of diuresis in the former hour.

**Figure 1.** Timeline of various intraoperative fluid protocols in recipients. L: liters; RL: Ringers' lactate; GDFT: goal directed fluid therapy, BW: body weight.

#### *2.4. Goal-Directed Fluid Therapy Protocol.*

GDFT was performed with the use of the FloTrac®in combination with the EV1000®monitor (Edwards Lifesciences Corporation, Irvine, CA, USA). The system was used according to manufacturer's instructions. A standard institutional GDFT protocol was used with adjustment of the goal. Instead of a stroke volume variation (SVV) < 12%, commonly used in abdominal surgery, we aimed for a SVV < 10% throughout the procedure. When the SVV was >10% additional fluid was given until SVV was <10%. If SVV < 10%, fluid administration was left to the discretion of the attending anesthesiologist, however, when cardiac index (CI) was below age-adjusted normal values, a noradrenaline infusion was started. If measurement of the SVV was not possible (e.g., due to cardiac arrhythmias) a protocol based on stroke volume (SV) was used. In this case, if a fluid bolus of 250 mL resulted in an increase of the SV of 10%, additional fluid was given, if not, the trend of the SV was monitored and fluid administration was left to the discretion of the attending anesthesiologist. When SV decreased >10%, additional fluid was given. The FloTrac®was used with the EV1000 monitor, which does not communicate with our digital PDMS. Therefore SV, SVV, and CI values could not be retrieved for this analysis.

#### *2.5. Patient Data*

Demographic and postoperative data were obtained from digital patient medical records. The following variables were taken into account: age, gender, BMI, smoking, hypertension, use of antihypertensive drugs, measured glomerular filtration rate (mGFR) with use of iodine 125-iothalamate in the donor, blood pressure (measured the day of hospital admission), difference in blood pressure between donor and recipient measured by systolic/diastolic/mean of the recipient minus systolic/diastolic/mean of the donor, underlying kidney disease, number of HLA mismatches, history of dialysis, related or unrelated donor transplantation. For all recipients, the age-adjusted Charlson comorbidity index (CCI) [24] and length of hospital stay was calculated. Intraoperative data were retrieved from our digital patient data monitoring system (PDMS, CS-EZIS, Chipsoft B.V., Amsterdam, the Netherlands) and consisted of duration of surgery, intraoperative volume and type of fluid, cumulative hypotensive periods defined as a systolic blood pressure < 80 mmHg and MAP < 60 mmHg, intraoperative use of vasoactive substances, ischemia times, left/right kidney, side of implantation, number of arteries, sacrifice of an accessory artery, and urinary output the first 2 h postoperatively. Regarding the use of vasoactive substances, patients were scored on receiving one or more boluses of ephedrine and/or phenylephrine and whether or not noradrenaline was administered as a continuous infusion. Additionally, the maximum noradrenaline infusion rate during the procedure was noted. This was grouped into 3 categories: low infusion rate (0.02–0.10 mg/h), intermediate (0.10–0.20 mg/h), and high (>0.20 mg/h) infusion rate.

#### *2.6. Statistics*

For the statistical analysis SPSS version 23 (IBM Corp, Armonk, NY, USA) and GraphPad Prism version 7.02 (GraphPad Software Inc, La Jolla, CA, USA) were used. We performed univariate analyses to identify factors associated with fDGF. Categorical data were analyzed by chi-square or Fisher's exact tests. Continuous data were analyzed with an unpaired t-test in the case of normally distributed values. If variables were not normally distributed Mann–Whitney test was applied. Multivariate analysis was performed by means of binary logistic regression. We adjusted the amount of fluid administered intraoperatively in recipients for potentially relevant confounders with high significance in the univariate analysis. Additionally, we were interested in the impact of implementation of our GDFT protocol on the incidence of fDGF and on the amount of fluid administered intraoperatively. We therefore analyzed these data between the different time periods 1–3 (described above) with the use of Fisher's exact test and Kruskal–Wallis test. Post-hoc analysis with Mann–Whitney was used. Values are given as number (%), mean ± standard deviation (SD) or median with interquartile range (IQR). All reported *p*-values are two-sided. A *p*-value of 0.05 or less was considered significant.

#### **3. Results**

#### *3.1. Univariate Analysis*

#### 3.1.1. Patient Characteristics

Between January 2014 and February 2017, 275 living donor kidney transplant procedures were performed in our center. Of the 275 recipients, 31 patients experienced fDGF and 244 recipients did not (nofDGF). Donor and recipients characteristics of fDGF and nofDGF kidneys are listed in Table 1. There were no statistically significant differences in baseline characteristics and kidney function (mGFR) in donors of kidneys with our without fDGF. Recipients developing fDGF were more likely to be dialysis-dependent at the time of transplantation (25 (81%) vs. 105 (43%), *p* < 0.001). The composition of the group of dialysis dependent patients did not differ between nofDGF and fDGF recipients. In the nofDGF group 76 (72%) patients were on hemodialysis at the time of transplantation and 29 (28%) on peritoneal dialysis. In the fDGF group, this was the case for 19 (76%) and six (24%), respectively. All patients on hemodialysis were dialyzed the day before transplantation to 1 kg above dry weight.


**Table 1.** Donor and recipient demographics. Data given as number (%), mean (SD), or median (IQR).


**Table 1.** *Cont.*

fDGF: functional delayed graft function; BMI: body mass index; S-RR: systolic blood pressure; D-RR: diastolic blood pressure; MAP: mean arterial pressure; ACE-I: angiotensin-converting enzyme inhibitor; AT-II-ant: angiotensin II receptor antagonist; CCI: Charlson comorbidity index; mGFR: measured glomerular filtration rate measured with use of iodine 125-iothalamate; DM: diabetes mellitus; PKD: polycystic kidney disease; HLA: human leucocyte antigen; LURD: living unrelated donation; \*: statistically significant.

#### 3.1.2. Intra- and Postoperative Data

Intraoperative data of donors of fDGF and nofDGF kidneys showed no differences with exception of the total amount of fluid, in which donors of fDGF kidneys received less fluid intraoperatively, which was the case for total volume (3545 mL (778.2) vs. 3845 mL (799.1), *p* = 0.050) and mL/kg BW (45 mL/kg BW (10.3) vs. 49 mL/kg BW (11.4), *p* = 0.053).

Recipients who developed fDGF received significantly less intraoperative fluid, which was the case for the total amount of fluid (3000 mL (2250–3680) vs. 3500 mL (2900–4075), *p* = 0.023) and mL kg-1BW (36 mL/kg BW (25.9–50.0) vs. 47 mL/kg BW (37.3–55.6), *p* = 0.007). Predominantly RL was given, but in case of hyponatremia RL was partially replaced by saline. This was the case in 48 (20%) of the recipients without fDGF and in 8 (26%) of the patients with fDGF (*p* = 0.477). Median volume replaced by saline was 1000 mL (500–2000) in the nofDGF group and 800 mL (500–1075) in the fDGF group (*p* = 0.865). Blood loss was comparable between groups and transfusion of red blood cells was

applied in 10 (4.1%) of the patients in the noFDGF group and two (6.4%) of the fDGF group. Patients showed no difference in hypotensive periods, but recipients experiencing fDGF were treated more frequently with noradrenaline continuous infusion (*p* = 0.034), which was only the case for low dose infusion with a maximum of 0.1 mg/h. For noradrenaline administered at higher dosage (>0.1 mg/h), there was no difference between the two groups. fDGF was associated with a lower urine output during the first two hours after transplantation (*p* = 0.005 for the first hour and *p* = 0.002 for the second hour). Ten patients in the fDGF group were dialyzed after transplantation versus zero patients in the nofDGF group (*p* < 0.001). Eight of these kidneys gained function after a mean of 10.3 (3.1) days. Two kidneys suffered primary nonfunction due to a combination of ATN and mild antibody-mediated rejection (patient 114, transplanted June 2015) and non-HLA-mediated hyperacute rejection (patient 273, transplanted November 2016). Recipients experiencing fDGF showed a longer hospital stay (14 (10–20) vs. 9 (7–13) days *p* < 0.001) (Table 2).


**Table 2.** Intra- and postoperative donor and recipient data. Data given as number (%), mean (SD), or median (IQR).



Min: minutes; BW: bodyweight; S-RR: systolic blood pressure; MAP: mean arterial pressure; WIT: warm ischemia time; CIT: cold ischemia time: WIT2: warm ischemia time 2; \*: statistically significant

#### *3.2. Multivariate Logistic Regression Analysis*

In the unadjusted analysis, a higher intraoperative administered fluid volume was associated with 3% lower odds for the development of fDGF per mL/kg BW (OR 0.967, CI (0.941–0.993), model 1). We adjusted for potentially relevant confounders with high significance in the univariate analysis, i.e., a history of dialysis and the use of intraoperative noradrenaline, after which the relationship was still apparent (OR 0.970, CI (0.943–0.998), model 2). Since the intraoperative amount of fluid in the donors approached significance in the univariate analysis with lower volumes given in the fDGF group, we also adjusted for amount of fluid in the donor, after which the relationship was still apparent (OR 0.969, CI (0.941–0.997), model 3) (Table 3).

**Table 3.** Multivariate logistic regression on risk factors of functional delayed graft function (fDGF).


#### *3.3. Influence of the GDFT Protocol on the Intraoperative Fluid Volume.*

Additionally, we were interested in the impact of implementation of our GDFT protocol on the incidence of fDGF and on the amount of fluid administered intraoperatively. The GDFT protocol was gradually implemented during 2015 and in 2016 (up to September) all recipients were treated following this protocol (Figure 1). Data of the EV1000 monitor were not recorded in our PDMS, therefore we were unable to see which patients in 2015 were treated according the GDFT protocol and disregarded this period (March 2015–December 2015) in this specific analysis. We compared patients transplanted between January 2014–February 2015 (period 1, *n* = 84, old protocol) to patients transplanted between January 2016–June 2016 (period 2, *n* = 52, GDFT protocol) and patients transplanted between September 2016–February 2017 (period 3, *n* = 61, new protocol).

Incidence of fDGF during the different periods are shown in Figure 2. Implementation of GDFT was accompanied by an increase in fDGF from 8.3% in period 1 to 23% in period 2. The implementation of the new protocol in period 3 resulted in a reduction of the incidence of fDGF back to baseline (8.2%, *p* = 0.029).

Total amount of intraoperative administered fluid and mL/kg BW in recipients in the different time periods are shown in Figure 3A,B, respectively. Total amount of fluid and mL/kg BW were significantly different between the three time periods (*p* < 0.001, *p* < 0.001). Implementation of the GDFT (period 2) resulted in a decrease of intraoperative fluid administration compared to our old protocol (period 1), which was the case for total volume (2775 mL (2313–3500) vs. 3625 mL (3213–4000), *p* < 0.001) and mL/kg BW (38 mL/kg BW (30.3–45.3) vs. 48 mL/kg BW (40–60), *p* < 0.001). The implementation of the new protocol (period 3) resulted in an increase in intraoperative fluid administration to 4150 mL (3475–4575) mL and 54 mL/kg BW (47.4–60.1) compared to the old (total volume *p* = 0.037, mL/kg BW *p* = 0.053) and GDFT (total volume *p* < 0.001, mL/kg BW *p* < 0.001).

**Figure 2.** Incidence of fDGF in recipients during the different time periods. Period 1: January 2014–February 2015, old protocol, 4–5 L RL. Period 2: January–June 2016, GDFT protocol. Period 3: September 2016–February 2017, new protocol, 50 mL/kg BW. *p* = 0.029.

**Figure 3.** Volume of fluid administered intraoperatively in recipients during the different time periods. Period 1: January 2014–February 2015, old protocol, 4–5 L RL. Period 2: January–June 2016, GDFT protocol. Period 3: September 2016–February 2017, new protocol, 50 mL/kg BW. Volumes are given in mL (**A**) and mL/kg BW (**B**).

#### **4. Discussion**

This retrospective cohort analysis study shows that intraoperative fluid restriction in recipients is associated with fDGF in living donor kidney transplantation. Additionally, we showed that the implementation of a GDFT with a goal set at SVV < 10% led to a reduction of intraoperative fluid administration, on average by 850 mL in total and 21% in mL/kg BW, compared to our old protocol of 4–5 L of RL. In our opinion, this analysis provides valuable information for other centers when changes in intraoperative fluid management during kidney transplantation are considered.

Four to five liters of RL was the standard intraoperative fluid protocol in kidney transplantation in our center for over 15 years. This may seem rather liberal, but problems due to hypervolemia were rarely seen. However, following new trends on GDFT [24], a personalized intraoperative fluid approach seemed more appropriate in this group of patients presenting with a variety of fluid states at the time of surgery. Therefore, when in 2015 an intraoperative GDFT protocol was introduced in our center for several surgical procedures, we included the kidney transplant program in this implementation. Since there is no evidence in current literature on what goal to aim for, we adjusted the standard institutional GDFT protocol of SVV < 12%, commonly used in abdominal surgery, to a more generous goal in fluid administration of SVV < 10%. The implementation of this protocol resulted in a reduction in the amount of fluid administered intraoperatively in contrast to previous studies comparing GDFT to a "standard" protocol, which generally reported an increase of the amount of fluid. This could be due to the fact that most of these studies compare GDFT with a rather restrictive fluid protocol, which was general practice before GDFT was introduced. Kidney transplantation, however, has always been an exception on this restrictive trend and most centers use a rather liberal fluid protocol during this procedure. Another factor could be the performance of the FloTrac®-system in predicting fluid responsiveness in this specific patient category. GDFT and the performance of the FloTrac®-system has predominantly been validated in cardiac and abdominal surgery, liver transplantation, and septic patients. Patients with end-stage renal disease (ESRD) and especially patients on HD develop morphologic and functional cardiovascular changes. They often present with severe arterio- and atherosclerosis, inducing arterial stiffening and systolic or diastolic dysfunction. Since SVV is calculated as the percentage change of SV to the mean, derived from an arterial pulse contour analysis, it is conceivable that these cardiovascular changes influence the performance of the FloTrac®-system in predicting patients fluid state. Only one pilot study presents the effect of fluid loading on SVV measured with the use of the FloTrac®-system in patients with ESRD on HD. In this study, HD patients undergoing vascular surgery presented with a broad range of SVV (16.2 ± 6.0) after

induction of anesthesia. After a fluid bolus of only 500 mL of a colloid solution almost all patients showed a SVV < 10% (6.2 ± 2.8), the threshold in our protocol [25].

The debate on perioperative fluid management is still ongoing. Controversy exists regarding assessment of the intravascular volume state, which goals to aim for, how to measure these goals, and what type of fluid should be used. Hypovolemia leads to a decreased oxygen supply to organs and tissues and may cause hypoxia, which can lead to organ dysfunction. Hypervolemia, on the other hand, can damage the endothelial glycocalyx resulting in a fluid shift from the intravascular compartment to the interstitial space and tissue edema [26]. Shin and colleagues report in their large cohort analysis of 92.094 patients undergoing noncardiac surgery that both too little and too much intraoperative fluid is associated with increased morbidity, mortality, costs, and length of hospital stay [27]. Myles and colleagues randomLy assigned 3000 patients undergoing a major abdominal procedure to a restrictive or liberal fluid regimen. In their study, a restrictive regimen was associated with increased risk of acute kidney injury with a hazard ratio of 1.71 (95% CI 1.29–2.27) [28]. These studies, however, do not take kidney transplant recipients into account. In the normal kidney, blood flow is regulated by an autoregulatory mechanism, ensuring adequate perfusion in a broad blood pressure range by afferent and efferent arterioles. In the transplanted, denervated kidney, this haemodynamic autoregulation is impaired making the renal blood flow linearly dependent on the systemic blood flow [29–31]. Furthermore, reperfusion of the ischemic kidney can be followed by vasoconstriction in the afferent arterioles. This may result in a reduced GFR due to a decrease in glomerular transcapillary hydraulic pressure difference [7,32,33]. Ensuring an adequate volume state in this specific patient category, therefore, is essential to obtain an adequate circulation both on macro- and microcirculatory level. Recently, Cavalari and colleagues reported the results of their prospective observational study, in which they compared a prospectively observed cohort of 33 deceased donor kidney transplant recipients treated with a GDFT protocol to a historical cohort of 33 kidney transplant recipients treated with their conventional fluid therapy [34]. They observed a significant reduction of cardiovascular complications, DGF. and surgical complications in the GDFT group. Surprisingly, in this study both groups received the same amount of fluid throughout the transplant procedure. Studies including deceased donor kidneys, however, comprise a variety of donor types with variable incidences of DGF hampering an adequate analysis and conclusions.

The most important predictor of fDGF in our analysis was dialysis dependency at the time of transplantation. A history of dialysis and especially hemodialysis prior to transplantation is a known risk factor of DGF [5,7,35,36]. Hypovolemia at the time of transplantation is one of the proposed underlying mechanisms [37]. Our hypothesis before implementation of the GDFT protocol was that these hypovolemic dialysis patients would present with higher SVV at time of surgery, demanding more fluid intraoperatively, compared to the relatively normovolemic or slightly hypervolemic preemptively transplanted patients. Surprisingly, comparable amounts of fluids were given to the two groups.

In our GDFT protocol, noradrenaline was used when CI was below an age-adjusted value. Therefore, an increased use of noradrenaline was seen in period 2 compared to period 1 (71% vs. 41% *p* = 0.001) due to the implementation of the GDFT. In period 3, the use of noradrenaline decreased to 50% of the patients. In the univariate analysis, the use of noradrenaline was correlated with development of fDGF, but after multivariate logistic regression this was no longer the case. However, Morita and coworkers showed that in a rat model, transplanted kidneys responded to sympaticomimetics with a reduction in renal blood flow (RBF) in contrast to the increase in RBF seen in native rat kidneys [38].

There are some limitations of this analysis that have to be addressed: A major limitation is that we were unable to evaluate outcome directly according to the fluid protocol (4–5L RL vs. GDFT) and are unable to present information or draw any conclusions regarding actual SV, SVV, CO or CI values and their relation to the observed increase of fDGF. Other limitations are those of a retrospective observational trial. There is the potential of confounding by unmeasured factors. Regarding postoperative fluid volume, the exact amount of fluid given could not be retrieved in a reliable way from our PDMS and is therefore not implemented in this analysis. Postoperative fluid

management was according to a standardized protocol and comprised of 1 L NaCl 0.45%-Glucose 2.5% per 24 h, complemented with the volume of diuresis in the former hour. This means that when the kidney produces less urine the patient will be given less fluid postoperatively. Since fDGF was associated with a lower urinary output the first two hours, it is very likely that patients experiencing fDGF received less fluid postoperatively. Whether this contributed to development of fDGF or is more of a symptom remains unknown. Backpressure from congested tubules obstructed with cellular debris may contribute to a reduction in GFR [39,40]. A higher volume of urine in the first hours may have led to washout of this debris.

Finally, due to the fact that there are only 31 events there is always the possibility of overestimating the strength of associations using a multivariate analysis. A strong argument, however, is that no policy changes were implemented during the study period with the exception of the intraoperative fluid regimen. Furthermore the incidence of fDGF in our LDKT population has been stable over many years and after changing the fluid regimen back to a more liberal fixed amount of 50 mL/kg BW the incidence of fDGF instantly returned to baseline.

DGF after transplantation is a clinically relevant problem. It is associated with an increase in morbidity, patient anxiety, increased risk of acute rejection, and additional diagnostic procedures and costs. In our population the median hospital stay in patients experiencing fDGF was prolonged by five days. Furthermore, this study shows that strict protocols for perioperative fluid management are needed when studies in kidney transplantation are designed. Fluid restriction can be an important risk factor for DGF, a frequently used primary end point, even in the setting of LDKT.

#### **5. Conclusions**

Implementation of a goal-directed approach to fluid administration with a goal set at a SVV < 10% throughout the procedure led to reduced intraoperative fluid administration in the LDKT recipients in our center. This intraoperative fluid restriction was associated with the development of more fDGF. A thorough validation of GDFT protocols in patients with renal insufficiency is warranted before these are implemented in this population.

**Author Contributions:** G.J.N.-M.: participated in research design, data collection, data analysis, writing of the paper. T.M.H.: participated in data collection, writing of the paper; R.A.P.: participated in writing of the paper; M.E.M.: participated in data analysis, writing of the paper; J.G.B.: participated in data analysis, writing of the paper; M.M.S.: participated in research design, data analysis, writing of the paper; S.P.B.: participated in research design, data analysis, writing of the paper.

**Funding:** This research received no external funding.

**Acknowledgments:** We would like to thank Cordelia Hempel and Tamar van den Berg for data retrieval and professor Anthony Absalom for his critical comments and help with language aspects.

**Conflicts of Interest:** The authors declare no conflict of interest

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Correction* **Correction: Fast Tac Metabolizers at Risk—It is Time for a C**/**D Ratio Calculation.** *J. Clin. Med.* **2019,** *8***, 587**

**Katharina Schütte-Nütgen 1,**†**, Gerold Thölking 1,**†**, Julia Steinke 1, Hermann Pavenstädt 1, René Schmidt 2, Barbara Suwelack <sup>1</sup> and Stefan Reuter 1,\***


Received: 31 October 2019; Accepted: 1 November 2019; Published: 4 November 2019

The authors wish to make the following corrections to this paper [1].

The authors made an error regarding the rejection-free survival curve in Figure 4A. Figure 4 needs to be corrected.

should be replaced with

The authors apologize to the readers for any inconvenience caused by these changes. It is important to state that this correction do not affect our study's results and involve no changes or modifications in the original data supporting our results. The original manuscript will remain online on the article webpage, with reference to this Correction.

#### **Reference**

1. Schütte-Nütgen, K.; Thölking, G.; Steinke, J.; Pavenstädt, H.; Schmidt, R.; Suwelack, B.; Reuter, S. Fast Tac Metabolizers at Risk—It is Time for a C/D Ratio Calculation. *J. Clin. Med.* **2019**, *8*, 587. [CrossRef] [PubMed]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Proton-Pump Inhibitors and Hypomagnesaemia in Kidney Transplant Recipients**

**Rianne M. Douwes 1,\*, António W. Gomes-Neto 1, Joëlle C. Schutten 1, Else van den Berg 1, Martin H. de Borst 1, Stefan P. Berger 1, Daan J. Touw 2, Eelko Hak 3, Hans Blokzijl 4, Gerjan Navis <sup>1</sup> and Stephan J. L. Bakker <sup>1</sup>**


Received: 30 October 2019; Accepted: 4 December 2019; Published: 6 December 2019

**Abstract:** Proton-pump inhibitors (PPIs) are commonly used after kidney transplantation and there is rarely an incentive to discontinue treatment. In the general population, PPI use has been associated with hypomagnesaemia. We aimed to investigate whether PPI use is associated with plasma magnesium, 24-h urinary magnesium excretion and hypomagnesaemia, in kidney transplant recipients (KTR). Plasma magnesium and 24-h urinary magnesium excretion were measured in 686 stable outpatient KTR with a functioning allograft for ≥1 year from the TransplantLines Food and Nutrition Biobank and Cohort-Study (NCT02811835). PPIs were used by 389 KTR (56.6%). In multivariable linear regression analyses, PPI use was associated with lower plasma magnesium (β: −0.02, *P* = 0.02) and lower 24-h urinary magnesium excretion (β: −0.82, *P* < 0.001). Moreover, PPI users had a higher risk of hypomagnesaemia (plasma magnesium <0.70 mmol/L), compared with non-users (Odds Ratio (OR): 2.12; 95% confidence interval (CI) 1.43–3.15, *P* < 0.001). This risk tended to be highest among KTR taking high PPI dosages (>20 mg omeprazole Eq/day) and was independent of adjustment for potential confounders (OR: 2.46; 95% CI 1.32–4.57, *P* < 0.005). No interaction was observed between PPI use and the use of loop diuretics, thiazide diuretics, tacrolimus, or diabetes (*P*interaction > 0.05). These results demonstrate that PPI use is independently associated with lower magnesium status and hypomagnesaemia in KTR. The concomitant decrease in urinary magnesium excretion indicates that this likely is the consequence of reduced intestinal magnesium absorption. Based on these results, it might be of benefit to monitor magnesium status periodically in KTR on chronic PPI therapy.

**Keywords:** proton-pump inhibitors; magnesium; hypomagnesaemia; kidney transplantation

#### **1. Introduction**

Proton-pump inhibitors (PPIs) are frequently used after kidney transplantation for their gastro- protective properties in the setting of immunosuppressive therapy, which usually includes glucocorticoids. Since their first introduction in the late 1980s, numerous case reports and observational studies have been published that associate PPI use with unfavorable clinical outcomes, including an increased risk of hypomagnesaemia [1–8]. Recently, this observation has been strengthened by a large population based cohort study which demonstrated a two times higher risk of hypomagnesaemia among subjects from the general populations on chronic PPI therapy compared to non-users [9].

Magnesium homeostasis depends mainly on the balance between intestinal Mg2<sup>+</sup> uptake, storage and resorption from bones and urinary excretion of Mg2<sup>+</sup> via the kidneys [10]. It is postulated that PPIs induce hypomagnesaemia through inhibition of pH-dependent active magnesium absorption via transient receptor potential melastatin (TRPM) 6 and 7 channels in the intestine [11,12]. Moreover, increased renal magnesium retention has been observed in magnesium depleted subjects using chronic PPI therapy, indicating a defect in intestinal magnesium absorption or increased losses into the gastrointestinal tract, rather than renal magnesium wasting [1,7,13].

Hypomagnesaemia is very common after kidney transplantation and it is generally thought to be a side effect of immunosuppressive therapy, especially of calcineurin inhibitors (CNI) which are known to induce renal magnesium wasting [14]. It has been shown that hypomagnesaemia is not only present in the immediate post transplantation period, but persists in about 20% of kidney transplant recipients (KTR) for many years after transplantation [15,16]. Importantly, hypomagnesaemia has been associated with onset of post-transplant diabetes mellitus (PTDM) in KTR [17,18] and has also been associated with increased risk of cardiovascular morbidity [19,20] and mortality [21] in the general population. Whether use of PPIs contributes to hypomagnesaemia in KTR has not been well established. To our knowledge only one cohort study investigated the association between PPI use and hypomagnesaemia in 512 KTR, with negative results [22]. Reasons for absence of an association were unclear, but may have included a low prevalence of PPI use of 20%, which could have led to low statistical power of the study. Thus, whether PPI use negatively affects magnesium status after transplantation remains to be determined. We aimed to investigate whether PPI use is associated with magnesium status and hypomagnesaemia in a large single center cohort of stable outpatient KTR, in which plasma magnesium measurements were not part of routine clinical care but were assessed from samples that had been stored in a biobank.

#### **2. Methods**

#### *2.1. Study Design and Population*

This is a cross-sectional analysis using data from a previously described prospective cohort study registered at clinicaltrials.gov as 'TransplantLines Food and Nutrition Biobank and Cohort-study', NCT02811835 [23]. In summary, all adult KTR with a functioning graft beyond the first year after transplantation and without known or apparent systemic illnesses (i.e., malignancies other than cured skin cancer, opportunistic infections, overt congestive heart failure) who visited the outpatient clinic of the University Medical Center Groningen (UMCG) between November 2008 and March 2011, were asked to participate. A total of 707 out of the initially 817 invited KTR signed informed consent. We excluded KTR with missing biomaterial (*n* = 8), missing data on PPI dosage (*n* = 1), with on-demand PPI use (*n* = 3) or using magnesium supplements (*n* = 6) from statistical analyses, leaving 689 cases eligible for analysis. Study measurements were performed during a single study visit at the outpatient clinic. All study procedures were conducted in adherence with the Declaration of Helsinki and Declaration of Istanbul. The institutional review board of the UMCG approved the study protocol (METC 2008/186, approved on 17 September 2008).

#### *2.2. Exposure Definition*

PPI type and daily dosage were obtained from electronic patient records and are demonstrated in Table S1. KTR using any PPI on a daily basis during a period of at least 3 months prior to the study visit were defined as chronic PPI users as described previously [24]. To investigate a potential dose–response relationship, KTR were divided into three groups based on daily PPI dose defined in omeprazole equivalents: no PPI, low PPI dose (≤20 mg omeprazole equivalents/day (Eq/day)) and high PPI dose (>20 mg omeprazole Eq/day) [24,25].

#### *2.3. Assessement of Plasma and Urinary Magnesium*

Plasma magnesium was measured in samples containing lithium heparin, using a xylidyl blue method (Roche Modular analyzer, Roche Diagnostics, Mannheim, Germany). Urinary magnesium excretion was assessed in 24 h-urine samples and measured on a MEGA clinical chemistry analyzer (Merck, Darmstadt, Germany). Hypomagnesaemia was defined as plasma magnesium <0.70 mmol/L.

#### *2.4. Assessment of Dietary Magnesium Intake*

Dietary magnesium intake was calculated using a validated semi quantitative food frequency questionnaire (FFQ) developed and updated at the Wageningen University, which was filled out at home [26,27]. Dietary data were converted into daily nutrient intake using the Dutch Food Composition Table of 2006 [28].

#### *2.5. Assessment of Covariates*

Medical history was obtained from electronic patient records as described previously [23]. History of cardiovascular disease was classified according to the International Classification of Diseases, 10th revision (ICD-10) code Z86.7. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Blood pressure was measured as described in detail previously [29]. Information on alcohol use and smoking behavior was obtained using a questionnaire. Medication use, including the use of PPIs, H2-receptor antagonists, diuretics, prednisolone, mycophenolate mofetil (MMF), tacrolimus, cyclosporine, and sirolimus was recorded at baseline. Routine immunosuppressive therapy consisted of: A combination of azathioprine and prednisolone from 1968 to 1989; a combination of cyclosporine and prednisolone from 1989 to 1996. In 1997 mycophenolate motefil was added to the standard immunosuppressive regimen and cyclosporine was slowly withdrawn after the first year in KTR without complications. In 2012 cyclosporine was replaced by tacrolimus, and KTR received triple-immunosuppressive therapy with prednisolone, tacrolimus and mycophenolate mofetil. PPIs were routinely prescribed after kidney transplantation for their gastro-protective properties with concurrent use of prednisolone. Blood samples were collected after an 8–12 h fasting period. Serum creatinine was measured using an enzymatic, isotope dilution mass spectrometry-traceable assay (P-Modular automated analyzer, Roche Diagnostics, Mannheim, Germany). Estimated glomerular filtration rate (eGFR) was calculated using the serum creatinine based Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation. Serum potassium, calcium, parathyroid hormone (PTH), glucose and hemoglobin A1c (HbA1c), were determined using standard laboratory methods. Proteinuria was defined as urinary protein excretion ≥0.5 g/24 h.

#### *2.6. Statistical Analyses*

Statistical analyses were performed using SPSS, version 23.0 (IBM corp., Armonk, NY, USA). Data are presented as mean ± SD for normally distributed data, median with interquartile range (IQR) for skewed data and number with percentage for nominal data. Differences between PPI users versus PPI non-users were tested using independent sample *T*-tests, Mann–Whitney U-tests and Chi-square tests or Fishers exact tests when appropriate.

To study the effect of PPI use on plasma magnesium linear regression analyses were performed with adjustment for potential confounders of magnesium status including: age, sex, BMI, eGFR, proteinuria, time since transplantation, alcohol use, diabetes, history of cardiovascular disease, use of loop diuretics, thiazide diuretics, tacrolimus, cyclosporine, MMF, and dietary magnesium intake. To investigate the association between PPI use and hypomagnesaemia we performed logistic regression analyses with adjustment for the same potential confounders used in multivariable linear regression analyses. Effect modification by loop diuretics, thiazide diuretics, tacrolimus and diabetes was tested

by inclusion of interaction terms. To investigate a potential dose–response relationship we performed additional analyses in which KTR were divided into three groups based on daily PPI dose defined in omeprazole equivalents: No PPI, low PPI dose (≤20 mg omeprazole Eq/day) and high PPI dose (>20 mg omeprazole Eq/day) [24,25]. Tests of linear trend were conducted by assigning the median of daily PPI dose equivalents in subgroups treated as a continuous variable. We performed sensitivity analyses in which H2-receptor antagonist (H2RA) users (*n* = 18) were excluded to assess the robustness of the association between PPI use and hypomagnesaemia. Lastly, we investigated which KTR are at increased risk of developing hypomagnesaemia. A two-sided *P*-value < 0.05 was considered statistically significant in all analyses.

#### **3. Results**

#### *3.1. Baseline Characteristics*

Baseline characteristics are shown in Table 1. PPIs were used by a small majority of 389 (56.5%) KTR and omeprazole was the most often prescribed PPI (*n* = 340). Other PPIs used were esomeprazole (*n* = 30), pantoprazole (*n* = 16), and rabeprazole (*n* = 3). KTR who used PPIs were older than KTR who did not use PPIs, had a higher BMI and had shorter time between transplantation and baseline measurements. Diabetes was significantly more prevalent in PPI users compared with non-users (28.3% vs. 18.3%, *P* < 0.002). Plasma magnesium and 24-h urinary magnesium excretion were significantly lower in PPI users and 102 (26.2%) PPI users had hypomagnesaemia compared with 43 (14.3%) non-users (*P* < 0.001). Dietary magnesium intake was not significantly different between PPI users and non-users. Loop diuretics, cyclosporine and MMF, were more often used by PPI users compared with non-users. Triple immunosuppressive therapy consisting of MMF, cyclosporine and prednisolone, was more common in PPI users compared with non-users. Duo therapy consisting of MMF-prednisolone, MMF-cyclosporine, and cyclosporine-prednisolone was more common in PPI users compared with non-users.


**Table 1.** Baseline characteristics of 689 kidney transplant recipients.


**Table 1.** *Cont.*

Data are presented as mean ± SD, median with interquartile ranges (IQR) or number with percentages (%). Abbreviations: BMI, body mass index; eGFR, estimated glomerular filtration rate; HbA1c, hemoglobin A1c; PTH, Parathyroid hormone; MMF, mycophenolate mofetil; Tac, tacrolimus; Pred, prednisolone.

#### *3.2. Association of PPI Use with Plasma Magnesium and 24-h urinary Magnesium Excretion*

PPI use was significantly associated with lower plasma magnesium (β = −0.03; 95% CI −0.04; −0.01 mmol/L, *P* = 0.001) and lower urinary magnesium excretion (β = −0.86; 95% CI −1.10; −0.06 mmol/24 h, *P* < 0.001) as compared to non-users, Table 2. After adjustment for potential confounders, PPI use remained significantly associated with lower plasma magnesium levels (β = −0.02, 95% CI −0.04; −0.003, *P* = 0.02) and 24-h urinary magnesium excretion (β = −0.82, 95% CI −1.07; −0.57, *P* < 0.001).

**Table 2.** Association of proton-pump inhibitor (PPI) use with plasma magnesium and 24-h urinary magnesium excretion in 689 kidney transplant recipients.


Multivariable analyses were adjusted for age, sex, BMI, eGFR, proteinuria, time since transplantation, alcohol use, diabetes, history of CV disease, loop diuretics, thiazide diuretics, tacrolimus use, cyclosporine use, MMF use and dietary magnesium intake. Abbreviations: CI, confidence interval.

#### *3.3. Association of PPI Use with Hypomagnesaemia*

In crude logistic regression analysis, PPI use was associated with a more than two times higher risk of hypomagnesaemia compared with no use (OR: 2.12; 95% CI 1.43–3.15, *P* < 0.001), as shown in Table 3. The association remained independent of adjustment for potential confounders including age, sex, eGFR, proteinuria, time since transplantation, alcohol use, diabetes, history of cardiovascular disease, medication use (loop diuretics, thiazide diuretics, tacrolimus, cyclosporine and MMF) and dietary magnesium intake (OR: 2.00; 95% CI 1.21–3.31, *P* = 0.007). No significant interaction was observed between PPI use and the use of loop diuretics, thiazide diuretics, tacrolimus, or diabetes for the association with hypomagnesaemia (*P*interaction = 0.2, *P*interaction = 0.7, *P*interaction = 0.7, *P*interaction = 0.9, respectively).


**Table 3.** Logistic regression analyses investigating the association of PPI use with hypomagnesaemia in 689 kidney transplant recipients.

Multivariable analyses were adjusted for age, sex, BMI, eGFR, proteinuria, time since transplantation, alcohol use, diabetes, history of CV disease, loop diuretics, thiazide diuretics, tacrolimus use, cyclosporine use, MMF use and dietary magnesium intake. Abbreviations: CI, confidence interval.

#### *3.4. Dose–Response Analyses*

Based on daily dose equivalents of omeprazole, 251 KTR received a low PPI dose (≤20 mg omeprazole Eq/day) and 138 KTR received a high PPI dose (>20 mg omeprazole Eq/day). As shown in Table 4 and Figure 1, risk of hypomagnesaemia tended to be highest among KTR taking a high PPI dose (OR: 2.53; 95% CI 1.55–4.11, *P* < 0.001). The association remained materially unchanged after multivariable adjustment (OR: 2.46; 95% CI 1.32–4.57, *P* < 0.005), Table 4. Moreover, a significant trend between PPI dose and risk of hypomagnesaemia was observed (*P*trend = 0.004).

**Table 4.** Subgroup analyses of the association of PPI use with hypomagnesaemia in 689 kidney transplant recipients.


Multivariable analyses were adjusted for age, sex, BMI, eGFR, proteinuria, time since transplantation, alcohol use, diabetes, history of CV disease, loop diuretics, thiazide diuretics, tacrolimus use, cyclosporine use, MMF use, dietary magnesium intake. Low PPI dose (≤20 mg omeprazole Eq/day), High PPI dose (>20 mg omeprazole Eq/day). Abbreviations: CI, confidence interval.

**Figure 1.** Crude association between PPI use and risk of hypomagnesaemia stratified by subgroups of PPI use. No PPI, Low PPI dose (≤20 mg omeprazole Eq/day), High PPI dose (>20 mg omeprazole Eq/day). Presented are odds ratio's with 95% confidence intervals. \* *P* = 0.004; \*\* *P* < 0.001; *P*trend < 0.001.

#### *3.5. Sensitivity Analyses for Risk of Hypomagnesaemia*

To account for the use of other important gastric acid reducing medication, we performed sensitivity analyses in which H2RA users (*N* = 18) were excluded form statistical analyses (Table S2). The association between PPI use and hypomagnesaemia remained materially unchanged when H2RA users were excluded (OR: 2.17, 95% CI 1.29–3.67, *P* = 0.004). We also performed analyses to investigate which KTR are at increased risk of developing hypomagnesaemia. These analyses are presented in Table S3. We found that patients with a history of cardiovascular disease, patients at shorter time after transplantation, not consuming alcohol, PPI users, thiazide diuretic users and patients using tacrolimus based immunosuppressive regimens were at increased risk of developing hypomagnesaemia. Moreover, KTR with hypomagnesaemia had higher fasting glucose levels, HbA1c and lower serum calcium levels compared with KTR without hypomagnesaemia.

#### **4. Discussion**

The present study is to our knowledge the largest cohort study to date exploring the association between PPI use and hypomagnesaemia in a cohort of KTR. Our results demonstrate a higher risk of hypomagnesaemia among KTR using PPIs, with subsequently lower plasma magnesium levels in combination with lower renal magnesium excretion. The association between PPI use and risk of hypomagnesaemia remained significant after adjustment for important potential confounders and tended to be highest among KTR taking high PPI dosages.

Our results confirm previous case-series and cohort studies investigating the association between PPI use and increased risk of hypomagnesaemia [1,2,7,9]. In a large cohort study (*N* = 9818) among subjects from the general population, it was shown that PPI users had significantly lower serum magnesium levels and had a two times higher risk of hypomagnesaemia compared with non-users [9]. Our results are in line with observations from this large cohort study and show a similar increased risk of hypomagnesaemia (OR 2.12).

So far, only one other study by van Ende et al. investigating the association between PPI use and magnesium status in KTR has been published [22]. Contrary to our findings, van Ende et al. found no association between PPI use and serum magnesium levels. Reasons for the lower proportion of PPI users in the study by van Ende et al. are unclear, though underreporting may have played a role, given that it was not specified how data regarding PPI use was obtained. It was also unclear whether the data of van Ende et al. were derived from routine outpatient assessment of plasma magnesium concentrations, which may have provided an incentive for stopping PPI use in KTR with low magnesium concentrations. This could have biased their results and could possibly also explain the large difference in PPI use between our study and their study, because in our center no plasma magnesium data were available at the time of the study. It was furthermore unclear whether it concerned on-demand or chronic PPI use. Furthermore, data regarding PPI dose, type and magnesium supplementation were not reported, which may have influenced the outcome.

In a recently published meta-analysis, a similar risk of hypomagnesaemia among KTR was demonstrated (pooled OR = 1.56, 95% CI 1.19–2.05) [30]. This meta-analysis by Boonpheng et al. was based on one published paper and seven abstracts presented at medical conferences. Our study adds that it investigated a dose–response relationship, and provides data on dietary magnesium intake and 24-h urinary magnesium excretion.

In the present study, both plasma magnesium and 24-h urinary magnesium excretion were lower in PPI users, suggesting that PPI induced hypomagnesaemia is caused by impaired gastrointestinal absorption rather than renal magnesium wasting. In general, hypomagnesaemia can be the consequence of either a decreased intestinal uptake, a decrease in dietary magnesium intake or an increase in renal magnesium excretion. It is postulated that PPIs inhibit the active magnesium absorption via the TRPM 6 and 7 channels in the intestine [11,12]. In KTR other contributing factors than PPI use may add to the risk of hypomagnesaemia. For example, decreased intestinal magnesium absorption can also be the consequence of chronic post-transplant diarrhea, which is highly prevalent and often complicated by hypomagnesaemia [10,31]. Data regarding symptoms of severe diarrhea were unfortunately unavailable in this study, therefore we could not correct for this potential confounder. Likewise, hypomagnesaemia can be the result of insufficient intake of foods rich in magnesium. In our study, mean dietary magnesium intake was 329.9 ± 88.7 mg/day, which was slightly lower than the mean habitual intake of magnesium among the general Dutch population, as reported in the Dutch National Food Consumption Survey 2007–2010 [32]. A low dietary magnesium intake can also be a reflection of an overall poor diet. Nonetheless, when we adjusted for dietary magnesium intake in our logistic regression analyses, the relationship between PPI use and risk of hypomagnesaemia remained materially unchanged, indicating that the observed risk associated with PPI use was not confounded by dietary magnesium intake.

The main strength of this study is measurement of three important pillars of magnesium status: plasma magnesium, 24-h urinary magnesium excretion and dietary magnesium intake. Because of this, we were able to confirm that PPI use does not lead to increased renal magnesium wasting but very likely impairs intestinal magnesium absorption. Furthermore, we only included KTR who were using PPIs for at least 3 months before blood sampling. It is previously noted that hypomagnesaemia occurs mainly in patients on prolonged PPI therapy suggesting that it takes time before magnesium stores are meaningfully depleted [6,7,33]. Moreover, we excluded KTR using magnesium supplements and adjusted for potential confounders, including CNI use, which did not alter the association.

A limitation of our study is its cross-sectional design. Therefore, a causal relationship between PPI use and hypomagnesaemia remains to be determined and changes over time in magnesium status parameters were unknown. Furthermore, no information regarding compliance to PPI treatment was available, which may have led to underestimation of effect sizes. PPI users had a shorter time between transplantation and baseline measurements. However, adjustment for time since transplantation did not alter the association between PPI use and hypomagnesaemia. Lastly, the possibility of residual confounding or bias by indication remains, which may have led to overestimation of the role of PPIs since on average PPI users were less healthy than non-users. A strength of the current study is, that no routine outpatient monitoring of plasma magnesium was performed and that we measured plasma and urine magnesium in samples that had been stored in a biobank, which reduces the change of selection bias in our cohort.

Our findings may be of clinical importance. KTR with low magnesium levels seem to develop post-transplant diabetes mellitus (PTDM) more frequently [17]. In this study we also found that KTR with hypomagnesaemia had higher fasting glucose levels and HbA1c. Next to that, a higher degree of arterial stiffness, as assessed by a carotid-femoral pulse wave velocity (PWV) measurement, has been found in KTR with low magnesium levels [34]. This same PWV measurement was found to be an independent predictor of cardiovascular events in KTR [35]. Moreover, hypomagnesaemia has been associated with cardiovascular morbidity [19,20] and mortality [21] in the general population. However, whether this association is also present in KTR is currently unknown. Another clinical significance lies in the association with lower calcium levels, which potentially points to an increased risk of developing osteoporosis. Long-term PPI use has indeed been associated with decreased bone mineral density and increased risk of fractures [36]. Because many patients use PPIs without evidence based indication [37–39], we believe that reevaluation of treatment indication in KTR on chronic PPI therapy might be of benefit. In situations in which PPIs are clinically needed, it would be judicious to assess and follow-up magnesium levels periodically during treatment, as recommended by the US Food and Drug administration and stated in the summary of product characteristics of all PPIs.

#### **5. Conclusions**

This study demonstrates that PPI use is associated with lower magnesium status and hypomagnesaemia in KTR. Moreover, risk of hypomagnesaemia was higher among KTR taking a high PPI dosage. Healthcare professionals should be aware of this additional risk and should consider regular monitoring of magnesium levels, especially in this patient population at high risk of hypomagnesaemia.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/8/12/2162/ s1, Table S1: Types and daily dosages of proton-pump inhibitors used by 389 kidney transplant recipients. Table S2: Logistic regression analyses investigating the association of PPI use with hypomagnesaemia in 617 kidney transplant recipients (H2RA users excluded). Table S3: Baseline characteristics of 689 RTR with and without hypomagnesaemia.

**Author Contributions:** Data curation, R.M.D., A.W.G.-N., E.v.d.B., and S.J.L.B.; Formal analysis, R.M.D., A.W.G.-N., J.C.S. and S.J.L.B.; Methodology, R.M.D, A.W.G.-N., J.C.S., E.v.d.B., M.H.d.B., S.P.B., D.J.T., E.H., H.B., G.J.N. and S.J.L.B. writing—original draft preparation, R.M.D., A.W.G.-N.; writing—review and editing, R.M.D., A.W.G.-N., J.C.S., M.H.d.B., E.v.d.B., S.P.B., D.J.T., E.H., H.B., G.J.N. and S.J.L.B. Supervision, H.B., G.N., S.J.L.B.; Funding acquisition, R.M.D., E.v.d.B. and S.J.L.B.

**Funding:** Generation of this study was funded by Top Institute Food and Nutrition (grant A-1003). R.M. Douwes is supported by NWO/TTW in a partnership program with DSM, Animal Nutrition and Health, The Netherlands; grant number: 14939.

**Conflicts of Interest:** M.H. de Borst is the principal investigator of a clinical trial supported by the Dutch Kidney Foundation (grant no 17OKG18) and Vifor Fresenius Medical Care Renal Pharma. He received consultancy fees (to employer) from Amgen, Astra Zeneca, Bayer, Kyowa Kirin, Sanofi Genzyme, and Vifor Fresenius Medical Care Renal Pharma. D.J. Touw has received grants (to employer) from ZONMw, Astellas and Chiesi Pharmaceuticals for research outside this project. None of these entities had any role in the design, collection, analysis, and interpretation of data for the current study, nor in writing the report or the decision to submit the report for publication. The other authors declare that they have no other relevant financial interests.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Urinary Biomarkers** α**-GST and** π**-GST for Evaluation and Monitoring in Living and Deceased Donor Kidney Grafts**

**Shadi Katou 1,\*, Brigitta Globke 2, M. Haluk Morgul 1, Thomas Vogel 1, Benjamin Struecker 1, Natalie Maureen Otto 3, Anja Reutzel-Selke 2, Marion Marksteiner 2, Jens G. Brockmann 1, Andreas Pascher <sup>1</sup> and Volker Schmitz <sup>1</sup>**


Received: 22 October 2019; Accepted: 5 November 2019; Published: 7 November 2019

**Abstract:** The aim of this study was to analyze the value of urine α- and π-GST in monitoring and predicting kidney graft function following transplantation. In addition, urine samples from corresponding organ donors was analyzed and compared with graft function after organ donation from brain-dead and living donors. Urine samples from brain-dead (*n*=30) and living related (*n*=50) donors and their corresponding recipients were analyzed before and after kidney transplantation. Urine αand π-GST values were measured. Kidney recipients were grouped into patients with acute graft rejection (AGR), calcineurin inhibitor toxicity (CNI), and delayed graft function (DGF), and compared to those with unimpaired graft function. Urinary π-GST revealed significant differences in deceased kidney donor recipients with episodes of AGR or DGF at day one after transplantation (*p* = 0.0023 and *p* = 0.036, respectively). High π-GST values at postoperative day 1 (cutoff: >21.4 ng/mg urine creatinine (uCrea) or >18.3 ng/mg uCrea for AGR or DGF, respectively) distinguished between rejection and no rejection (sensitivity, 100%; specificity, 66.6%) as well as between DGF and normal-functioning grafts (sensitivity, 100%; specificity, 62.6%). In living donor recipients, urine levels of α- and π-GST were about 10 times lower than in deceased donor recipients. In deceased donors with impaired graft function in corresponding recipients, urinary α- and π-GST were elevated. α-GST values >33.97 ng/mg uCrea were indicative of AGR with a sensitivity and specificity of 77.7% and 100%, respectively. In deceased donor kidney transplantation, evaluation of urinary α- and π-GST seems to predict different events that deteriorate graft function. To elucidate the potential advantages of such biomarkers, further analysis is warranted.

**Keywords:** kidney transplantation; urinary biomarkers; α-GST; π-GST; acute rejection; delayed graft function; nephrotoxicity

#### **1. Introduction**

Kidney transplantation is by far the best therapeutic option for patients with end-stage renal disease (ESRD). After transplantation, the main challenges, besides surgical complications, are acute graft rejection, delayed graft function, and adverse effects of immunosuppressants [1]. Acute graft rejection still occurs in up to 25% of recipients and is a significant prognostic factor for long-term graft survival [2]. The improvements of immunosuppressive drugs have turned transplantation into a safe and widely predictable therapy; however, many of the agents used today still contribute to graft failure due to their nephrotoxic potential [3]. Delayed graft function, defined by the need for dialysis within the first week after transplantation and mainly caused by acute tubular necrosis, is mostly due to long ischemia times, advanced donor age, and comorbidities [4,5]. Recognizing the cause of graft dysfunction may be challenging, yet immediate diagnosis and therapy are essential for optimal graft survival.

The signs of graft dysfunction are decreased diuresis and impaired creatinine blood levels. Monitoring immunosuppressive drugs and their toxicity through serum levels is of limited value, since the difference between therapeutic and toxic levels is not fixed [6]. In order to define the pathomechanism of graft dysfunction, a graft biopsy is required in most cases. However, this is an invasive procedure and risks associated complications endangering the transplanted kidney [7].

α- and π-GST, which are specifically present in the kidney tubules, are two isotypes of the glutathione-S-transferases. Beyond their biochemical differences, they are also located in different parts of the tubule system [8]. α-GST is found in cells of the proximal tubules, which are predominantly affected by ischemia time and nephrotoxic substances. π-GST is located in distal tubules, which are damaged during acute graft rejection [9]. Their release into the urine as a result of cell damage gives an accurate prediction of the impaired part of the tubules system and therefore the underlying cause of graft dysfunction [10,11]. Analyses of α- and π-GST have been reported to be promising for discriminating between the different causes of graft dysfunction [11–13].

The aim of our study was to determine the value of measuring α- and π-GST concentrations in urine as biomarkers for monitoring graft function and predicting postoperative events in the first week after transplantation in living and deceased donor kidney transplantation.

#### **2. Experimental Section**

#### *2.1. Samples and Data Collection*

This study was approved by the Ethics Committee of Berlin's Charité University Hospital (EA2/137/10). We prospectively analyzed blood and urine samples as well as demographic data from 160 patients: 30 brain-dead donors and their 30 corresponding recipients; and 50 living kidney donors as well as their 50 corresponding living donor kidney recipients. All surgeries were carried out at the Department of General, Visceral and Transplantation surgery of Charité University Hospital, Berlin. Machine perfusion was not performed for any kidney allograft in this study. Except for brain-dead donors, all patients were followed during the first week after surgery. Blood and urine samples were collected at the following time points: day 0, day 1, day 3, day 5, and day 7. Samples from brain-dead organ donors were obtained on day 0, the day of the organ donation surgery. Urine samples were collected from recipients after transplantation through an externalized uretero-vesico-cutaneous stent and therefore exclusively reflected the α- and π-GST content of the transplanted grafts.

#### *2.2. Immunosuppression Events and Subgroups*

All recipients received a triple immunosuppressant consisting of prednisolone, mycophenolate mofetil (MMF), and a calcineurin inhibitor. All deceased donor kidney recipients received tacrolimus, whereas living donor kidney recipients were treated with either tacrolimus or cyclosporine. Recipients were divided into subgroups according to the events in the first postoperative week: acute graft rejection (AGR, G1), calcineurin-induced nephrotoxicity (CNI, G2), both acute kidney rejection and calcineurin-induced nephrotoxicity (AGR + CNI, G3), delayed graft function (DGF, G4), and event-free (healthy, G5) subgroup (Figure 1). An acute graft rejection was confirmed by graft biopsy and classified according to the BANFF criteria. Calcineurin-induced nephrotoxicity was defined by serum levels of agents (tacrolimus >15 ng/mL, cyclosporin >250 ng/mL). It is worth noting that delayed graft function, characterized by the need for dialysis in the first week after transplantation, was not recorded in any of the living donor recipients. In case dialysis was required due to a known graft-damaging event such as AGR or CNI, those patients were enrolled in subgroups of primary cause and not defined as DGF. Recipients with no signs for any of the above events were considered healthy and were used as our control group.

**Figure 1.** Patients, groups, and subgroups.

#### *2.3. GST Analysis and Statistics*

Urinary α- and π-GST values were measured using a commercially available ELISA test kit provided by Argutus Medical Ltd. (Dublin, Ireland). To consider the physiological differences in urine concentrations, α- and π-GST values were standardized to urine creatinine. The resulting unit for GST was ng/mg uCrea. Reference ranges of urinary α- and π-GST were determined as recommended from the measured GST values in the healthy population of this study group. For this purpose, the αand π-GST values in the urine of healthy living donors were considered before donor nephrectomy. The reference interval for α-GST is 2.7–7.6 ng/mg uCrea and for π-GST 4.1–13 ng/mg uCrea in this study.

For statistical analysis, GraphPad Prism 6 (GraphPad Software, San Diego, CA, USA) was used. Quantitative data are given as the mean and standard deviation. To compare normally distributed variables, *t*-tests such as Mann-Whitney and Wilcoxon were performed. For the comparison of multiple variables, we used two-way ANOVA; here we applied Tukey and Holm-Sidak tests for post hoc analysis of the subgroups. The area under the curve was calculated in ROC analysis and a log-rank (Mantel-Cox) test was performed in survival analysis. A *p*-value less than 0.05 was defined as significant.

#### **3. Results**

#### *3.1. Demographic Data*

There were no differences regarding gender and BMI between the patient groups. The mean eGFR before graft recovery or transplantation (d0) did not differ between the groups. Recipients of deceased donor kidneys were significantly older than those of living donors (58 ± 13 and 48 ± 15 respectively, *p* = 0.006), whereas the donor's age was not different. Cold as well as warm ischemic time were significantly higher in the deceased donor group (both *p* < 0.0001). The demographic data of donors and recipients are given in Table 1. The increase in the eGFR of recipients after transplantation was, as expected, more noticeable in living donation scenarios (Figure 2).


**Table 1.** Demographic data of donors and recipients.

BMI, body mass index; eGFR, estimated glomerular filtration rate; n.s., not significant.

**Figure 2.** eGFR comparison between deceased and living donor recipients before and during the first week after transplantation.

#### *3.2.* α*- and* π*-GST in Recipients*

Neither α- nor π-GST correlated with age, BMI, and cold or warm ischemic time in any group. However, both GST isoenzymes correlated with renal function in living donors and in the healthy recipients subgroup. In deceased donor recipients and living donor recipients we observed acute rejection in four (13.4%) and eight patients (16%), calcineurin-induced nephrotoxicity in five (16.7%) and 12 (24%) patients, and both simultaneously in three (10%) and three (6%) patients, respectively. Delayed graft function occurred only with deceased donors (nine patients, 30%). Patients with an uneventful postoperative course in the deceased donor recipient group numbered nine (30%) and, in the living donor recipient group, 27 (54%). Both α- and π-GST were significantly elevated at 1st postoperative day (POD) in deceased donor recipients, with acute rejection when compared with the corresponding healthy subgroup (α-GST: Mean 473.5 ± 818 vs. 15.6 ± 21.2 ng/mg uCrea,

*p* = 0.0094; π-GST: mean 477.8 ± 804 vs. 8 ± 6.4 ng/mg uCrea, *p* = 0.0023). In living donor recipients, only π-GST showed an increase, reaching a peak at day 5 (mean 21.5 ± 28.2 ng/mg uCrea); however, there was no significant difference between this group and the healthy subgroup. In patients with CNI toxicity, α-GST performed better in both recipient groups; in deceased donor recipients, the mean was 316 ± 704.5 ng/mg uCrea at 1 POD and was easily discriminated from the uneventful subgroup (*p* = 0.06). Also, in the living donor group, a rise of α-GST to 68.9 ± 219 ng/mg uCrea was noted at day 5, when CNI toxicity occurred. When both acute rejection and CNI toxicity were recorded, neither αnor π-GST was able to distinguish those patients; however, it is worth mentioning that the number of subjects in this subgroup was very low in our study. In the case of delayed graft function, present only in deceased donor recipients in our study, α- as well as π-GST were elevated in the urine, with means at 1st POD of 81.5 ± 201.3 and 151.6 ± 270.6 ng/mg uCrea, respectively. Only π-GST levels proved significant when compared to the control subgroup (*p* = 0.036) (Figure 3).

**Figure 3.** Urinary α- and π-GST levels in subgroups of deceased and living donor recipients during the first week after transplantation; G1: Acute graft rejection (AGR), G2: Calcineurin-induced nephrotoxicity (CNI), G3: Simultaneous acute graft rejection and calcineurin-induced nephrotoxicity (AGR + CNI), G4: Delayed graft function (DGF), G5: Event-free (control).

Furthermore, we performed ROC curve and survival analysis on the most outstanding biomarkers at a specific point in the study. π-GST showed the most promising results in deceased donor recipients with acute graft rejection and delayed graft function at day 1 after transplantation. Patients who developed a biopsy confirmed acute graft rejection within the first week after transplantation had significantly higher levels of urinary π-GST at POD 1. With an estimated cutoff of 21.4 ng/mg uCrea, π-GST was able to distinguish the occurrence of AGR from a rejection-free course with 100% and 66.6% sensitivity and specificity, respectively (Figure 4). Similar reliability for π-GST was observed in the DGF subgroup; however, the estimated cutoff at POD 1 was slightly lower at 18.3 ng/mg uCrea (sensitivity, 100%; specificity, 62.6%). Higher urinary π-GST levels could be seen in recipients with

delayed graft function; this was observed at POD1, POD 3 as well as POD 7. π-GST was not able to differentiate between the causes of graft dysfunction in the early postoperative period (Figure 5).

**Figure 4.** Correlation between urinary π-GST levels in recipients of brain-dead donors' grafts on POD 1, and probability of graft survival without acute graft rejection (AGR) during the first week after transplantation.

**Figure 5.** Correlation between urinary π-GST levels in recipients of brain-dead donors' grafts on POD 1, 3, and 7; and probability of graft survival without delayed graft function (DGF) during the first week after transplantation.

#### *3.3.* α*- and* π*-GST in Donors*

α- and π-GST showed interesting results when measured in deceased donor urine before organ harvesting as they were elevated in those with poorer graft function after transplantation, and seemed to predict a foreseeable event such as acute rejection or delayed graft function, which was 3- to 8-fold more likely to occur than in those recipients with an uneventful course of treatment. α-GST stood out in the subgroup with both acute rejection and CNI toxicity (*p* = 0.02), while α- and π-GST were remarkably higher (but not significantly so) in the DGF subgroup compared to the control group (Figure 6). As for the predictive value of GST when measured in donor urine, α-GST stood out with significant results in the AGR subgroup when ROC curve analysis was performed, and a cutoff value of >33.97 ng/mg uCrea was calculated (AUC, 0.86; sensitivity, 77.7%; specificity, 100%). Based on the donor's urinary α-GST alone, all four renal grafts from deceased donors that showed acute rejection in recipients were distinguished from those who had an AGR-free course in the survival curve analysis (*p* = 0.0109) (Figure 7).

On the other hand, urinary GST in living donors showed no differences between subgroups and the corresponding control group, and therefore failed to predict future events in recipients.

**Figure 6.** Urinary α- and π-GST levels in deceased donors before transplantation; G1: Acute graft rejection (AGR), G2: Calcineurin-induced nephrotoxicity (CNI), G3: Simultaneous acute graft rejection and calcineurin-induced nephrotoxicity (AGR+CNI), G4: Delayed graft function (DGF), G5: Event-free (control).

**Figure 7.** Correlation between urinary α-GST levels in brain-dead donors before transplantation and probability of graft survival without acute graft rejection in corresponding recipients during the first week after transplantation.

#### *3.4. Six- and 12-Months Graft Survival*

We followed up recipients of the subgroups G1 (AGR) and G4 (DGF) at six and 12 months after transplantation, as correlations of α- and π-GST in those cohorts showed the most promising results. However, in subgroup G1 only one patient out of four lost the graft due to recruiting nephritis; in subgroup G4 all three grafts were lost due to death not associated with graft function (cancer or cardiac arrest).

#### **4. Discussion**

The results of our prospective study, evaluating urinary α- and π-GST in deceased as well as living kidney donors and their corresponding recipients as biomarkers for graft quality and function, suggest the potential value of these enzymes. Previous studies showed the ability of urinary α- and π-GST to predict acute renal damage in kidney graft recipients and demonstrated a release of these enzymes in malfunctioning grafts [10–15]. However, none of these groups has compared the course of αand π-GST in donors as well as in corresponding recipients. In addition, we investigated the differences in the markers in two settings, brain-dead/deceased and living organ donation. Our analyses reveal that the determination of urinary π-GST concentration in deceased donor recipients, especially on the first day after graft transplantation, could be valuable and indicative of kidney allograft function and survival without AGR or DGF. Secondly, we found higher concentrations of α- and π-GST in the urine of deceased kidney donors, whose grafts performed poorly in the corresponding recipients; α-GST was able to predict AGR before transplantation. A determination between the causes of impaired allograft function could not be reached by assessing urinary α- and π-GST alone, though it is unlikely that in the complex setting of transplantation a single biomarker will reliably distinguish between the pathogenesis of multifactorial elements; therefore, the proposed markers should be seen as an useful tool in addition to established methods.

Research studies investigating α- and π-GST in living donation are extremely limited: in our review of the literature we only found one publication on this issue, and this concerned liver rather than kidney transplantation [16]. Our findings showed lower concentrations of α- and π-GST in the urine of living donor kidney recipients than in that of deceased donor kidney recipients. This was observed in almost all subgroups, and especially on POD 1. Except for a notable, yet insignificant, rise of π-GST in living donor recipients with AGR when compared to the control group, our results find no further significances of urinary α- and π-GST in living donor recipients when harmful events occurred. Considering the superior organ quality and logistics in living donation transplantation, as demonstrated by over half of living donor grafts surviving the first week after transplantation event-free compared to 30% of deceased donor grafts, lower urinary concentrations of α- nor π-GST in living donor transplantation are to be expected. Daemen et al. found a correlation between α-GST and warm ischemic time in grafts from donation after cardiac death [17]. In our study neither α- nor π-GST had a proven correlation with ischemic time, although it should be noted that a different type of graft was investigated in the work mentioned.

The toxic effect of CNI agents on renal grafts and its association with the excretion of α-GST has been described in the past [10,18]. Our results failed to indicate such a correlation in deceased or living donor recipients. This might be due to our definition of the toxic range of CNI serum levels. Therapeutic and toxic serum levels of several drugs and especially immunosuppressants have been known to be inconsistent and even overlapping [6,19]. In the 1990s, serum levels of tacrolimus in the early period after transplantation were suggested to be below 20 ng/mL in order to avoid side effects, whereas later on levels above 15 ng/mg were proven to be associated with a higher risk of toxicity [20,21]. On the other hand, the risk of acute rejection is significantly higher when there are low concentrations of the agent [22]. A helpful step would be to find the toxic serum level of immunosuppressants that is agreed upon by transplant communities; currently, despite all efforts, this varies significantly between transplant centers. Another concerning factor is the design of the study, which did not include daily and therefore more precise surveillance on that matter. In order to investigate the correlation between toxic exposure to immunosuppressants and the excretion of GST into the urine, a closer observation with more frequent sample collection is required.

DGF was observed only in deceased donor recipients in our study. It occurred with an incidence of 30%, which is similar to the findings of a recent work by Willicombe et al. [23]. Risk factors and characteristics of donors and recipients associated with DGF such as cold ischemic time and donor age have been established in previous publications [4,5,24]. However, taking these factors into consideration, it is to be expected that DGF is less common in living donation. It has been demonstrated that α-GST excretion would be increased in the case of DGF due to its location in the renal tubular system and the association between DGF and proximal tubular necrosis [8,25]. On the other hand, π-GST has been shown to be of predictive value in terms of the need for dialysis in a publication by Seabra et al. including 245 patients with acute kidney injury [26]. Hall et al. investigated α- and π-GST in a perfusate solution during machine perfusion of kidney allografts from deceased donors, and suggested an independent association between π-GST and DGF [27]. Our findings demonstrated the consistent significance of urinary π-GST in differentiating between DGF and normally functioning grafts when measured in deceased donor recipients. This was observed at several time points of the study and had a strong power of sensitivity and specificity. Little is known about the behavior of the proteins under dialysis, so it is unclear whether α-GST is more dialyzable than π-GST or the other way round. The sample collection from patients undergoing dialysis in our study did not occur with respect to dialysis time, which should be seen as another possible disturbance factor.

Further limitations of this study are the small number of patients in certain subgroups and the overall high standard deviations. We distinguished well between the causes of impaired graft function and took into consideration simultaneous events. The time frame of our observation was limited in that it focused only on the first week after transplantation. We believe that multiple serial samples and an extended study design would be beneficial in future projects.

#### **5. Conclusions**

In summary, the elevation of urinary π-GST in deceased donor kidney recipients at day 1 after kidney transplantation could be a helpful monitoring parameter, in addition to urinary output and serum creatinine, to determine graft function in recipients. It might be indicative of acute rejection or a need for dialysis. The measuring of urinary α- and π-GST should also be considered in deceased donors as this seems to be of predictive value in terms of graft outcome and might help with assessing allograft quality. Our findings reveal an association between urinary α-GST in deceased donors and AGR in corresponding recipients. Thus, further investigation of α- and π-GST in a larger population and daily sample collection should be considered. Although urinary α- and π-GST in living kidney donation showed no relevant correlation with harmful events in our analyses, this is, to the best of our knowledge, the first study demonstrating differences in biomarkers between deceased and living kidney donation, so subsequent investigations will be needed in order to confirm or contradict our findings.

**Author Contributions:** Conceptualization, S.K. and V.S.; methodology, S.K., V.S. and B.G.; software, S.K., A.R.-S. and B.S.; validation, B.G., M.M. and B.S.; formal analysis, S.K. and A.R.-S.; investigation, S.K., B.G. and V.S.; resources, A.P., V.S. and N.M.O.; data curation, S.K., B.G., M.M. and N.M.O.; writing—original draft preparation, S.K.; writing—review and editing, J.G.B., M.H.M. and T.V.; visualization, S.K., M.H.M. and T.V.; supervision, V.S., A.P. and J.G.B.; project administration, S.K. and V.S. All authors have read and approved the final version of the manuscript.

**Acknowledgments:** We acknowledge support from the Open Access Publication Fund of the University of Münster. The authors also extend their gratitude to Argutus Medical Ltd. (Dublin, Ireland) for providing the kits.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


#### *J. Clin. Med.* **2019**, *8*, 1899


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Urinary Oxalate Excretion and Long-Term Outcomes in Kidney Transplant Recipients**

**Alwin Tubben 1,\*, Camilo G. Sotomayor 1, Adrian Post 1, Isidor Minovic 2, Timoer Frelink 3, Martin H. de Borst 1, M. Yusof Said 1, Rianne M. Douwes 1, Else van den Berg 1, Ramón Rodrigo 4, Stefan P. Berger 1, Gerjan J. Navis <sup>1</sup> and Stephan J. L. Bakker <sup>1</sup>**


Received: 21 October 2019; Accepted: 22 November 2019; Published: 2 December 2019

**Abstract:** Epidemiologic studies have linked urinary oxalate excretion to risk of chronic kidney disease (CKD) progression and end-stage renal disease. We aimed to investigate whether urinary oxalate, in stable kidney transplant recipients (KTR), is prospectively associated with risk of graft failure. In secondary analyses we evaluated the association with post-transplantation diabetes mellitus, all-cause mortality and specific causes of death. Oxalate excretion was measured in 24-h urine collection samples in a cohort of 683 KTR with a functioning allograft ≥1 year. Mean eGFR was 52 <sup>±</sup> 20 mL/min/1.73 m2. Median (interquartile range) urinary oxalate excretion was 505 (347–732) μmol/24-h in women and 519 (396–736) μmol/24-h in men (*p* = 0.08), with 302 patients (44% of the study population) above normal limits (hyperoxaluria). A consistent and independent inverse association was found with all-cause mortality (HR 0.77, 95% CI 0.63–0.94, *p* = 0.01). Cause-specific survival analyses showed that this association was mainly driven by an inverse association with mortality due to infection (HR 0.56, 95% CI 0.38–0.83, *p* = 0.004), which remained materially unchanged after performing sensitivity analyses. Twenty-four-hour urinary oxalate excretion did not associate with risk of graft failure, post-transplant diabetes mellitus, cardiovascular mortality, mortality due to malignancies or mortality due to miscellaneous causes. In conclusion, in KTR, 24-h urinary oxalate excretion is elevated in 44% of KTR and inversely associated with mortality due to infectious causes.

**Keywords:** oxalate; hyperoxaluria; kidney transplant recipients; graft failure; post-transplantation diabetes mellitus; all-cause mortality; cardiovascular mortality; infectious mortality

#### **1. Introduction**

Kidney transplantation is considered the gold standard treatment for end-stage renal disease (ESRD) [1,2]. Short-term survival of kidney transplant recipients (KTR) has improved markedly in the past decades [3,4]. Although a better understanding of modifiable risk factors has been achieved over the recent years [5,6], patients perceive the ever existing threat of premature graft failure (GF) as most compelling, and would like to know whether factors such as lifestyle and diet can contribute to prevention of it [7,8]. Another factor of interest influencing long-term KTR survival is

post-transplantation diabetes mellitus (PTDM), which has become increasingly common and may affect patient and graft survival [9]. Further, an increased risk of premature mortality, in particular, increased risk for premature death from cardiovascular and infectious causes remain significant problems in the post-transplantation setting. In KTR, conventional risk factors for cardiovascular mortality are abundantly present, such as hypertension, diabetes mellitus and dyslipidemia. On top of that, KTR had pre-existent renal diseases, which additionally increases the cardiovascular risk [10]. Mortality due to infection is significantly higher in KTR than in the general population due to multiple reasons, to which immunosuppressive therapy is a large contributing factor [11]. Furthermore, KTR are at a two to threefold higher risk of cancer-related mortality compared to the general population [12].

Although different mechanisms underlying these long-term complications of kidney transplantation have been found, substantial unknown mechanisms particular to the post-kidney transplantation setting remain to be identified in order to provide rationale for the markedly high risk of premature mortality in KTR [13]. A recent prospective cohort study in 3123 patients with chronic kidney disease (CKD) stages 2 to 4, found urinary oxalate as a potential risk factor for progression of CKD [14]. In the post-kidney transplantation setting, the study of oxalate remains overlooked. Whether urinary oxalate (reference value ≤455 μmol/24-h) [15] may be prospectively associated with adverse outcomes in KTR remains unknown.

The current study aims to assess the potential association of urinary oxalate excretion with adverse long-term outcomes in a large cohort of extensively phenotyped KTR with a functioning graft ≥1 year. For this purpose, the prospective associations of 24-h urinary oxalate excretion with GF, PTDM, and overall and cause-specific mortality were systematically investigated.

#### **2. Experimental Section**

#### *2.1. Study Design and Population*

This is a single-center prospective cohort study, initiated in 2008 on with follow-up of endpoints until 2015. KTR with a functioning allograft for at least one year or more who visited the outpatient clinic of the University Medical Center Groningen (Groningen, The Netherlands) between November 2008 and March 2011. Exclusion criteria were no known or apparent systemic illnesses, insufficient knowledge of the Dutch language and history of drug or alcohol addiction according to their patient files. KTR received anti-hypertensive and standard maintenance immunosuppressive therapy. Of the 817 invited KTR, 706 (87%) signed informed consent. Patients missing 24-h urinary oxalate excretion were excluded from the analyses, resulting in 683 KTR eligible for statistical analyses. The study was conducted in concordance with the guidelines formulated in the Declaration of Helsinki and Istanbul, and approved by the Institutional Review Board of the UMCG (METc 2008/186) [16]. The continuous surveillance system according to the American Society of Transplantation was followed for the correct collection of data [17]. When status of patients was unknown, the referring nephrologist or general practitioners were contacted in order to obtain the missing information. There was no loss due to follow-up.

#### *2.2. Study Endpoints*

The primary endpoint of this study is death-censored GF. Secondary endpoints are PTDM, all-cause mortality and cause-specific mortality. GF occurrence in this study is defined as ESRD requiring re-transplantation or return to dialysis. A subject was considered to have developed PTDM when the fasting plasma glucose exceeded 7 mmol/L, the HbA1c exceeded 6.5% or use of antidiabetics after transplantation as registered in the patient database [18,19]. Among specific causes of death, we studied cardiovascular mortality, death from infection, death from malignancies, and other causes of death (miscellaneous). Cardiovascular mortality is defined as mortality caused by cardiovascular pathophysiology, coded by ICD-10 codes I10-I52. This information was obtained from linking the patient number to the database of the Central Bureau of Statistics and then, by physicians, reported

mortality cause. Infectious mortality and mortality due to malignancies were defined as mortality caused by infectious causes or malignant causes. Miscellaneous causes of mortality have been defined as other causes of death besides the previously described outcomes.

#### *2.3. Baseline Measurements and Definitions*

At the outpatient clinic, baseline data was gathered according to a strict and detailed protocol described previously [20]. Anthropometrics were obtained without shoes and heavy garments. Systolic and diastolic blood pressures (SBP and DBP) were measured by means of an automatic device (Philips Suresign VS2+, Andover, MA, USA) according to a standard clinical protocol [16]. Mean arterial pressure (MAP) was automatically calculated by (SBP + DBP × 2)/3. History of cardiovascular disease was searched for in the patient files under ICD-10 code Z86.7.

Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m2) [21]. Estimated glomerular filtration rate (eGFR) was calculated using the CKD Epidemiology Collaboration (CKD-EPI) creatinine equation as shown in Formula (S1) [22].

#### *2.4. Assesments of Physical Activity and Dietary Intake*

Physical activity was quantified using the Short Questionnaire to Assess Health-enhancing physical activity (SQUASH). Activity was expressed in intensity multiplied by the amount of hours [23]. Dietary intake was assessed using a semi-quantitative Food Frequency Questionnaire (FFQ) [24,25]. To obtain the energy of a certain product, the Dutch Food Composition Table of 2016 was used [26]. Micro and macronutrients were adjusted for total energy intake (kCal), because of the potential of correlation and confounding [27].

#### *2.5. Laboratory Measurements*

For the collection of 24-h urine samples, the patients were asked to start the collection the day prior to their visit to the outpatient clinic. Collection was done in concordance with a strict protocol, i.e., discarding the first morning urine, collecting the subsequent in 24 h including the next morning's urine [16]. Subsequently, urine samples for oxalate analysis were acidified and stored at −80 ◦C. Urine oxalate analysis was performed using a validated ion-exchange chromatography assay with conductivity detection (Metrohm, Herisau, Switzerland). Inter-assay precision was monitored using three urine pool samples. Inter-assay precision was 8.2% at 0.17 mmol/L, 7.0% at 0.38 mmol/L and 9.0% at 0.52 mmol/L. Comparison of this method with a routine laboratory GC-MS method showed no systemic difference and no proportional difference. Reverse-phase high performance liquid chromatography (HPLC) was used to measure urinary thiosulfate [28].

#### *2.6. Statistical Analyses*

Data was analyzed using IBM SPSS version 23.0 (SPSS Inc., Chicago, IL, USA); Stata version 14.0 (StataCorp., College Station, TX, USA), GraphPad Prism version 7.02 (GraphPad Software, La Jolla, CA, USA), and Rstudio version 3.2.3 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided *p* < 0.05 was considered significant in all following analyses.

Normally distributed variables are expressed as mean ± standard deviation (SD), skewed data as medians (Interquartile range (IQR)), and categorical data as given number and percentage. Baseline characteristics were described for the overall population and by sex-stratified tertiles of 24-h urinary oxalate excretion. Data are presented in tertiles to allow for assessment of linearity of cross-sectional associations of 24-h urinary oxalate excretion with other variables. Sex-stratified tertiles were created by first separately distributing all female subjects according to tertiles and distributing all male subjects according to tertiles, and thereafter combining the tertiles of females and males. We generated sex-specific tertiles because of differences between women and men in oxalate excretion [29–32]. Analyses of difference in baseline characteristics across sex-stratified tertiles of 24-h urinary oxalate excretion were tested by ANOVA for normally distributed continuous variables, Kruskal-Wallis

for skewed continuous variables and χ2 test for categorical data. Sex-stratified tertiles of 24-h oxalate excretion were tested for associations with outcomes by Kaplan-Meier analysis, including the log-rank test.

Linear regression analyses were performed to investigate the association of baseline characteristics with 24-h urinary oxalate excretion. Normality was assessed by means of a *p*–*p* plot, and a natural log transformation was performed when appropriate. Homoscedasticity was controlled in a scatterplot.

Cox regression analyses were used to investigate the association of 24-h urinary oxalate with primary and secondary outcomes. Model 1 of the Cox proportional-hazards regression analysis was adjusted for demographics, i.e., sex and age. Model 2 was additionally adjusted for transplantation related variables, namely primary renal disease, BMI, donor age, time from transplantation to follow-up, eGFR and proteinuria. In the next models, baseline characteristics which were cross-sectionally associated with 24-h urinary oxalate excretion were subsequently included, and potential confounding of urinary thiosulfate was investigated due to its role in the anion transporters in the proximal renal tubuli (Model 3) [33]. In addition, we also looked for lactate dehydrogenase (LDH) because of its importance in the conversion of glyoxylate (Model 4) [34], for 24-h urinary pH because of its influence on the reaction of oxalate with calcium (Model 5) [35], for fibroblast growth factor 23 (FGF23) because of the relationship with gastrointestinal calcium absorption and oxalate bioavailability [36,37] (Model 6), and for fruits and vegetables as main dietary sources of oxalate [38–40] (Model 7). To allow for detection of a potential threshold effect, which was found in an earlier study on urinary oxalate excretion and CKD [14], Cox regression analyses were also performed according to sex-stratified tertiles with the first tertile as reference.

Spline regression were created to visualize the association of 24-h urinary oxalate excretion for outcomes, for which we consistently found significant associations. Nonlinearity was tested by using the likelihood ratio test, comparing models with linear or linear and cubic spline terms. Restricted cubic splines were knotted at the minimum, median and maximum. The splines were adjusted according to Model 6 of the primary prospective analyses.

#### Sensitivity Analyses

Several sensitivity analyses were performed to examine the robustness of the associations between 24-h urinary oxalate excretion and outcomes. For that purpose, we reanalyzed the data excluding subjects with potential inadequate 24-h urine collection (i.e., overcollection or undercollection), which was defined as the upper and lower 2.5% of the difference between the estimated and measured volume of a subject's 24-h urine sample. The following formula was used to calculate the estimated 24-h urine volume: <sup>1</sup> <sup>4</sup> ((urine creatinine) \* (24-h urine volume)/(serum creatinine)), where creatinine clearance was estimated using the Cockcroft-Gault Formula [41,42]. These analyses were analogous to Model 6 of the primary prospective analyses.

Furthermore, we performed competing risk analyses of outcomes of interest with all-cause mortality as competing event according to Fine and Gray [43]. For that purpose, we performed multivariable Cox regression analyses analogously to Model 6 of the primary prospective analyses.

#### **3. Results**

#### *3.1. Baseline Characteristics*

In total 683 KTR were included in the analyses (mean age 53 ± 13, 43% female, 99.6% Caucasian ethnicity). Median urinary oxalate excretion was 505 (IQR, 347–732) μmol/24-h in women and 519 (IQR, 396–736) μmol/24-h in men (*p* = 0.08). Forty-four percent of the patients were above the range of clinical hyperoxaluria of ≤455 μmol/24-h. All 227 study subjects in tertile 3 were above the clinical cutoff point for hyperoxaluria, and all 227 subjects in tertile 1 were below the clinical cutoff point. Mean eGFR was 52 <sup>±</sup> 20 mL/min/1.73 m2. Additional baseline characteristics and analyses are shown overall and by sex-stratified tertiles of 24-h urinary oxalate excretion in Table 1.


a


*J. Clin. Med.* **2019**, *8*, 2104


**Table** 


Abbreviations: ♀, female; ♂, male; KTR, kidney transplant recipients; *n*, number; β, standardized beta; BSA, body surface area; BMI, body mass index; SQUASH, Short Questionnaire to Assess Health-enhancing physical activity; SBP, systolic blood pressure; MAP, mean arterial pressure; LDL, low density lipoprotein; Av., average; hs-CRP, high sensitivity C-reactive protein; LDH, lactate dehydrogenase; eGFR, estimated glomerular filtration rate; UUN, urinary urea nitrogen; FGF-23, fibroblast growth factor 23. a Normally distributed variables are expressed as mean ± standard deviation (SD), skewed data as medians (25th–75th inter quartile range (IQR)), categorical data is given as number and percentage, *n,* (%). Analyses of difference in baseline characteristics across sex-stratified tertiles of 24-h urinary oxalate excretion were tested by ANOVA for normally distributed continuous variables; Kruskal-Wallis for skewed continuous variables; χ<sup>2</sup> test for categorical data. b To convert oxalate in μmol/24-h to mg/24-h, multiply by 0.088. c Adjusted for energy intake.

#### *3.2. Cross-Sectional Analysis*

We found that age (*p* = 0.04), current smoking status (*p* = 0.01), and cystatin C (*p* = 0.03) were inversely associated with 24-h urinary oxalate excretion, whereas plasma glucose (*p* = 0.01), ascorbic acid (*p* < 0.001), fruit consumption (*p* < 0.001), vitamin B6 (*p* < 0.001), urinary urea nitrogen excretion (*p* < 0.001), and phosphate excretion (*p* < 0.001) were positively associated with 24-h urinary oxalate excretion.

#### *3.3. Prospective Analyses*

GF and mortality were recorded during a follow-up of 5.3 years (IQR, 4.5–6.0). During follow-up, 83 (12%) patients developed GF, 55 (9%) patients developed PTDM and 149 (22%) patients died, of which 59 deaths (40%) were due to cardiovascular causes, 41 deaths (28%) due to infectious causes, 26 deaths (17%) due to malignancies and 23 deaths (15%) due to miscellaneous causes (Table 2).


**Table 2.** Association of baseline characteristics with 24-h urinary oxalate excretion. <sup>a</sup>

<sup>a</sup> Multivariate linear regression, adjusted for age, sex and eGFR.

#### 3.3.1. GF, PTDM, Cardiovascular Mortality, Mortality due to Malignancies, and Miscellaneous Mortality

A Kaplan-Meier curve for the association of tertiles of 24-h urinary oxalate excretion with GF is shown in Figure 1A (*p* = 0.20, *p* for trend 0.08). Results of multivariate Cox regression analyses did not show a consistent association of 24-h urinary oxalate excretion with GF (HR 0.71, 95% CI 0.53–0.98) (Table 3). Uni- and multivariate analyses of the associations of 24-h urinary oxalate excretion and potential confounders with GF are shown in Table S1.

A Kaplan-Meier curve for the association of tertiles of 24-h urinary oxalate excretion with PTDM is shown in Figure 1B (*p* = 0.24, *p* for trend 0.37). Results of multivariate Cox regression analyses showed no association of 24-h urinary oxalate excretion with PTDM (HR 0.95, 95% CI 0.71–1.27) (Table 3).

A Kaplan-Meier curve for the association of tertiles of 24-h urinary oxalate excretion with cardiovascular mortality is shown in Figure 1C (*p* = 0.08, *p* for trend 0.08). Results of multivariate Cox regression analyses showed cardiovascular mortality is not associated with 24-h urinary oxalate excretion (HR 0.78, 95% CI 0.56–1.10) (Table 4).

**Figure 1.** (**A**) Graft failure, (**B**) PTDM, (**C**) cardiovascular mortality, (**D**) death due to malignancy, (**E**) miscellaneous mortality (**F**) all-cause mortality, and (**G**) death due to infection according to sex-stratified tertiles of 24-hour urinary oxalate excretion over approximately 7 years of follow-up.


**Table 3.** Association of 24-h urine oxalate excretion with graft failure and PTDM.

Multivariate Cox regression were performed for the association of 24-h urinary oxalate excretion with graft failure and PTDM. Model 1: age and sex adjusted. Model 2: Model 1 + adjustment for BMI, primary renal disease, donor age, transplant vintage, eGFR, and proteinuria. Model 3: Model 2 + adjustment for thiosulfate in 24-h urine. Model 4: Model 3 + adjustment for LDH in blood. Model 5: Model 4 + adjustment for pH of 24-h urine. Model 6: Model 5 + adjustment for FGF23. Model 7: Model 6 + adjustment for fruit and vegetables intake.


**Table 4.** Association of 24-h urine oxalate excretion with all-cause and cardiovascular mortality.

Multivariate Cox regression were performed for the association of 24-h urinary oxalate excretion with all-cause and cardiovascular mortality. Model 1: age and sex adjusted. Model 2: Model 1 + adjustment for BMI, primary renal disease, donor age, transplant vintage, eGFR, and proteinuria. Model 3: Model 2 + adjustment for thiosulfate in 24-h urine. Model 4: Model 3 + adjustment for LDH in blood. Model 5: Model 4 + adjustment for pH of 24-h urine. Model 6: Model 5 + adjustment for FGF23. Model 7: Model 6 + adjustment for fruit and vegetables intake.

A Kaplan-Meier curve for the association of tertiles of 24-h urinary oxalate excretion with death due to malignancy is shown in Figure 1D (*p* = 0.51, *p* for trend 0.29). Results of multivariate Cox regression analyses showed mortality due to malignancies is not associated with 24-h urinary oxalate excretion (HR 1.14, 95% CI 0.73–1.77) (Table 5).

A Kaplan-Meier curve for the association of tertiles of 24-h urinary oxalate excretion miscellaneous mortality is shown in Figure 1E (*p* = 0.11, *p* for trend 0.10). Results of multivariate Cox regression analyses showed miscellaneous death causes are not associated with 24-h urinary oxalate excretion (HR 0.75, 95% CI 0.45–1.26) (Table 5).

**Table 5.** Association of 24-h urine oxalate excretion with death due to infection, malignancy and other causes.


Multivariate Cox regression were performed for the association of 24-h urinary oxalate excretion with death due to infection, malignancy and other causes. Model 1: age and sex adjusted. Model 2: Model 1 + adjustment for BMI, primary renal disease, donor age, transplant vintage, eGFR, and proteinuria. Model 3: Model 2 + adjustment for thiosulfate in 24-h urine. Model 4: Model 3 + adjustment for LDH in blood. Model 5: Model 4 + adjustment for pH of 24-h urine. Model 6: Model 5 + adjustment for FGF23. Model 7: Model 6 + adjustment for fruit and vegetables intake.

#### 3.3.2. All-Cause and Infectious Mortality

A Kaplan-Meier curve for the association of tertiles of 24-h urinary oxalate excretion with all-cause mortality is shown in Figure 1F (*p* = 0.06, *p* for trend 0.02). Results of multivariate Cox regression analyses showed, however, that all-cause mortality is independently associated with 24-h urinary oxalate excretion (HR 0.77, 95% CI 0.63–0.94) (Table 4). Uni- and multivariate analyses of the associations of 24-h urinary oxalate excretion and potential confounders with all-cause mortality are shown in Table S2. The association of 24-h urinary oxalate excretion with all-cause mortality demonstrated a nonlinear relationship, as shown by a restricted cubic spline (Figure 2A).

A Kaplan-Meier curve for the association of tertiles of 24-h urinary oxalate excretion with infectious mortality is shown in Figure 1G (*p* = 0.03, *p* for trend 0.008). Results of multivariate Cox regression analyses showed infectious mortality was independently associated with 24-h urinary oxalate excretion (HR 0.58, 95% CI 0.38–0.83) (Table 5). The association between 24-h urinary oxalate excretion and infectious mortality demonstrated a nonlinear relationship, as shown by a restricted cubic spline (Figure 2B).

**Figure 2.** Adjusted association of standardized log 24–hour urinary oxalate excretion with (**A**) all-cause mortality, and (**B**) infectious mortality, based on restricted cubic spline regression, fitted with Model 6. The black line in the graph represents the HR, 95% CI is shown and the gray area.

#### *3.4. Sensitivity Analyses*

When we restricted the analyses to subjects with no potential over or undercollection of 24-h urine samples based on differences in expected and observed 24-h urinary creatinine excretions (n = 650), generally similar results were found for GF (HR 0.74, 95% CI 0.55–0.99), PTDM (HR 0.93, 95% CI 0.69–1.26), cardiovascular mortality (HR 0.68, 95% CI 0.47–0.98), mortality due to malignancies (HR 1.10, 95% CI 0.70–1.72), mortality due to miscellaneous causes (HR 0.71, 95% CI 0.41–1.24), all-cause mortality (HR 0.74; 95% CI 0.59–0.92), and infectious mortality (HR 0.58, 95% CI 0.37–0.89).

When competing risk analyses were performed, generally similar results were found for PTDM (HR 1.15, 95% CI 0.91–1.48), cardiovascular mortality (HR 0.82, 95% CI 0.55–1.23), mortality due to malignancies (HR 1.16, 95% CI 0.69–1.95), mortality due to miscellaneous causes (HR 0.84, 95% CI 0.54–1.31), and infectious mortality (HR 0.61, 95% CI 0.45–0.85). The risk of GF was not consistently significant (HR 1.14, 95% CI 0.64–2.26).

#### **4. Discussion**

In KTR, median excretion of 24-h oxalate was higher than the clinical cut-off point for hyperoxaluria. No association of 24-h urinary oxalate excretion was found with GF, PTDM, cardiovascular mortality, or mortality due to malignancy or miscellaneous causes, but an independent, inverse association with all-cause mortality and infectious mortality was found. There was respectively a 23% and 44% decrease in hazard ratio per standard deviation increase of 24-h urinary oxalate excretion. The associations remained materially unchanged after adjusting for potential confounders. The association with all-cause and infectious mortality remained materially unchanged after performing sensitivity analyses.

A single-centered prospective study had previously already found an elevated plasma oxalate level in KTR [44]. However, no previous study has provided data on oxalate excretion. The elevated urinary oxalate excretion reflects one of the major findings of this study, being that 44% of the stable KTR are within the clinical range of hyperoxaluria.

To the best of our knowledge, there have not been any previous studies investigating the association of urinary oxalate with GF, PTDM and (cause-specific) mortality in stable KTR. However, a recent study of Waikar et al. with CKD patients stage 2 to 4 found 24-h urinary oxalate excretion to be positively associated with all-cause mortality [14]. With regards to the study of Waikar et al., their first four quintiles can be considered to be below the range of hyperoxaluria of 455μmol/24-h, whereas in our population, only the first tertile can be considered normal with regard to urinary oxalate excretion. This, in part, might explain the difference in association of urinary oxalate excretion with the outcome variables.

The difference between Waikar et al.'s and our study cannot be explained by a higher BMI or diabetes contributing to hyperoxaluria through higher effective renal plasma flow and glomerular hyperfiltration in the Waikar et al. population (respectively, BMI of 32.1 ± 7.7 and 26.6 ± 4.8 and diabetes in 48.9% and 24% of the population) [45]. Low density lipoprotein (LDL) profile was not published in the Waikar et al. report, therefore, difference in oxalate excretion through dyslipidemia cannot be determined [46]. In both studies, the urinary samples were stored at −80 ◦C. Storage at this temperature can lead to underestimation of oxalate levels through calcium oxalate precipitation [14]. Since the difference of storage time of the samples until measurement is not known, we cannot exclude this as a potential clarification of the found difference. Additionally, spontaneous oxalate generation over the course of the storage might have increased the sample oxalate levels in either studies. Another hypothesis for the interesting difference in 24-h urinary oxalate excretion between the study of Waikar et al. might be found in the possible absence of *Oxalobacter formigenes* in the gut microbiome. KTR have been exposed to antimicrobial prophylactic therapies to lower the risk of opportunistic infections. This greatly affects the diversity of the human microbiome and can cause dysbiosis [47]. Dysbiosis in KTR could contribute to a decrease of *O. formigenes* and therefore, increased gastrointestinal absorption of oxalate, leading to an increased oxalate serum concentration and consequently, elevated urinary excretion.

We found no association with GF, PTDM, mortality due to malignancies, nor mortality due to miscellaneous causes. The results of the proportional hazards models show that the inverse overall association with mortality is mainly driven by infectious mortality. We hypothesized that because 24-h urinary oxalate excretion was positively associated with ascorbic acid, which is inversely associated with overall mortality in RTR through reducing inflammation, an increase in oxalate might contribute to a lower infectious mortality [48]. However, the exact mechanism behind the association of 24-h urinary oxalate excretion with infectious mortality remains to be further investigated, since to our knowledge, there are no studies available showing a potential theoretical explanation.

The strength of this study lays in its prospective design, with a large cohort of stable KTR who were closely monitored according to standardized protocols and continuous surveillance system according to the American Society of Transplantation without loss due to follow-up during a median follow-up of 5.4 years for (specific cause) mortality. The KTR were extensively phenotyped at baseline measurement, providing a broad array of potential confounders to adjust for. The inclusion of the FFQ gives the possibility to assess the associations with dietary intake, rather than just the urinary excretion. Furthermore, urine was collected as 24-h collecting samples, according to a previously described strict protocol, which eliminates possible daily variances in fluid balance and excretion to give a more accurate excretion estimate. Additionally, potential over- or undercollection of the 24-h urine samples was accounted for by means of sensitivity analyses, which showed that the results remained materially unchanged after restricting the study population as described previously.

However, we also acknowledge limitations of the current study. First, we were unable to adjust our results for socioeconomic status at baseline. Next, although the FFQ and SQUASH are validated questionnaires, they are self-reported, which may lead to possible over or underreporting of dietary intake and physical activity. We also acknowledge that our population consists almost entirely of Caucasian ethnicity, therefore, our results call for caution to extrapolate our results to different populations with regard to ethnicity. Finally, data on nephrolithiasis was not documented; therefore, we were unable to assess the association of urinary oxalate with the outcome nephrolithiasis, which remains a rather overlooked topic in KTR. Nevertheless, our results show for the first time a high prevalence of hyperoxaluria in the post-kidney transplant setting, thus emphasizing the need for future studies in which such analyses are performed. Additionally, because the study of the microbiome was beyond the scope of the current study, the hypothesized mechanism of increased gastrointestinal absorption of oxalate to explain the observed levels of hyperoxaluria cannot be further confirmed.

In conclusion, in stable KTR, 24-h urinary oxalate excretion is quantitatively higher than in the general population. Forty-four percent of the current study population showed urinary oxalate levels above the range of clinical hyperoxaluria. This hyperoxaluria might suggest a role of dysbiosis by leading to diminished *O. formigenes* and therefore, higher oxalate absorption and excretion in the current study population. Twenty-four-hour urinary oxalate excretion was not associated with risk of graft failure, post-transplant diabetes mellitus, cardiovascular mortality, mortality due to malignancies, nor death from miscellaneous causes. However, a consistent and independent inverse association was found with infectious mortality. Our data encourages further studies to validate our findings on the associations of oxalate with long-term outcomes in KTR. Future studies are warranted to investigate specific causes of death and the effect of hyperoxaluria post-kidney transplantation.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/8/12/2104/s1, Formula (S1): CKD-EPI creatinine according to Levey et al., Supplementary Table S1: Uni- and multivariate analyses of the associations of 24-h urinary oxalate excretion and potential confounders with graft failure; Supplementary Table S2: Uni- and multivariate analyses of the associations of 24-h urinary oxalate excretion and potential confounders \* with all-cause mortality.

**Author Contributions:** Conceptualization, S.J.L.B.; data curation, E.v.d.B.; formal analysis, A.T. and C.G.S.; funding acquisition, C.G.S. and S.J.L.B.; investigation, A.T., C.G.S. and M.Y.S.; methodology, A.T., C.G.S., I.M., T.F., M.H.d.B., M.Y.S., G.J.N. and S.J.L.B.; project administration, S.P.B. and S.J.L.B.; software, M.Y.S.; supervision, C.G.S.; visualization, A.T. and S.J.L.B.; writing—original draft, A.T.; writing—review & editing, C.G.S., A.P., M.Y.S., R.M.D., R.R. and S.J.L.B.

**Funding:** Sotomayor is supported by a doctorate studies grant from CONICYT (F 72190118).

**Acknowledgments:** This study is based on data of the TransplantLines Food and Nutrition Biobank Cohort Study (ClinicalTrials.gov Identifier: NCT02811835).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Plasmapheresis Reduces Mycophenolic Acid Concentration: A Study of Full AUC0–12 in Kidney Transplant Recipients**

**Sudarat Piyasiridej 1, Natavudh Townamchai 1,2,3,\*, Suwasin Udomkarnjananun 1,2,3, Somratai Vadcharavivad 4, Krit Pongpirul 5,6, Salin Wattanatorn 1,2, Boonchoo Sirichindakul 2,7, Yingyos Avihingsanon 1,3, Kriang Tungsanga 1, Somchai Eiam-Ong <sup>1</sup> and Kearkiat Praditpornsilpa <sup>1</sup>**


Received: 18 October 2019; Accepted: 21 November 2019; Published: 1 December 2019

**Abstract:**Background: Mycophenolic acid (MPA), a crucialimmunosuppressive drug, and plasmapheresis, an effective immunoreduction method, are simultaneously used for the management of various immune-related diseases, including kidney transplantation. While plasmapheresis has been proven efficient in removing many substances from the blood, its effect on MPA plasma levels remains unestablished. Objectives: To evaluate the full pharmacokinetics of MPA by measuring the area under the time–concentration curve (AUC0–12), which is the best indicator for MPA treatment monitoring after each plasmapheresis session, and to compare the AUC0–12 measurements on the day with and on the day without plasmapheresis. Methods: A cross-sectional study was conducted in kidney transplantation recipients who were taking a twice-daily oral dose of mycophenolate mofetil (MMF, Cellcept®) and undergoing plasmapheresis at King Chulalongkorn Memorial Hospital, Bangkok, Thailand, during January 2018 and January 2019. The MPA levels were measured by an enzymatic method (Roche diagnostic®) 0, 1/2, 1, 2, 3, 4, 6, 8, and 12 h after MMF administration, for AUC0–12 calculation on the day with and on the day without plasmapheresis sessions. Plasmapheresis was started within 4 h after administering the oral morning dose of MMF. Our primary outcome was the difference of AUC0–12 between the day with and the day without plasmapheresis. Results: Forty complete AUC measurements included 20 measurements on the plasmapheresis day and other 20 measurements on the day without plasmapheresis in six kidney transplant patients. The mean age of the patients was 56.2 ± 20.7 years. All patients had received 1000 mg/day of MMF for at least 72 h before undergoing 3.5 ± 1.2 plasmapheresis sessions. The mean AUC on the day with plasmapheresis was lower than that on the day without plasmapheresis (28.22 ± 8.21 vs. 36.79 ± 10.29 mg × h/L, *p* = 0.001), and the percentage of AUC reduction was 19.49 ± 24.83%. This was mainly the result of a

decrease in AUC0–4 of MPA (23.96 ± 28.12% reduction). Conclusions: Plasmapheresis significantly reduces the level of full AUC0–12 of MPA. The present study is the first to measure the full AUC0–12 in MPA-treated patients undergoing plasmapheresis. Our study suggests that a supplementary dose of MPA is necessary for patients undergoing plasmapheresis.

**Keywords:** mycophenolic acid; immunosuppression; plasmapheresis; kidney transplantation

#### **1. Introduction**

Mycophenolic acid (MPA) is one of the main powerful immunosuppressive drugs widely used for many immunological diseases. There are two MPA compounds available, i.e., mycophenolate mofetil (MMF, Cellcept®) and enteric-coated mycophenolate sodium (EC-MPS, Myfortic®). Both MMF and EC-MPS are similar in terms of efficacy and safety. EC-MPS was developed to improve the side effects of upper gastrointestinal symptoms. The time to reach maximum plasma MPA concentration (tmax) of MMF is usually within 1–2 h after an oral dose, while EC-MPS reveals a median lag time from 0.25 to 1.25 h [1]. After absorption from the gastrointestinal tract, 97 to 99% of MPA, which is the active form, will bind to serum albumin. MPA is converted by uridine diphosphate-glucuronosyltransferase (UGT) into inactive mycophenolic acid glucuronide (MPAG), which is mainly excreted by the renal tubules. MPAG can also be excreted in the biliary tract by multidrug-resistant protein (MRP), which can lead to enterohepatic recycling. [1]

Plasmapheresis is one of the most effective methods utilized for rapid immunoglobulin removal in various immunological diseases. Many proteins and protein-bound substances, including medications, can also be removed during plasmapheresis sessions [2,3]. Substances which are likely to be removed during plasmapheresis have the following characters: (1) high blood concentration, (2) high protein bound, (3) low volume of distribution (Vd), and (4) undergoing high-dose/high-efficiency plasmapheresis [4].

Several immunologically mediated diseases can be treated by MPA together with plasmapheresis, i.e., systemic lupus erythematosus (SLE), lupus nephritis, myasthenia gravis, Guillain–Barré syndrome, psoriatic arthritis, relapsed/refractory thrombotic thrombocytopenic purpura (TTP), severe polymyositis/dermatomyositis, inflammatory bowel disease, pemphigus vulgaris, and kidney transplantation [5–7]. Unintentional removal of MPA may result in inadequate immunosuppression and unfavorable outcomes. Of interest, the effect of plasmapheresis on MPA concentration has been studied only in a case series of two patients, one kidney transplant recipient and one patient with myasthenia gravis [8]. MPA removal were measured by considering MPA levels at only two time points—before and after each plasmapheresis session. The MPA removal was calculated on the basis of MPA concentration in plasma effluent. The authors concluded that plasmapheresis of 3 L of plasma did not significantly alter post-plasmapheresis MPA concentration. Currently, there are no available data regarding the effect of plasmapheresis on the area under the concentration–time curve from 0 to 12 h (AUC0–12) of MPA, which is the best indicator of MPA exposure of patients.

The present study was conducted in kidney transplant recipients who were taking stable doses of MMF and had indication for plasmapheresis to examine the effects of plasmapheresis on MPA exposure.

#### **2. Methods**

An observational study of patients who were taking MMF (Roche, Basel, Switzerland) in combination with plasmapheresis treatment was conducted in King Chulalongkorn Memorial Hospital, Bangkok, Thailand, during January 2018 and January 2019. The inclusion criteria were kidney transplant recipients older than 18 years, who were under an immunosuppressive regimen of tacrolimus, MMF, low-dose prednisolone and had an indication for plasmapheresis. The dosage of MMF had to be

500 mg orally every 12 h for at least one week. Exclusion criteria were patients with serum albumin concentration lower than 2 g/dL and patients who were coadministered a proton pump inhibitor.

Plasmapheresis sessions were initiated within 4 h after the morning dose of MMF. The plasmapheresis machine was Plasauto EZ®, and the dialyzer was Plasmaflo® with a maximum pore size of 0.3 μm. The total treatment volume was 1.5 plasma volume per session. The blood flow rate was 150 mL/h. The replacement fluid was 5% albumin in the same volume as the treatment volume. The number of sessions required was determined on the basis of the clinical judgment of the attending nephrologists.

Plasmapheresis was performed on an alternate day basis for patients who were prescribed more than one plasmapheresis session.

Patients had to strictly take a stable dose of MMF, i.e., 500 mg orally every 12 h for at least one week, before entering the study. MMF dosage adjustment was not allowed during the study period. Patients were not allowed to have a meal for one hour before and two hours after taking the MMF dose. MPA level was measured by an enzymatic immunoassay method (Roche-diagnostic®). The AUC0–12 was calculated with the trapezoidal rule from the MPA levels at nine time points after the morning dose of MMF (C0, C0.5, C1, C2, C3, C4, C6, C8, and C12) (Figure 1). The full AUC0–12 was measured on the day just before the day patients underwent plasmapheresis and compared with the AUC0–12 of the following day, in which patients received the plasmapheresis treatment. Blood samples were taken via a heparin lock in the arm by using the double-syringe technique.

**Figure 1.** Timing of mofetil (MMF) dosage, plasmapheresis sessions, and meal on the day before and on the day with a plasmapheresis session. MPA: mycophenolic acid.

A complete clinical evaluation including vital signs and body weight was performed. The baseline characteristics including age, cause of end-stage renal disease, type of kidney transplantation, time after kidney transplantation, renal function, indications for plasmapheresis, session of plasmapheresis, and plasma volume per session were recorded.

Absolute and relative frequencies were used for qualitative data. Mean and standard deviation were utilized for numerical data. The chi-squared test was used for comparisons between categorical data. Paired-samples *t*-test was used to compare the AUC0–12 of the day with plasmapheresis and the AUC0–12 of the day without plasmapheresis. Data were analyzed using the SPSS statistic version 22 (IBM; New York, NY, USA).

This study was approved by The Research Ethics Review Committee for Research Involving Human Research Participants, Health Sciences Group, Chulalongkorn University (IRB No.CF 333/61). The study was registered with the Thai Clinical Trials Registry (TCTR20190211001).

#### **3. Results**

Six kidney transplant recipients were enrolled, with a total of 20 plasmapheresis sessions. There were 40 AUC0–12 measurements (each AUC consisted of measurements of MPA levels at 9 time points), 20 of which were recorded on the day just before the day patients underwent plasmapheresis, and the other 20 were recorded on the following day, when patients underwent a plasmapheresis session. The mean (±SD) age of the patients was 56.2 ± 20.7 years, and five patients (83.3%) were men (Table 1). At baseline, the mean (±SD) estimated glomerular filtration rate (eGFR) was 49.7 ± 10.9 mL/min/1.73 m2, serum albumin concentration was 3.8 <sup>±</sup> 0.4 g/dL, and hemoglobin concentration was 10.3 ± 1.4 g/dL. Indication for plasmapheresis was antibody-mediated rejection (ABMR) for all

six patients, who were diagnosed by pathological presentation and donor-specific antibody (DSA) detection. The number of plasmapheresis sessions per patient was 3.5 ± 1.2 (range of 1–4 sessions).


**Table 1.** Baseline characteristics of the patients.

ESRD: end-stage renal disease, ABMR: antibody-mediated rejection; eGFR: estimated glomerular filtration rate; CKD-EPI: chronic kidney disease epidemiology collaboration; SGOT: serum glutamic-oxaloacetic transaminase; SGPT: serum glutamate-pyruvate transaminase.

The mean of MPA AUC0–12 of the day with plasmapheresis was significantly lower than that of the day without plasmapheresis (28.22 ± 8.21 vs. 36.79 ± 10.29 mg × h/L, *p* = 0.001) (Figure 2). The percentage reduction of AUC0–12 was 19.49 ± 24.83% (Table 2). The early part of the AUC was affected by plasmapheresis sessions. The AUC0–4 of the day with plasmapheresis was significantly lower than that of the day without plasmapheresis (15.79 ± 6.46 vs. 21.78 ± 5.66 mg × h/L, *p* < 0.001), while the AUC4–12 was not significantly different between the day with and that without plasmapheresis (12.43 ± 5.02 vs. 15.00 ± 7.56 mg × h/L, *p* = 0.125).

**Figure 2.** MPA levels on the day with plasmapheresis (20 sessions) compared with those on the day without plasmapheresis (20 sessions). PP: plasmapheresis.



(AUC; area under the time–concentration curve).

The reduction of MPA AUC0–12 was detected as early as the first session of plasmapheresis. The MPA AUC0–12 of the day before and of the day of the first session of plasmapheresis were 41.66 ± 10.66 and 32.26 ± 9.42mg × h/L, respectively (*p* = 0.001) (Table 2 and Figure 3). The percentage reduction of MPA AUC0–12 of the first day of plasmapheresis session was 22.86 ± 6.99%. The AUC0–12 of the day before the second to that of the day of the forth plasmapheresis sessions could be rebounded from the AUC0–12 of the day with plasmapheresis. However, the rebounded AUC0–12 gradually decreased with the number of sessions of plasmapheresis that the patients received (Figure 4). Given that the target therapeutic AUC0–12 of MPA is 30 to 60 mg × h/L for kidney transplantation recipients [9], 17 out of 20 (85%) AUC0–12 measured on the day without plasmapheresis achieved the target therapeutic range, compared with only 9 out of 20 (45%) AUC0–12 measured on the day with plasmapheresis (*p* = 0.008) (Figure 5).

**Figure 3.** MPA levels on the day before the first plasmapheresis session (*N* = 6) compared with MPA levels on the day with the first plasmapheresis session (*N* = 6).

**Figure 4.** Comparison of the mean MPA AUC0–12 between the day with and that without plasmapheresis from the first plasmapheresis session to the fourth session.

**Figure 5.** The MPA AUC0–12 achieved the target level between the day just before a plasmapheresis session (20 measurements) and the following day, when plasmapheresis was administered (20 measurements).

#### **4. Discussion**

The present study is the first to demonstrate the effect of plasmapheresis on MPA exposure by using the full MPA AUC0–12. The AUC0–12 of MPA was significantly affected by plasmapheresis. This effect was found starting from the first session of plasmapheresis (Figures 2 and 3). One-fifth of the total AUC0–12 was lowered by plasmapheresis. The component of AUC most affected by plasmapheresis was the early part (AUC0–4). Undergoing plasmapheresis treatment immediately after an oral dose of MMF can lower the MPA peak level, leading to exposure to a subtherapeutic level of MPA. Consecutive sessions of plasmapheresis could increase the risk of underimmunosuppression by lowering the rebound of MPA AUC0–12 (Figure 4).

MMF is one of the major immunosuppressive agents widely used to treat many immunological diseases. Since overimmunosuppression can lead to many side effects and underimmunosuppression can cause unfavorable treatment outcomes, MPA level monitoring has been recommended to maintain MPA concentration at the therapeutic level [9,10]. Plasmapheresis is one of the most effective methods for rapid immunoglobulin G (IgG) reduction [5]. Many high-molecular-weight substances can also be removed during a plasmapheresis session, especially proteins and albumin, which makes albumin replacement necessary. Since 97 to 99% of MPA is protein-bound, MPA should be theoretically removed from patients during plasmapheresis treatment.

The effect of plasmapheresis on MPA plasma level was reported in only two patients who were administered MMF in combination with plasmapheresis [8]. Plasmapheresis sessions were started 4 h after MMF administration, and MPA removal was assessed at only two time points (pre- and post-plasmapheresis) together with MPA concentration in plasma waste. The authors concluded that a plasmapheresis session starting later than 4 h after the administration of an oral MMF dose did not significantly alter MPA concentration. Since serum proteins can be trapped in the dialyzer and bloodline, monitoring of MPA removal by only measuring MPA in plasma waste may not reflect total MPA removal. Our study monitored MPA exposure by full AUC0–12 measurement on the day with a plasmapheresis session as the study arm and on the day without plasmapheresis as the control arm. Alteration in AUC0–12 between the day with and the day without plasmapheresis is the best indicator of the effect of plasmapheresis on MPA plasma levels. The early phase of the full MPA AUC (peak level, AUC0–4) is the one mostly affecting MPA exposure and represents more than 50% of AUC0–12. The plasmapheresis sessions designed in the present study started within one hour after oral administration of an MMF dose which is the most crucial period for determining the effects of plasmapheresis on MPA.

MPA together with plasmapheresis is mainly utilized for the treatment of many immunologic conditions and diseases which require potent immunosuppression, such as kidney transplant rejection, severe lupus nephritis, or relapsed/refractory thrombotic thrombocytopenic purpura. The patients enrolled in the present study were kidney transplant recipients who were taking MMF and experienced antibody-mediated rejection, which is indicated for plasmapheresis treatment. The present study reveals that MPA administration without dosage adjustment during consecutive sessions of plasmapheresis can lead to unexpected underimmunosuppression and may increase the failure rate of treatment. The present study demonstrated that MPA AUC0–12 is reduced by 20% when a plasmapheresis session is started within 4 h after oral administration of MMF (Table 2, Figure 2). The higher the number of consecutive sessions of plasmapheresis performed, the higher the chance of MPA underexposure (Figure 4). We also further examined the role of MMF dose increments in two patients who underwent plasmapheresis and found that increasing the MMF dose from 1000 mg/day to 1250 mg/day can prevent subtherapeutic AUC0–12 during plasmapheresis sessions (unpublished data). An MMF dosage increment of 20% may be required to maintain a therapeutic level of MPA on the day patients undergo plasmapheresis. A further comprehensive study of therapeutic drug monitoring in patients with increased dose of MPA before undergoing plasmapheresis is crucially required. Otherwise, a 4 h delay of the plasmapheresis session after administration of an MMF dose may reduce the effect of plasmapheresis on MPA exposure (Figure 6).

**Figure 6.** Recommendations for MMF dose or plasmapheresis adjustment in patient receiving concomitant MMF and plasmapheresis treatment.

The MMF dose used in the present study was relatively low. This is because the target population of patients enrolled in this study were kidney transplant recipients who were in the maintenance phase of immunosuppression. Moreover, a study on Asian patients showed that most of the patients achieved the target MPA level with an MMF dose of 1000 mg/day [11]. Besides conventional plasmapheresis, a study of the effects of others apheresis techniques such as double-filtration plasmapheresis and immunoadsorption, which have different kinetics of protein removal, should be carried out.

#### **5. Conclusions**

Plasmapheresis significantly reduces MPA plasma levels, particularly in the early phase after oral administration of an MPA dose. This effect should be addressed when combining MPA administration together with plasmapheresis in a treatment protocol.

**Author Contributions:** Conceptualization, S.P., N.T. and S.U.; methodology, S.P., N.T. and S.V.; formal analysis, N.T. and K.P. (Krit Pongpirul); investigation, S.P. and S.W.; writing-original draft preparation, S.P. and N.T.; writing-review and editing, N.T., B.S., Y.A., K.T., S.E.-O. and K.P. (Kearkiat Praditpornsilpa); visualization, N.T., supervision, N.T.; project administration, N.T.; funding acquisition, S.P. and N.T.

**Funding:** This research was funded by Ratchadaphiseksomphot Fund of Chulalongkorn University, grant number RA61/106.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Plasma Vitamin C and Cancer Mortality in Kidney Transplant Recipients**

**Tomás A. Gacitúa 1, Camilo G. Sotomayor 1,\*, Dion Groothof 1, Michele F. Eisenga 1, Robert A. Pol 2, Martin H. de Borst 1, Rijk O.B. Gans 1, Stefan P. Berger 1, Ramón Rodrigo 3, Gerjan J. Navis <sup>1</sup> and Stephan J.L. Bakker <sup>1</sup>**


Received: 15 October 2019; Accepted: 21 November 2019; Published: 23 November 2019

**Abstract:** There is a changing trend in mortality causes in kidney transplant recipients (KTR), with a decline in deaths due to cardiovascular causes along with a relative increase in cancer mortality rates. Vitamin C, a well-known antioxidant with anti-inflammatory and immune system enhancement properties, could offer protection against cancer. We aimed to investigate the association of plasma vitamin C with long-term cancer mortality in a cohort of stable outpatient KTR without history of malignancies other than cured skin cancer. Primary and secondary endpoints were cancer and cardiovascular mortality, respectively. We included 598 KTR (mean age 51 ± 12 years old, 55% male). Mean (SD) plasma vitamin C was 44 ± 20 μmol/L. At a median follow-up of 7.0 (IQR, 6.2–7.5) years, 131 patients died, of which 24% deaths were due to cancer. In Cox proportional hazards regression analyses, vitamin C was inversely associated with cancer mortality (HR 0.50; 95%CI 0.34–0.74; *p* < 0.001), independent of potential confounders, including age, smoking status and immunosuppressive therapy. In secondary analyses, vitamin C was not associated with cardiovascular mortality (HR 1.16; 95%CI 0.83–1.62; *p* = 0.40). In conclusion, plasma vitamin C is inversely associated with cancer mortality risk in KTR. These findings underscore that relatively low circulating plasma vitamin C may be a meaningful as yet overlooked modifiable risk factor of cancer mortality in KTR.

**Keywords:** Kidney transplant; vitamin C; cancer mortality; oxidative stress.

#### **1. Introduction**

Although kidney transplantation improves the prognosis of patients with end-stage renal disease (ESRD), kidney transplant recipients (KTR) remain at higher mortality risk compared to healthy individuals [1]. Since the beginning of kidney transplantation, the main cause of death has been cardiovascular [2–4]. In recent years, however, there has been a changing trend in mortality causes in KTR, with a decline in death due to cardiovascular causes along with a relative increase in cancer mortality [2,5–7]. Among non-cardiovascular deaths, malignancies lead the individual causes of death [8,9]. Noteworthy is that overall risk of death associated with cancer in KTR is ten-fold higher than in the general population [9]. Given this relative increase in cancer mortality in KTR, further studies to explore potential risk factors and underlying mechanisms are needed.

Post-transplantation immunosuppression as well as chronic uremic state have been recently proposed as risk factors, with oxidative stress as a potential underlying mechanism [2,10,11]. Vitamin C is a well-known radical scavenger and reducing agent [12], and due to its antioxidant, anti-inflammatory and immune system enhancement properties, it could offer protection against cancer incidence in KTR [13]. There is evidence supporting that low plasma vitamin C may lead to an increased risk of dying from cancer in the general male population [13], and is also inversely associated with gastric cancer risk in the general population [14].

Increased oxidative stress occurs when there is an imbalance between antioxidant and pro-oxidant species, leading to oxidative damage. Malondialdehyde (MDA), a decomposition product of peroxidized polyunsaturated fatty acids, is a widely used and sensitive biomarker of oxidative damage [15]. Gamma-glutamyl transpeptidase (GGT) is also currently used as an indicator of whole body oxidative stress [16,17]. Uric acid in plasma acts as antioxidant in presence of vitamin C [18]. Higher levels of free thiol groups have been proposed to be protective against oxidative damage, similarly to vitamin C [19]. Under the hypothesis that anti-carcinogenic properties of vitamin C are mainly driven by its antioxidant properties, the potential protective effect of vitamin C against cancer mortality would be expected to vary upon changes in oxidative stress biomarkers.

This evidence suggests that vitamin C could be a simple and widely available modifiable risk factor for cancer mortality in KTR. Nevertheless, studies focusing on the prospective association of vitamin C and long-term cancer mortality in this clinical setting are lacking. In this study, in primary analyses we aimed to investigate the association of circulating plasma vitamin C concentrations with long-term cancer mortality in a large cohort of KTR. As oxidative stress is considered a potential underlying mechanism, we aimed to assess whether the potential association of plasma vitamin C with cancer mortality would vary upon changes in oxidative stress biomarkers, i.e., uric acid, free thiol groups, MDA and GGT. In secondary analyses, we aimed to investigate the association of circulating plasma vitamin C concentrations with cardiovascular mortality.

#### **2. Materials and Methods**

#### *2.1. Study Design and Patients*

We performed a post hoc analysis in the TransplantLines Insulin Resistance and Inflammation Biobank and Cohort Study, number NCT03272854. Outpatient KTR (≥18 years old) with a functioning graft for at least 1 year were invited to participate between August 2001 and July 2003. Patients with overt congestive heart failure and patients diagnosed with cancer other than cured skin cancer (squamous cell or basal cell carcinoma successfully treated by a dermatologist) were not considered eligible for the study. The outpatient follow-up constitutes a continuous surveillance system in which patients visit the outpatient clinic with declining frequency, in accordance with the American Transplantation Society guidelines [20]. A total of 847 KTR were invited to be enrolled, of which 606 (72%) patients provided written informed consent to participate. Data were extensively collected at baseline. Patients with missing plasma vitamin C concentration (*n* = 8) were excluded for the statistical analysis, resulting in 598 KTR, of whom data are presented in the current study (Figure S1). The present study was approved by the Institutional Review Board (METc 2001/039), and was conducted in accordance with declarations of Helsinki and Istanbul.

#### *2.2. Kidney Transplant Recipients Characteristics*

Relevant characteristics including recipient age, gender, and transplant date were extracted from the Groningen Renal Transplant Database. This database contains detailed information on all kidney transplantations that have been performed at the University Medical Center Groningen since 1968. Details of the standard immunosuppressive treatment were described previously [21]. Smoking status was obtained using a self-report questionnaire at inclusion. Details about collection of dietary history have been described before [22]. In brief, a semi-quantitative food-frequency questionnaire was used

to assess fruit and vegetable intake. Fruit intake was assessed by asking participants 'How many servings of fruit do you eat per day on average?' Vegetable intake was assessed by asking participants 'How many tablespoons of vegetable do you eat per day on average?' Respondents were asked to choose among five possible frequency categories: 0, 1, 2, 3, ≥4 per day. Collection of data on use of vitamin C or multivitamin supplements containing vitamin C was systematically performed, by means of self-report, at baseline.

#### *2.3. Laboratory Measurements*

All measurements were performed during a morning visit to the outpatient clinic. Diabetes mellitus was defined according to the guidelines of the American Diabetes Association [23]. Proteinuria was defined as urinary protein excretion ≥0.5 g/24 h. Kidney function was assessed by estimated Glomerular Filtration Rate (eGFR) applying the Chronic Kidney Disease Epidemiology Collaboration equation [24].

Blood was drawn after a fasting period of 8–12 h, which included no medication intake. According to a strict protocol, patients were instructed to collect a 24-hour urine sample the day before their visit to the outpatient clinic. Total cholesterol, low-density lipoprotein cholesterol (LDL), plasma triglycerides, plasma glucose levels, plasma insulin concentration, and glycated hemoglobin (HbA1C) were determined as described previously [25]. Plasma high sensitivity C-reactive protein (hs-CRP) was measured by enzyme-linked immunosorbent assay, as described previously [26]. MDA was measured fluorescently after binding to thiobarbituric acid as described before [27]. Ellman's reagent was used for the determination of free thiol groups in cell culture and a cell-free solution of L-cysteine as described previously [28]. Plasma creatinine concentration was determined using a modified version of the Jaffé method (MEGA AU510; Merck Diagnostica). Total urinary protein concentration was analyzed using the Biuret reaction (MEGA AU510; Merck Diagnostica).

#### *2.4. Plasma Vitamin C Measurement*

After phlebotomy, blood was directly transferred to the laboratory on ice, deproteinized and stored in the dark at −20◦C until analysis. For quantitative measurement ascorbic acid is enzymatically transformed to dehydroascorbic acid, which in turn is derivatized to 3-(1,2-dihydroxyethyl) furo-[3,4-b] quinoxaline-1-one. Then, reversed phase liquid chromatography with fluorescence detection is applied (excitation 355 nm, emission 425 nm).

#### *2.5. Cause-Specific Mortality and Graft Failure*

The primary endpoint for analyses was mortality from cancer, defined according to a previously specified list of International Classification of Diseases, Ninth Revision (ICD-9) codes 140–239 [29]. Secondary endpoint was mortality from cardiovascular causes, defined as death due to cerebrovascular disease, ischemic heart disease, heart failure, or sudden cardiac death according to ICD-9 codes 410–447. Information on the cause of death was derived from the patients' medical records and was assessed by an adjudication committee. Information about death-related type of cancer was ascertained by contacting the general practitioners who were in charge of deceased cancer patients. Graft failure was defined as return to dialysis or need for a re-transplantation. The continuous surveillance system of the outpatient program ensures up-to-date information on patient status and cause of death. There was no loss to follow-up.

#### *2.6. Statistical Analyses*

Data analysis was performed using SPSS version 23.0 software (SPSS Inc., Chicago, IL, USA), STATA 14.1 (STATA Corp., College Station, TX, USA), and R version 3.2.3 (R Foundation for Statistical Computing, Vienna, Austria). In all analyses, a two-sided *p* < 0.05 was considered significant. Continuous variables were summarized using mean (standard deviation; SD) for normally distributed data, whereas skewed distributed variables are given as median (interquartile range; IQR). Categorical variables were summarized as numbers (percentage). Multiple imputation was performed to account

for missingness of data among variables other than data on plasma vitamin C [30]. The percentages of missing data were 0.2, 0.2, 0.2, 0.2, 0.3, 0.3, 0.3, 0.3, 0.5, 0.7, and 0.7% for waist circumference, HbA1C, albumin, alkaline phosphatase, proteinuria, leukocyte concentration, MDA, cumulative dose of prednisolone, uric acid, GGT, and prior history of cardiovascular disease, respectively. The percentages of missing data were maximally 11, 21, and 33% for free thiol groups, free fatty acids, and fruit and vegetable intake, respectively.

Age- and sex-adjusted linear regression analyses were performed to evaluate the association of plasma vitamin C concentrations with baseline characteristics. Residuals were checked for normality and variables were natural log-transformed when appropriate. In order to study in an integrated manner which patient- and transplant-related variables of interest were independently associated with and were determinants of plasma vitamin C concentrations, we performed forward selection of baseline characteristics by including all the variables that were associated with plasma vitamin C with a *p* < 0.1 in the preceding age- and sex-adjusted linear regression analyses. Selected variables were then used to perform stepwise backwards multivariable linear regression analyses (*Pout* > 0.05). Standardized beta coefficients represent the difference (in standard deviations) in plasma vitamin C per 1 standard deviation increment in continuous baseline characteristics, or for categorical characteristics the difference (in standard deviations) in plasma vitamin C compared to the implied reference group.

To analyze whether plasma vitamin C was prospectively and independently associated with cancer mortality, we performed multivariable-adjusted Cox proportional hazards regression analyses. For these analyses plasma vitamin C concentrations were used as log-transformed values with a log2 base, in order to obtain the best fitting model. We tested proportionality assumptions of Cox proportional hazards regression analyses, and they were satisfied, indicating that the association of baseline vitamin C with outcome is constant over follow-up time of the current study. The selection of covariates was made a priori, considering their potential confounding effect based on previously described risk factors for all-cause mortality in KTR and generally accepted risk factors for cancer mortality in the general population and in KTR [9,10,13,31]. We adjusted for age, sex, and smoking status (Model 1); eGFR, dialysis vintage, time since transplantation and proteinuria (Model 2); and, fruit and vegetable intake (Model 3). To avoid overfitting and inclusion of too many variables for the number of events, further models were performed with additive adjustments to Model 3 [32]. We performed additional adjustments for diabetes mellitus, hs-CRP and prior history of cardiovascular disease (Model 4); immunosuppressive therapy (use of calcineurin inhibitors (CNI), use of antimetabolites, use of mammalian target of rapamycin (m-TOR) inhibitors, and cumulative dose of prednisolone, calculated as the sum of maintenance dose of prednisolone since kidney transplantation until inclusion in the study and the dose of prednisolone or methylprednisolone required for treatment of acute rejection (a conversion factor of 1.25 was used to convert methylprednisolone to prednisolone dose). For acute rejection, different amounts of prednisolone or methylprednisolone were administered, which was taken into account in the calculations. Rejection episodes after inclusion were not included [33]; Model 5); and transplantation era (Model 6). Transplantation eras, with corresponding immunosuppressing medications, have been previously well described [34]. In secondary analyses, the aforementioned Cox proportional hazards regression analyses were performed for cardiovascular mortality. The analyses for both cancer death and cardiovascular death were performed by fitting multivariable-adjusted proportional cause-specific hazard models. In each of these models, the competing events were treated as censored observations, causing the regression parameters to directly quantify the hazard ratio among those individuals who are actually at risk of developing the event of interest, i.e., cancer mortality or cardiovascular mortality [35]. Hazard ratios (HR) are reported with 95% confidence interval (CI). The HR of each model is given per doubling of vitamin C concentration.

To adhere to existing recommendations for good reporting on survival analyses [36,37], we tested for potential interaction of all potential confounders and the oxidative stress biomarkers with vitamin C, namely, uric acid, free thiol groups (corrected by total serum protein) [19], MDA, and GGT by fitting models containing both main effects and their cross product terms. For these analyses, *P*interaction < 0.05 was considered to indicate significant interaction. We also performed subgroup analyses according to the aforementioned oxidative stress biomarkers, with adjustment for age, sex, smoking status, eGFR, dialysis vintage, time since transplantation, proteinuria, and fruit and vegetable intake. Cut-off points of originally continuous variables used in the stratified analyses were determined so they would allow for an as much as possible similar number of events in each subgroup, and thus allow for similar statistical power for the assessment of the primary association under study (plasma vitamin C and cancer mortality) in each subgroup after stratification of the overall population. Whenever and as much as possible, these criteria were matched with clinical cut-off points.

In sensitivity analyses, we performed graft failure-censored Cox proportional hazards regression analyses of the association of plasma vitamin C with cancer mortality and cardiovascular mortality. In addition, we performed Cox proportional hazards regression analyses of the association of plasma vitamin C with cancer mortality with adjustment for HbA1c instead of diabetes mellitus.

#### **3. Results**

#### *3.1. Baseline Characteristics*

A total of 598 patients (51 ± 12 years old, 55% male) were included at a median of 5.9 (IQR, 2.6–11.4) years after kidney transplantation. None of the patients used vitamin C supplements or multivitamin supplements containing vitamin C. Mean plasma vitamin C concentration was 44 ± 20 μmol/L, mean eGFR was 47 <sup>±</sup> 16 mL/min/1.73 m2. Patient-related variables of interest, including transplant-related characteristics and immunosuppressive therapy are summarized in Table 1. The results of the ageand sex-adjusted linear regression analyses are shown in Table 2. In stepwise backward multivariable linear regression analysis, fruit intake (std. β = 0.22; *p* < 0.01), dialysis vintage (std. β = −0.09; *p* < 0.05), proteinuria ≥0.5 g/24 h (std. β = −0.11; *p* < 0.05), HbA1C (std. β = −0.14; *p* < 0.01), diastolic blood pressure (std. β = −0.16; *p* < 0.01), alkaline phosphatase (std. β = −0.15; *p* < 0.01), hs-CRP (std. β = −0.17; *p* < 0.01) and male sex (std. β = −0.18; *p* < 0.01) were identified as independent determinants of plasma vitamin C (Table 2). The overall *R*<sup>2</sup> of the final model was 0.21.



**Table 1.** *Cont.*

Data available in: <sup>a</sup> 597, <sup>b</sup> 596, <sup>c</sup> 594, <sup>d</sup> 400, <sup>e</sup> 471, <sup>f</sup> 595. Abbreviations: hs-CRP, high-sensitive C reactive protein; HDL, high-density lipoprotein; IQR, interquartile range; LDL, low-density lipoprotein; HbA1C, glycated hemoglobin; SD, standard deviation.



**Table 2.** *Cont.*

\* *<sup>p</sup>* Value <sup>&</sup>lt; 0.1; \*\* *<sup>p</sup>* Value <sup>&</sup>lt; 0.05; \*\*\* *<sup>p</sup>* Value <sup>&</sup>lt; 0.01. † Linear regression analysis; adjusted for age and sex. § Stepwise backwards linear regression analysis; for inclusion and exclusion in this analysis, *<sup>p</sup>* Values were set at 0.1 and 0.05, respectively. <sup>~</sup> Excluded from the final model. Abbreviations: Std. β, standardized beta coefficient; hs-CRP, high-sensitive C reactive protein; HDL, high-density lipoprotein; LDL, low-density lipoprotein; HbA1C, glycated hemoglobin.

#### *3.2. Primary Prospective Analyses*

At a median follow-up of 7.0 (IQR, 6.2–7.5) years, 131 (22%) patients died, of which 32 (24%) deaths were due to cancer (summary of types of cancer can be found in Table S1). Median time from kidney transplantation to cancer death was 12.0 (IQR, 6.2–20.0). In multivariable-adjusted Cox proportional hazards regression analyses, plasma vitamin C concentration was inversely associated with cancer mortality risk (HR 0.50; 95%CI 0.34–0.74; *p* < 0.001), independent of potential confounders including age, sex, smoking status, eGFR, dialysis vintage, time since transplantation, proteinuria, fruit and vegetable intake, diabetes mellitus, hs-CRP, prior history of cardiovascular disease, immunosuppressive therapy and transplantation era (Table 3, Models 1–6) (Figure 1). Full report of coefficient estimates for both the variable of interest plasma vitamin C as well as for potential confounders included in every multivariable model (Models 1–6) are shown in Table S2. Neither significant interaction of the association of vitamin C with cancer mortality was found for potential confounders (Table S3) nor for oxidative stress biomarkers. Results of interaction and subgroup analyses of oxidative stress biomarkers are presented in Figure 2.

**Figure 1.** Association of plasma vitamin C with cancer mortality risk in 598 KTR. Data were fitted by a Cox proportional hazards regression model adjusted for age, sex, smoking status, estimated Glomerular Filtration Rate, dialysis vintage, time since transplantation, proteinuria, fruit and vegetable intake, diabetes mellitus, high-sensitivity C-reactive protein, and prior history of cardiovascular disease (Model 4). The gray areas indicate the 95% CIs. The line in the graph represents the hazard ratio.


**Figure 2.** Interaction and subgroup analyses of the association of plasma vitamin C with cancer mortality. *P*interaction was calculated by fitting models which contain both main effects as continuous variables and their cross-product term. Hazard ratios were calculated with adjustment for age, sex, smoking status, estimated Glomerular Filtration Rate, dialysis vintage, time since transplantation, proteinuria, and fruit and vegetable intake, analogous to Model 3 of the overall prospective analyses. Abbreviations: CI, confidence interval; MDA, malondialdehyde; GGT, gamma-glutamyl transpeptidase.



Cox proportional hazards regression analyses were performed to assess the association of plasma vitamin C with cancer mortality. Model 1: adjustment for age, sex and smoking status. Model 2: Model 1 + adjustment for estimated Glomerular Filtration Rate, dialysis vintage, time since transplantation and proteinuria. Model 3: Model 2 + adjustment for fruit and vegetable intake. Model 4: Model 3 + adjustment for diabetes mellitus, high-sensitivity C-reactive protein and prior history of cardiovascular disease. Model 5: Model 3 + adjustment for immunosuppressive therapy. Model 6: Model 3 + adjustment for transplantation era. Abbreviations: HR, hazard ratio; CI, confidence interval. <sup>a</sup> Each model hazard ratio is given per doubling of vitamin C concentration.

#### *3.3. Secondary Prospective Analyses*

In secondary analyses, at a median follow-up of 7.0 (IQR, 6.2–7.5) years, 131 (22%) patients died, of which 67 (49%) deaths were due to cardiovascular causes. Median time from kidney transplantation to cardiovascular death was 11.0 (IQR, 7.6–14.8). There was no significant association of plasma vitamin C with cardiovascular mortality (HR 1.16; 95%CI 0.83–1.62; *p* = 0.40) (Table 4). This finding remained unaltered after adjustment for potential confounders, analogous to Models 1 to 6 of the primary analyses.


**Table 4.** Association of plasma vitamin C with cardiovascular mortality in 598 kidney transplant recipients.

Cox proportional hazards regression analyses were performed to assess the association of plasma vitamin C with cardiovascular mortality. Model 1: adjustment for age, sex, and smoking status. Model 2: Model 1 + adjustment for estimated Glomerular Filtration Rate, dialysis vintage, time since transplantation and proteinuria. Model 3: Model 2 + adjustment for fruit and vegetable intake. Model 4: Model 3 + adjustment for diabetes mellitus, high-sensitivity C-reactive protein and prior history of cardiovascular disease. Model 5: Model 3 + adjustment for immunosuppressive therapy. Model 6: Model 3 + adjustment for transplantation era. Abbreviations: HR, hazard ratio; CI, confidence interval.

#### *3.4. Sensitivity Analyses*

After performing graft failure-censored Cox proportional hazards regression analyses, our primary findings of the association of plasma vitamin C with both cancer mortality and cardiovascular mortality remained materially unchanged (Tables S4 and S5, respectively). After performing Cox proportional hazards regression analyses of the association of plasma vitamin C with cancer mortality with adjustment for HbA1c instead of diabetes mellitus the association remained materially unchanged (Table S6).

#### **4. Discussion**

In the current study, we show that cancer is a substantially prevalent individual cause of death after kidney transplantation, and that plasma vitamin C concentrations are inversely and independently associated with long-term cancer mortality risk in stable KTR. Secondary analyses did not reveal significant associations with cardiovascular mortality. To the best of our knowledge, this is the first study that provides prospective data supporting vitamin C as a potential risk factor for cancer mortality in KTR.

Our results are in line with previously reported cancer mortality risk data in KTR. Au et al. reported that 16.7% of deaths in a large cohort of KTR were due to cancer after a median follow-up of 6.3 (IQR, 2.3–12.0) years. Although cancer mortality has been previously described as an increasing and imperative problem in KTR [2,5,6,10], there is a paucity of studies exploring potential risk factors and underlying mechanisms leading to this increased cancer mortality in KTR. Immunosuppression following kidney transplant is the most accepted risk factor, specifically CNI [4,6,38,39]. In fact, there is extensive research focused on finding the best combination of immunosuppressants in order to reduce de novo malignancy incidence without increasing rejection rates, where m-TOR inhibitors could have a role in reducing cancer risk [6,40–42]. Noteworthy is that according to our findings, the association of plasma vitamin C concentrations with cancer mortality is independent of immunosuppressive therapies after a kidney transplant.

Low plasma vitamin C has been previously associated with gastric cancer risk in the general population. In this patient setting, mean plasma vitamin C concentration was 39.9 ± 25.2 μmol/L for cases and 41.5 ± 19.4 μmol/L for controls, both comparable to those from our study [14]. Likewise, in the general male population, low plasma vitamin C was linked to an increased risk of mortality with cancer playing a key role. In this study, median plasma vitamin C was 49.4 (IQR, 47.7–51.7) μmol/L [13], also comparable to our study. Furthermore, the anti-cancer properties from vitamin C and other antioxidants have drawn much attention in the oncology research field [43–46]. According to the results of cross-sectional analyses of our study, daily fruit intake was independently associated

with plasma vitamin C levels, congruent with evidence suggesting a diet high in fruits to be associated with decreased cancer risk in various patient settings, with antioxidants playing a key-role [47–53]. Surprisingly, our results show that the association of lower plasma vitamin C with cancer mortality risk is independent of fruit and vegetable intake, introducing vitamin C as a specific therapeutic target in this setting of patients.

A possible explanation for the association we found could be the important role that vitamin C plays as epigenetic modulator in health and disease [43–46], and specifically in cancer cell lines [54]. On the other hand, it is well known that oxidative stress can cause cancer [55,56], due to oxidative damage to deoxyribonucleic acid (DNA) [57]. This oxidative damage is usually counteracted by DNA repair enzymes, but in a pro-oxidant environment, e.g., chronic inflammation and uremic state [58,59], this defense-mechanism is held back [56,60,61]. It has been suggested that antioxidant treatment cannot prevent occurrence of gastrointestinal cancer and that it may even increase overall risk of mortality [55]. However, it has been described that kidney transplant recipients (KTR) have increased oxidative stress [19], which in turn can lead to increased oxidative damage to DNA [57]. Together with decreased immunological surveillance secondary to post-transplant immunosuppression, these phenomena can play a role in increased cancer mortality in KTR and an increased contribution of oxidative stress therein. It can therefore not be excluded that other than subjects of the general population, KTR could benefit from anti-oxidant treatment. High dosages of vitamin C supplementation have been linked to higher risk of development of oxalate kidney stones in male subjects of the general population [62,63]. Vitamin C supplementation may also enhance immunity, which could result in increased risk of rejection. Such effects could limit the utility of vitamin C supplementation in clinical practice and should be taken into account when considering vitamin C supplementation strategies in KTR. Of note, no significant interaction of the association of vitamin C with cancer mortality was found by oxidative stress biomarkers. In light of these results, it could be hypothesized that the inverse association of vitamin C with cancer mortality hereby reported may be explained by its potential role as epigenetic modulator rather than through its antioxidant properties. The latter may be further supported by the finding that plasma vitamin C was inversely associated with cancer mortality independently of fruit and vegetable intake, which suggests that the beneficial effect of vitamin C would not be fully related to the classic theory of dietary intake of natural antioxidants as anticarcinogens [53,57].

Our study has important strengths, including its large sample size of stable KTR, which were closely monitored during a considerable follow-up period by regular check-up in the outpatient clinic, without loss of participants to follow-up. Furthermore, data were extensively collected, allowing to adjust our findings for several potential confounders and predictors of the main results, including current or former smoking status. We acknowledge the study's limitations as the following. First, vitamin C was measured at baseline. Like the current study, most epidemiological studies use a single baseline measurement to predict outcomes, which adversely affects predictive properties of variables associated with outcomes [64–67]. If intra-individual variability of predictive biomarkers using repeated measurements is taken into account, this results in strengthening of predictive properties, particularly in case of markers with high intra-individual variation [64,67]. The lower the intra-individual variation from one measurement to the next would be, the more accurate the single measurement represents the usual level of the marker [64–67]. Noteworthy, evidence available for intra-individual variability of plasma vitamin C suggests that its concentrations relatively stable over time, with a single plasma vitamin C measurement being representative of an individual's status for long periods of time [65]. Moreover, previous epidemiological studies have used a baseline measurement of plasma vitamin C to predict clinical outcomes over a period of several years [68–70]. Second, we measured plasma vitamin C rather than leukocyte vitamin C, which could have provided assessment of tissue vitamin C, and therefore additional information on the role of vitamin C in disease prevention [71]. Third, initiation of vitamin C supplementation during follow-up was not recorded, which could have introduced bias that cannot be accounted for in our analyses. Fourth, incidence and types of non-fatal cancer were not documented, while this information would have been

of added value to the reported findings. With the presented data, we had no power to discriminate the association with cancer mortality by types of cancer, which does not necessarily imply that associations are similar for all types of cancer. Nevertheless, our results show, for the first time, a prospective association of plasma vitamin C with long-term risk of cancer mortality in stable kidney transplant recipients, which holds a plea for future studies in which data on incidence and types of non-fatal cancer are collected. To allow for such studies we have started a new large, long-lasting prospective cohort study in kidney transplant recipients in which collection of such data is included [72]. Another limitation is that history of cured skin cancer was not documented, it could therefore not be included in multivariable analyses. Finally, due to its observational design, conclusions on causality cannot be drawn from our results.

In conclusion, we show that cancer is a substantially prevalent individual cause of death after kidney transplantation, and that plasma vitamin C concentrations are inversely and independently associated with cancer mortality risk. Remarkably, our findings link for the first time plasma vitamin C concentrations with cancer mortality risk in KTR, which underscores that vitamin C may be a meaningful as yet overlooked modifiable risk factor of cancer mortality in KTR. Considering the relative increase in cancer mortality rates in kidney transplant recipients along with the decline in deaths due to cardiovascular causes, it is expected that novel risk management strategies are to emerge. Whether a novel vitamin C-targeted strategy may represent an opportunity to decrease the burden of cancer mortality in KTR requires further studies.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/8/12/2064/s1, Figure S1: Strobe flow diagram, Table S1: Death-related type of cancer, Table S2: Association of plasma vitamin C with cancer mortality, all models, Table S3: Interaction analyses for potential confounders on the association of vitamin C with cancer mortality, Table S4: Sensitivity analysis; association of plasma vitamin C with cancer mortality in 598 kidney transplant recipients, censored for graft-failure, Table S5: Sensitivity analysis; association of plasma vitamin C with cardiovascular mortality in 598 kidney transplant recipients, censored for graft-failure, Table S6: Sensitivity analysis; association of plasma vitamin C with cancer mortality in 598 kidney transplant recipients, with HbA1c instead of diabetes mellitus as potential confounder.

**Author Contributions:** Formal analysis, T.A.G., C.G.S. and D.G.; investigation, T.A.G., C.G.S., D.G., M.F.E., R.A.P., M.H.d.B., R.O.B.G., S.P.B., R.R., G.J.N. and S.J.L.B.; data curation, T.A.G., C.G.S. and D.G.; writing—original draft preparation, T.A.G. and C.G.S.; writing—review and editing, T.A.G., C.G.S., D.G., M.F.E., R.A.P., M.H.d.B., R.O.B.G., S.P.B., R.R., G.J.N. and S.J.L.B.; supervision, R.R., G.J.N. and S.J.L.B.; project administration, R.O.B.G., S.P.B., G.J.N. and S.J.L.B.; funding acquisition, T.A.G., C.G.S. and S.J.L.B.

**Funding:** This study is based on data from the TransplantLines Insulin Resistance and Inflammation Biobank and Cohort Study (TxL-IRI; ClinicalTrials.gov identifier: NCT03272854), which was funded by the Dutch Kidney Foundation (grant C00.1877). Camilo G. Sotomayor is supported by a doctorate studies grant from Comisión Nacional de Investigación Científica y Tecnológica (F 72190118). Funders of this study had no role in the study design, collection of the data, analyzing the data, interpretation of results, writing the manuscript or the decision to submit the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
