**4. Discussion**

In this bi-centric study, an analysis of 105 kidney transplantations of deceased donors, allocated within the Eurotransplant Senior Program, was conducted. We aimed to identify risk factors for ICU admission after KT during a hospital stay in times of shortened PPE and ICU capacities because of the SARS-CoV-2 pandemic.

Overall, recipient and graft characteristics were comparable with other cohorts [19–22]. CIT was lower, lasting on average 9.5 h, while most other ESP programs have CITs averaging 10 to 12 h [19,20]. ESP aims to reduce CIT by prioritizing local organ allocation, as longer CITs have been clearly linked with higher DGF rates. Nonetheless, our DGF rate of 40% is higher than that of one of the largest ESP cohorts so far, with 1406 KTs, by Frei et al. They reported a median DGF rate of 29.7% [19]. In contrast, other groups have comparable DGF rates ranging between 34.7 to 41.1% in their ESP cohorts [22,23]. Chavalitdhamrong et al. even stated a DGF rate of 60.4% for 601 KTs, but for organs allocated by ECD (extended criteria donors) for donors aged 50–69 years, and 63.9% for donors aged ≥70 years [24].

In a high-risk cohort like ESP recipients, complications are common. There were 11.4% intraoperative complications, and 26.7% minor and 33.3% major complications occurred postoperatively, according to Clavien–Dindo. Reports on complication rates state highly variable results, mainly due to inconsistent definitions. Bentas et al. have "surgical complications" in 47% of cases in their ESP program, whereas Bahde et al. reported 15.7% intraoperative and 22.5% post-operative surgical complications among their recipients [23,25]. Only Gallinat et al. defined postoperative complications according to Clavien–Dindo. In their comparison of very old donors in the ECD program, the rate for major complications was 48%, defined as <sup>≥</sup>grade 3b [26].

During follow-up, death-censored graft survival (1- and 5-year: 84% and 73%) and patient survival (1- and 5-year: 85% and 62%) were superior to Frei et al. and comparable with Quast et al., who retrospectively analyzed 217 ESP transplantations at their department from 1998 to 2014, considering donor age [19,20] (see Table 6). In accordance with Boesmueller and Giessing et al., the main reason for graft loss was death with functioning graft [18,22]. Our analysis comprises one of the longest follow-ups in ESP so far. Overall, graft-survival after 9 years was 42%, and patient survival was 38%. Quast et al. reported a 10-year patient survival of 40% for old donors, and 35% for very old donors, whereas graft survival was 30% and 10%, respectively.


**Table 6.** Comparison of death-censored graft and patient survival in ESP programs.

1 Only considering old, but not very old, donors.

Based on this data, we have identified risk factors for ICU admission during a hospital stay in the ESP. In times of the SARS-CoV-2 pandemic with a shortage of ICU capacities, risk stratification is crucial to identify patients at high risk for ICU admission (after KT). This aspect has rarely been addressed so far. To the best of our knowledge, only three working groups have stratified their data for ICU admission [27–29]. Two working groups focused on ICU admission at any time after KT, even years after KT, which clearly does not help when trying to decide whether or not to perform a KT during the present SARS-CoV-2 pandemic. Abrol et al. retrospectively analyzed 1527 kidney transplantations between 2007 and 2016 and found higher age, increasing BMI, pre-transplant dialysis

and deceased donor transplantation to be associated with ICU admission in their multivariate analysis. Living donor KT and preemptive KT were associated with a lower risk [27]. Nonetheless, 82.8% of the included KTs were living kidney transplantations. As such, we are the first to report on the risk for ICU admission immediately after kidney transplantation in the ESP.

17 (16.2%) patients in our cohort were admitted to the ICU for a mean time of 2 days. More than 80% of patients were admitted directly postoperatively or within four days after KT. The main cause for ICU admission was significant hypotension requiring catecholamines. Overall, patients admitted to the ICU had a lower BMI, and CAD as well as hypertensive nephropathy were more common. Graft characteristics and surgical outcomes during transplantation did not differ. The DGF rate of patients admitted to the ICU was high, with 52.9%, but did not significantly differ from patients without an ICU stay (37.5%).

As stated elsewhere, neither the donor nor the recipient's age had an impact on the postoperative course [18,20]. Therefore, age did not affect ICU admission rates in the regression analysis. We assume that within this (very) old patient cohort, age differences were not as important as in younger patient cohorts due to preselection during the workup for listing. As patients admitted to the ICU had a lower BMI, an increasing BMI lowered the risk for ICU admission (OR 0.8, see Table 4). This is an interesting finding, referring to the 'obesity paradox', which describes the association of obesity with higher mortality in the general population on the one hand, but with a survival advantage among obese patients with several diseases on the other hand. In this regard, meta-analyses have shown that patients with a higher BMI might have (i) a reduced risk of ICU admission or death when suffering from pneumonia, (ii) a reduced adjusted mortality when admitted to the ICU with sepsis, severe sepsis or shock, and (iii) a lower mortality on mechanical ventilation in an ICU [30–32]. Although the concept of the obesity paradox has been questioned, there is also convincing evidence for underlying molecular mechanisms, i.e., that a lower energy reservoir in underweight patients cannot equally counteract the adverse influence of increased catabolic stress [33,34].

As further variables, hypertensive nephropathy and CAD increased the OR for ICU admission by 4 and 4.5, respectively. Most patients were admitted to the ICU because of hypotension as a major symptom for cardiac insufficiency, which is more likely in patients with CAD. In addition, hypertensive nephropathy has been linked with a higher risk for cardiovascular events and death [35]. When combining these three independent risk factors in a risk model, it gained a c-index of 0.789 with a sensitivity of 94.1%, a FNR of 5.9% and a NPV of 97.8% (see Table A1). For this reason, our risk model is highly valuable for the identification of patients at high risk for ICU admission. When applied to our cohort, the risk model was false negative in only one case. We are aware that it has a rather low specificity and PPV, whereas the FPR is high. Furthermore, the confidence intervals for the corresponding odds ratios are large, because only 17 (16.2%) patients were admitted to the ICU and not all of them suffered from hypertensive nephropathy or CAD (see Table 1). However, the high sensitivity and NPV of more than 94% render our risk model an ideal search test.

Our patient who had an organ offer in ESP during the SARS-CoV-2 pandemic had a probability of 92.8% to be admitted to the ICU according to our risk model, with a hypertensive nephropathy, CAD and BMI of 29.4 kg/m<sup>2</sup> (see A1 for further explanation). Of note, this patient was not included within the analyzed cohort. Indeed, after transplantation, he had to be admitted to the ICU on postoperative day seven due to urosepsis and suspected cardiac infarction. Infectious complications are common among old kidney recipients and have been shown to be their second most frequent cause for DWFG [16]. In our cohort, 3 out of 17 (17.6%) patients had to be admitted to the ICU because of sepsis. Especially in the context of the ongoing SARS-CoV-2 pandemic, the question of how to manage immunosuppression for KT recipients is still a matter of debate [11,12].

As standard, all patients were administered tacrolimus, mycophenolate mofetil, (methyl)prednisolone and basiliximab for induction therapy. Consequently, the regimen did not affect ICU admission rates. Since lymphopenia has been associated both with a higher risk for SARS-CoV-2 infection and for severe forms of Covid-19, the questions (i) whether or not to perform the transplantation at all and (ii) whether the induction therapy should be reduced were intensively discussed at the transplant center which had an organ offer in ESP during the SARS-CoV-2 pandemic [11]. Finally, the patient was transplanted and received an induction therapy with basiliximab, and unfortunately suffered from sepsis and neutropenia. For this reason, mycophenolate mofetil was stopped and the dose of prednisolone reduced. Of note, SARS-CoV-2 had been ruled out prior to transplantation and after the onset of sepsis again; as we have not experienced a major shortage of ICU capacities, we could guarantee maximum care for this patient at all times. However, we might decide differently if we receive another organ offer in the ESP program during the ongoing SARS-CoV-2 pandemic again.

Interestingly, ICU admission also proved to be an excellent indicator for the identification of patients at risk for short graft and patient survival. In Kaplan–Meier analysis, patients admitted to the ICU had a significantly shorter graft survival of 59.1 months; all of them died within five years (see Figure 2). Consequently, ICU admission impacted patient survival with a HR of 4.72, but did not impact graft survival in Cox regression (see Table 5). Diabetes mellitus was the only other covariate impacting patient survival. Other studies were inconclusive about the effect of pre-transplant diabetes mellitus or new-onset diabetes mellitus (NODAT) on patient survival. Some studies have found associations with NODAT, but not pre-transplant diabetes, with mortality and graft failure, and others the inverse [36–38]. By contrast, ICU admission did not impact death-censored graft survival in Cox regression. The individual number of kidney transplantations per patient (HR 9.66), number of HLA-mismatches (HR 1.53) and the serum creatinine one month after transplantation (HR 1.37) were significant. The negative impact of increasing HLA-mismatches on graft survival was reported more than two decades ago [39]. To shorten waiting times for old recipients, ESP does not integrate HLA-matching in the allocation algorithm.

This analysis is not devoid of limitation. To exclude center-specific factors and enlarge cohort size, we performed a bicentric analysis and included 105 patients. This is a rather low sample size, but big sample sizes in ESP programs are rare. Due to its retrospective nature, we could not test our new risk model for ICU admission in a prospective, independent manner. Before extrapolating our results to other centers, an external validation of our risk model will be needed. For this reason, we encourage other transplantation centers to test our risk model to further enhance its validity. With a bigger cohort size, the confidence intervals for the risk factors BMI, CAD and hypertensive nephropathy will potentially be reduced. Currently, our risk model is an excellent search test, but has a rather low PPV and therefore cannot replace individual and local risk assessment in times of reduced ICU capacities during the SARS-CoV-2 pandemic.
