**Preoperative Health-Related Quality of Life Predicts Minimal Clinically Important Di**ff**erence and Survival after Surgical Resection of Hepatocellular Carcinoma**

**Chong-Chi Chiu 1,2,3, King-Teh Lee 4,5, Jhi-Joung Wang 6, Ding-Ping Sun 2, Hao-Hsien Lee 1, Chien-Cheng Huang 7,8 and Hon-Yi Shi 5,9,10,\***


Received: 23 February 2019; Accepted: 22 April 2019; Published: 27 April 2019

**Abstract:** Despite the growing use of minimal clinically important difference (MCID) as a cancer outcome measure, no study has reported clinically significant outcomes in cancer patients. We defined MCID and evaluated the use of preoperative HRQoL for predicting MCID and survival after surgical resection of hepatocellular carcinoma (HCC). In total, 369 patients completed the Functional Assessment of Cancer Therapy-Hepatobiliary (FACT-Hep) and the SF-36 at baseline and at two years post-operative at three tertiary academic hospitals. The corresponding MCID values were 3.6 (SF-36 physical component summary), 4.2 (SF-36 mental component summary), 5.4 (FACT-General total score), and 6.7 (FACT-Hep total score). The predictors of achieving postoperative MCID were significantly higher in patients who had low preoperative HRQoL score, advanced age, high education level, and high BMI (*p* < 0.05). However, patients with a high preoperative HRQoL score, high education level, high BMI, and low Charlson comorbidity index score were significantly associated with survival (*p* < 0.05). Preoperative HRQoL scores were predictive of MCID and overall survival after surgical resection of HCC. The findings of this study may be useful for managing the preoperative expectations of candidates for HCC resection and for developing shared decision-making procedures for patients undergoing surgical resection of HCC.

**Keywords:** hepatocellular carcinoma; health-related quality of life; minimal clinically important difference; survival

#### **1. Introduction**

Hepatic resection is the mainstay curative treatment for patients with hepatocellular carcinoma (HCC), even in some patients with early-stage HCC [1–3]. Health-related quality of life (HRQoL) is a recognized indicator of healthcare outcomes and, since the 1990s, evaluations of cancer treatment

outcomes have increasingly emphasized assessment of HRQoL [4,5]. Disease-specific and generic HRQoL measures are often reported together and provide complementary assessments of patient well-being before and after an intervention. It is important not to mix up the concept of quality of life with a recently growing area of HRQoL. Quality of life is an essential concept in the field of international development since it allows analysis of development on a measure broader than the standard of living [6,7]. Within development theory, however, there are varying ideas concerning what constitutes desirable change for a particular society, and the different ways that institutions define the quality of life, therefore, shapes how these organizations work for its improvement as a whole. The Functional Assessment of Cancer Therapy-Hepatobiliary (FACT-Hep) measure is one of the most widely used patient-reported questionnaires for measuring HRQoL in cancer research [8]. Unfortunately, research has shown that most of the studies that have used the FACT-Hep lack any reporting of clinical significance, even though guidelines for assessing clinical significance do exist.

A clinically important difference is a change that a patient or clinician would consider meaningful or worthwhile, such that an intervention or treatment would be considered worthy of repeating or such that patients would consider the change as an improvement in HRQoL. As such, measures of clinical significance such as the minimal clinically important difference (MCID) are increasingly used as a standard clinical outcome measure [6,7]. The MCID is defined as the smallest outcome change that the patient perceives as clinically important [9]. Despite the growing use of MCID as a cancer outcome measure, no study has reported clinically significant outcomes after surgical resection of HCC.

For cancer patients and their families, clinical data for HRQoL outcomes provide a useful indicator of the expected course of recovery and the expected effects of treatment. Thus, HRQoL data can help them make informed treatment decisions [10,11]. Baseline assessments of HRQoL have proven useful for predicting survival in various cancers, including colorectal, esophageal, breast, oropharyngeal, and lung cancers [10–13]. For varying severity of cancer, HRQoL has shown higher sensitivity compared to conventional prognostic indicators and compared to physician assessments [11,14]. Gotay et al. assessed the use of patient-reported HRQoL as a prognostic indicator of cancer outcomes [15]. Out of 39 clinical trials reviewed by the authors, 36 reported at least one HRQoL domain that was a significant predictor of survival. However, comparisons of results published in the literature are difficult because studies differ in the HRQoL measures applied and studied populations differ in patient attributes such as the type, site, and stage of disease [14–17]. Given the variability in reported overall survival rates and overall study heterogeneity, the evidence base for overall survival and determinants of overall survival after surgical resection of HCC are still evolving, and continued investigation is warranted.

A growing body of evidence indicates that preoperative functional status or HRQoL are important determinants of cancer surgery outcomes [14–17]. To the best of our knowledge, no prior study has systematically evaluated the role of preoperative HRQoL in achieving MCID and in overall survival after surgical resection of HCC. Therefore, the purpose of this study was to investigate the use of preoperative HRQoL scores for predicting achievement of MCID and for predicting overall survival after surgical resection of HCC.

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

#### *2.1. Subjects and Data Collection*

This study recruited all patients who had received surgical resection of HCC performed at one of three southern Taiwan medical centers between February 2013 and February 2017. For accurate assessment of postoperative outcome measures, the analysis was limited to patients who had received surgical resection performed by a director of surgery in a medical institution or by a senior attending doctor specializing in HCC surgery or treatment. Inclusion criteria were the following: (1) a histologic or combined radiographic and laboratory diagnosis of HCC, (2) ability to communicate in Chinese or Taiwanese, and (3) agreement to participate in a questionnaire survey performed in the hospital ward or by telephone. Major exclusion criteria included concurrent malignancy or participation in another quality-of-life study that might have interfered with this study. Figure 1 shows that, during the sample selection period, 496 subjects were eligible for participation. Of these, 62 were excluded due to benign tumor or cognitive impairment. Therefore, 369 subjects were assessed preoperatively (baseline) and at 2 years postoperatively. Baseline demographic and clinical data were collected through questionnaire surveys and medical records reviews. This study was approved by the Institutional Review Board of Chi Mei Medical Center (10002-L01).

**Figure 1.** Flow chart showing population changes during the study, including subjects who met initial exclusion criteria, those who later declined to participate and those who lost to follow-up. SF-36: 36-Item Short Form Survey; FACT-Hep: Functional Assessment of Cancer Therapy-Hepatobiliary.

### *2.2. Study Protocol*

Patients were asked to complete the questionnaires during follow-up visits at our outpatient clinic. To maximize compliance and minimize volunteer bias, a research assistant was available to help patients complete the questionnaires during each outpatient session. All HRQoL data were collected by the same research assistant. Patients were informed that their questionnaire responses would not be revealed to their attending surgeons and, hence, would not affect their treatment.

#### *2.3. Measures of HRQoL*

The Short Form-36 (SF-36) Health Survey measures eight dimensions: physical function, role limitation due to physical health, bodily pain, general health, vitality, social function, role limitation due to emotional health, and mental health. To compare the overall physical and mental functioning of the study population with those in the general Taiwan population, physical component summary scores (PCS) and mental component summary scores (MCS) were calculated by norm-based scoring methods and used as dependent variables [18]. Based on a previous study [19], the PCS and MCS were computed in comparison with the general population of Taiwan. Values below 50 indicated that the examined PCS or MCS were below the average values for the general Taiwan population, and vice versa.

The FACT-Hep measure contains five dimensions: physical well-being, social/family well-being, functional well-being, emotional well-being, and additional concerns. The subscales for the physical well-being, social/family well-being, and functional well-being dimensions each contained seven items with a subscale score range of 0–28 points; the subscale for emotional well-being contained six items with a subscale score range of 0–24 points; the subscale for additional concerns about HCC contained 18 items with a subscale score range of 0–72 points [8]. The Functional Assessment of Cancer Therapy-General (Fact-G) and additional concerns for HCC scores were summed to obtain the FACT-Hep total score, which ranged from 0 to 180. Higher scores on all FACT-Hep dimensions were interpreted as better HRQoL and fewer symptoms.

#### *2.4. Statistical Analysis*

Studies show that a distribution-based method can reliably derive MCID calculated as one-half the standard deviation (SD) in outcome score change from baseline to the two-year follow up for a given instrument in a patient cohort [9,20]. Therefore, this methodology was used to determine MCID values for the SF-36 PCS, SF-36 MCS, FACT-G total score, and FACT-Hep total score.

Multivariable logistic regression models were used to identify predictors of the achievement of MCID after surgical resection of HCC. A Cox multivariable proportional hazard regression model was also used to evaluate how other prognostic factors affect survival. Survival distribution was estimated by Kaplan–Meier method. Significant differences in survival probability were stratified by a log-rank test. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated from regression coefficients.

Variables included in the multivariable analyses were gender, age, marital status, education, body mass index (BMI), Charlson co-morbidity index (CCI) score, co-residence with family, smoking, drinking, tumor stage, chemotherapy, radiotherapy, and average length of stay (ALOS). Multivariable analyses also included preoperative HRQoL score. Variables that fell out of the model were excluded from the tables of multivariable results. Statistical analyses were performed using SPSS software (IBM SPSS Statistics for Windows, Version 20.0, Armonk, NY, USA). All statistical tests were two tailed with a significance level of 0.05.

#### **3. Results**

#### *3.1. Patient Demographics*

The SF-36 and FACT-Hep measures were completed by 369 HCC surgery patients preoperatively and at two years postoperatively. We compared the patients who remained in the study throughout the two-year period with those who were lost or dead to follow up between the baseline and the second year after discharge. There was no difference in terms of gender, age, marital status, education, BMI, CCI score, co-residence with family, smoking, drinking, tumor stage, chemotherapy, radiotherapy, ALOS, or in any preoperative HRQoL parameters mentioned above (data not shown). Table 1 presents their demographic and clinical characteristics.


**Table 1.** Demographic and clinical characteristics of 369 patients with hepatic resection for hepatocellular carcinoma.

SD: standard deviation.

#### *3.2. HRQoL Outcomes*

The patients had a mean age of 60.2 ± 10.8 years, and 73.4% (271) patients were male. Table 2 shows that mean patient-reported HRQoL scores at two years after surgery were significantly higher than those before surgery (*p* < 0.001). The MCID values were 3.6 for the SF-36 PCS; 4.2 for the SF-36 MCS; 5.4 for the FACT-G total score; and 6.7 for the FACT-Hep total score.

**Table 2.** Mean ± standard deviation for SF-36 and Functional Assessment of Cancer Therapy-Hepatobiliary (FACT-Hep) before and after resection for hepatocellular carcinoma (*n* = 369) \*.


\* Both PCS and MCS scores were converted to obtain a mean of 50 and a standard deviation of 10 compared to the normal (nationwide) group. SF-36, 36-Item Short Form Survey; PCS, physical component summary; MCS, mental component summary; FACT-G, Functional Assessment of Cancer Therapy-General; FACT-Hep, Functional Assessment of Cancer Therapy-Hepatobiliary.

#### *3.3. Multivariable Analyses*

Multivariable analyses of HRQoL and survival were performed to identify predictors of the achievement of MCID after surgical resection for HCC. For each HRQoL measure, a high preoperative score negatively predicted achievement of MCID (*p* < 0.001) (Table 3). According to the SF-36 PCS data, the odds of achieving MCID were lower in males than in females (odds ratio (OR), 0.31; 95% CI, 0.12–0.83) but were higher in patients with advanced age (OR, 1.05; 95% CI, 1.01, 1.10), high education level (OR, 1.14; 95% CI, 1.01, 1.30), and high BMI (OR, 1.11; 95% CI, 1.09, 1.12) compared to their counterparts with young age, low education level, and low BMI, respectively. According to the SF-12 MCS data, the odds of achieving MCID were lower in patients with high BMI compared to those with low BMI (OR, 0.91; 95% CI, 0.84, 0.99); however, the odds of achieving MCID were higher in those with advanced age (OR, 1.01; 95% CI, 1.01, 1.02) and high CCI score (OR, 1.53; 95% CI, 1.13, 1.94) compared to their counterparts with young age and low CCI score, respectively. According to the FACT-G total data, the odds of achieving MCID were higher in patients with advanced age (OR, 1.04; 95% CI, 1.01, 1.07), high education level (OR, 1.12; 95% CI, 1.04, 1.21), and high BMI (OR, 1.19; 95% CI, 1.09, 1.29) compared to their counterparts with young age, low education level, and low BMI, respectively. According to the FACT-Hep total data, the odds of achieving MCID were higher in patients with advanced age (OR, 1.05; 95% CI, 1.02, 1.07), high education level (OR, 1.11; 95% CI, 1.03, 1.19), and high BMI (OR, 1.09; 95% CI, 1.01, 1.18) compared to their counterparts with young age, low education level, and low BMI.


**Table 3.** Odds of achieving minimal clinical important difference (MCID) in health-related quality of life according to multivariate logistic regression model \*.

\* The full model was adjusted for preoperative functional status, gender, age, marital status, education, body mass index, Charlson co-morbidity index, co-residence with family, smoking, drinking, tumor stage, chemotherapy, radiotherapy, and average length of stay. SF-36, 36-Item Short Form Survey; PCS, physical component summary; MCS, mental component summary; FACT-G, Functional Assessment of Cancer Therapy-General; FACT-Hep, Functional Assessment of Cancer Therapy-Hepatobiliary.

Multivariate analyses of each pre-operative HRQoL score showed that a high score was a positive predictor of overall survival (Table 4). Lower pre-operative HRQoL scores were significantly associated with post-operative morbidity (*p* < 0.05). Education level, BMI, and CCI also showed significant associations with overall survival (*p* < 0.001).

**Variable HR (95% CI)** *p* **Value** Preoperative SF-36 PCS score 1.05 (1.03, 1.08) <0.001 Preoperative SF-36 MCS score 1.03 (1.01, 1.05) <0.001 Preoperative SF-36 physical function 1.06 (1.01, 1.10) <0.001 Preoperative SF-36 role physical 1.03 (1.00, 1.05) <0.001

**Table 4.** Overall survival analysis by Cox multivariable proportional hazard regression model \*.


\* The full model was adjusted for preoperative functional status, gender, age, marital status, education, body mass index, Charlson co-morbidity index, co-residence with family, smoking, drinking, tumor stage, chemotherapy, radiotherapy, and average length of stay. SF-36, 36-Item Short Form Survey; PCS, physical component summary; MCS, mental component summary; FACT-G, Functional Assessment of Cancer Therapy-General; FACT-Hep, Functional Assessment of Cancer Therapy-Hepatobiliary; HR, hazard ratio; CI, confidence interval.

#### **4. Discussion**

This study investigated how patient-reported preoperative HRQoL affects two outcomes of surgical resection of HCC: MCID and overall survival. Multivariate analyses showed that each preoperative HRQoL score was predictive of both MCID and overall survival. Low preoperative HRQoL score, advanced age, high education level, and high BMI were significantly associated with achievement of postoperative MCID (*p* < 0.05). Additionally, high preoperative HRQoL score, high education level, high BMI, and low CCI had significant positive associations overall survival (*p* < 0.05). It demonstrates that, at baseline, preoperative HRQoL scores relates to postoperative mortality. Lower scores in physical and functional domains are associated with an increased risk of postoperative mortality. The importance of preoperative HRQoL scores for predicting outcomes of surgical resection in HCC patients is now well recognized [9,21]. The current study found that, for a given HRQoL outcome measure, a high preoperative score was significantly for not achieving a postoperative MCID in the outcome measure. The likely explanation for this finding is that patients who already have high HRQoL scores before surgery and less potential for achieving a HRQoL score improvement that meets the criteria for an MCID.

This study aimed to calculate and report the MCID value of commonly used HRQoL scales. Changes in HRQoL by time and/or treatment may not correlate with the direction (positive vs. negative) as well as the magnitude of clinical improvements in outcomes as they are perceived by the patients. Furthermore, cancer treatment has a more significant impact on HRQoL among HCC surgical patients. The MCID for (SF-36 PCS, SF-36 MCS, and FACT-G total score) value changes differed across domains, and they differed for perceived improvement and deterioration. We knew that domain scores related to physical function diminished from pre-treatment to on- or immediately after treatment, and emotional

function improved. Additionally, the initial anxiety of the diagnosis and treatment initiation period may have been ameliorated by subsequent familiarity and supportive psychosocial care provided by the clinical service teams after patients finished the initial questionnaire. Thus, this might explain why those patients with a low preoperative score did achieve a postoperative MCID in the outcome measure and regarded as "efficacy" when compared with that achieved by those patients who had full pre-operative familiarity and mental support and thus noted with a high preoperative score.

The significant associations revealed by the HRQoL instruments investigated in this study underscore the relationship between HRQoL measures and medical outcomes. The study showed that preoperative HRQoL scores accurately predict postoperative MCID and overall survival, which is consistent with the literature [11,15,17]. Therefore, counseling is essential for apprising HCC resection candidates of expected postoperative improvements and impairments. If medical outcomes are considered benchmarks, then the preoperative HRQoL score, which is an important predictor of postoperative MCID and overall survival, is crucial.

Until now, no studies have described the implications of significant changes in disease-specific and generic HRQoL outcome measures in patients who have undergone HCC surgery. Steel et al. evaluated the clinical meaningfulness of FACT-Hep scores in HCC patients [22]. The authors combined distribution-based analyses with cross-sectional anchor-based analyses to obtain minimally important differences (MIDs) in FACT-G subscale scores (2–3 points), FACT-G total scores (6–7 points), Hepatobiliary Cancer Subscale scores (MID 5–6 points), and FACT-Hep scores (MID 8–9 points). However, data for clinically significant improvements in HCC surgery outcomes, particularly patient-reported outcomes, are very limited [4–6]. For patients who undergo surgical resection of HCC, the current study obtained MCID values of 3.6 for SF-36 PCS; 4.2 for SF-36 MCS; 5.4 for FACT-G total score; and 6.7 for FACT-Hep total score. These data are a novel addition to the literature and provide a useful reference for further studies of HRQoL outcomes after surgical resection of HCC. Notably, this study used patient-reported data for the period from before surgery until two years after surgery. A distribution-based method was used to calculate MCID. Since MCID values may change depending on the time point studied and the psychometric method used for analysis, further studies are needed to investigate MCID after surgical resection of HCC using different time points and mixed anchor/distribution-based methods of deriving MCID values.

The evidence base for overall survival after surgical resection of HCC is growing but is still relatively limited. The overall survival rates reported in the literature are somewhat variable, and the studied populations have been heterogeneous. Quinten et al. investigated the prognostic relationship between HRQoL and survival in a dataset for 30 randomized controlled trials performed by the European Organization for Research and Treatment of Cancer [11]. Their study showed that, in each cancer site, at least one HRQoL domain had an additive prognostic value that exceeded the prognostic values of clinical and sociodemographic variables. A systematic literature review by Quinten et al. confirmed that baseline HRQoL and at least one HRQoL domain were significantly associated with overall survival [11]. The current study found that preoperative HRQoL scores, education level, and BMI had significant positive associations with overall survival (*p* < 0.05) whereas CCI score had a significant negative association with overall survival (*p* < 0.05).

The HRQoL factors identified in this study varied from those in previous studies [9,20,22]. One possible explanation is differences in study populations. Our study focused on patients in both early and advanced stages of HCC whereas previous studies have only focused on patients in advanced stages of the disease. Secondly, patients with different cultural backgrounds may have different perceptions of HRQoL. Thirdly, patients in recent studies have more treatment options compared to patients in earlier studies, which can result in different perceptions of the implications of HCC and thus different perceptions of HRQoL. Fourthly, even studies that use the same HRQoL measure may have very different data analysis methodologies.

The findings of this study have important implications for preoperative counseling of patients, management of patient expectations, and stratification of outcomes. Healthcare providers increasingly emphasize shared decision making and are now using predictive modeling to inform surgery patients about potential clinical outcomes and the likelihood of success [11,22]. The results of the present study suggest that healthcare providers should consider routinely administering HRQoL instruments preoperatively as screening tools and for informing shared decision-making strategies. Patients with low HRQoL scores can be referred for counseling to modify their outcome expectations or referred for targeted interventions to optimize their HCC resection outcomes.

Certain limitations of this study are noted. Firstly, the patient data were derived from a multiinstitutional HCC registry containing data contributed by multiple surgeons. As such, other than institutional best practices, surgical techniques and rehabilitation protocols were not standardized. However, since patient-reported HRQoL outcomes were obtained at the time of each clinical encounter, the reports of functional status are assumedly accurate. Another limitation is that sensitivity of the MCID values was not analyzed. The MCID can be calculated according to a consensus of or by using an anchor-based method or a distribution-based method [9,23]. Each methodology for deriving MCID has its associated pitfalls, and none has consistently proven to be superior. The applied methodology should be selected according to the characteristics of the data and the disease under study [23]. Additionally, the role of the FACT-Hep for assessing outcomes after surgical resection of HCC has not been robustly studied or validated. As such, the responsiveness of the HRQoL outcomes for this population subset is not clear; for example, the reliability (and change over time) of emotional, psychological, and social responses after surgical resection of HCC needs further study.

#### **5. Conclusions**

In conclusion, this study revealed that HRQoL scores are independent predictors of MCID and overall survival after surgical resection of HCC. The MCID values were 3.6 for the SF-36 PCS; 4.2 for the SF-36 MCS; 5.4 for the FACT-G total score; and 6.7 for the FACT-Hep total score. The findings of this study may be useful for preoperative management of patient expectations and for developing shared decision-making measures for patients undergoing surgical resection of HCC.

**Author Contributions:** Study concept and design: C.-C.C., and H.-Y.S.; acquisition of data: K.-T.L., J.-J.W., D.-P.S., H.-H.L., and C.-C.H.; analysis and interpretation of data: C.-C.C., and H.-Y.S.; drafting of the manuscript: C.-C.C., and H.-Y.S.; revised the manuscript: K.-T.L., J.-J.W., D.-P.S., H.-H.L., and C.-C.H.; approved the manuscript: all authors.

**Funding:** This study was supported by grants from the Chi Mei Medical Center, Liouying, Research Grant (CLFHR 10301 and 10409) and from the Chi-Mei Medical Center and the Kaohsiung Medical University Research Foundation (107CM-KMU).

**Acknowledgments:** We thank all the individuals who participated in this study, without which this study would not have been possible.

**Conflicts of Interest:** There 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* **Re-Evaluating the Protective Effect of Hemodialysis Catheter Locking Solutions in Hemodialysis Patients**

### **Chang-Hua Chen 1,2,3,4,\*, Yu-Min Chen 5, Yu Yang 6, Yu-Jun Chang 7, Li-Jhen Lin <sup>2</sup> and Hua-Cheng Yen <sup>8</sup>**


Received: 5 March 2019; Accepted: 20 March 2019; Published: 25 March 2019

**Abstract:** Catheter-related bloodstream infections (CRBSIs) and exit-site infections (ESIs) are common complications associated with the use of central venous catheters for hemodialysis. The aim of this study was to analyze the impact of routine locking solutions on the incidence of CRBSI and ESI, in preserving catheter function, and on the rate of all-cause mortality in patients undergoing hemodialysis. We selected publications (from inception until July 2018) with studies comparing locking solutions for hemodialysis catheters used in patients undergoing hemodialysis. A total of 21 eligible studies were included, with a total of 4832 patients and 318,769 days of catheter use. The incidence of CRBSI and ESI was significantly lower in the treated group (citrate-based regimen) than in the controls (heparin-based regimen). No significant difference in preserving catheter function and all-cause mortality was found between the two groups. Our findings demonstrated that routine locking solutions for hemodialysis catheters effectively reduce the incidence of CRBSIs and ESIs, but our findings failed to show a benefit for preserving catheter function and mortality rates. Therefore, further studies are urgently needed to conclusively evaluate the impact of routine locking solutions on preserving catheter function and improving the rates of all-cause mortality.

**Keywords:** effect; protection; catheter; hemodialysis; meta-analysis; trial sequential analysis

#### **1. Introduction**

#### *1.1. Variety of New Strategies for Locking Solutions to Avoid Catheter Infection and Catheter Malfunction in Hemodialysis Patients*

Infections are widely prevalent in patients on chronic hemodialysis, and mortality from infection account for 10% of deaths observed in patients undergoing hemodialysis [1]. The use of central venous catheters in hemodialysis has been associated with catheter-related bloodstream infections (CRBSIs) and exit-site infections (ESIs) [2–4]. Although recent efforts have minimized the use of catheters, the proportion patients with end-stage renal disease undergoing dialysis using central venous catheters has not yet declined [5]. Protective strategies against CRBSI and catheter malfunction are necessary [6], and to this end, the use of heparin as a routine locking solution for central venous catheters has become an accepted clinical practice [4]. However, heparinized locking solutions might cause unintended complications, such as systemic anticoagulation effects, bleeding episodes, heparin-induced thrombocytopenia, and susceptibility to bacterial biofilm formation [7–9]. A variety of new locking solutions have been developed; this includes citrate, which has antimicrobial properties [10–12]. However, the disadvantages of citrate compared with heparin have been raised and included the ability of avoiding catheter malfunction, citrate toxicity, and induction of cardiac arrhythmia [13]. Weijmer et al. showed that a 30% citrate solution was superior to heparin in preventing CRBSI [14]. In contrast, other studies have reported that the use of citrate does not have an advantage over heparin in preventing CRBSI [4,12]. Currently, the findings of the studies comparing citrate with heparin locking solutions are inconclusive for protecting against CRBSI and ESI and preserving the catheter function. Clinicians question if locking solutions should be considered a modifiable risk factor for CRBSIs in patients undergoing hemodialysis. Furthermore, the recommended locking solution for the routine care of patients undergoing hemodialysis continue to remain questionable.

#### *1.2. Rationale for Re-Evaluating the Protective Effect of Hemodialysis Catheter Locking Solutions in Hemodialysis Patients*

Routine locking solutions for hemodialysis catheters are recommended with category II evidence according to the guideline by the Healthcare Infection Control Practices Advisory Committee in 2011 [15]; however, there are some limitations of the studies providing the current and update evidence. Mostly, conclusions of meta-analysis could be influenced by the heterogeneity between individual studies and insufficient information size. Quantification of the required information size [16] is important to ensure the reliability of the data. In addition, current meta-analyses lack information size calculation [17–24]. Additionally, the incidence of CRBSI is difficult to evaluate because of their subjectivity for case finding, lack of specificity, and high inter-observer variability. CRBSI is associated with high morbidity and mortality in patients undergoing hemodialysis [1], and the prevention of CRBSI and ESI is becoming increasingly essential. Given these limitations, we performed a meta-analysis and trial sequential analysis to assess the impact of routine locking solutions on the incidence of CRBSI and ESI, in preserving catheter function, and on the rate of all-cause mortality in patients undergoing hemodialysis. We grouped the eligible publications according to combination regimen, antimicrobial activity, and concentration of the locking solutions; thereafter, we grouped according to the study design to assess its potential effect on the reported outcomes.

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

#### *2.1. Search Strategy and Inclusion Criteria*

The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the institutional review board of Changhua Christian Hospital (CCH IRB No. 180801). From the earliest record to July 2018, we searched PubMed, Scopus, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, ClinicalTrials.gov, Embase, and Web of Science databases for studies on locking solutions for central venous catheters used in hemodialysis of patients. Full search strategies for each database are available in the Appendix A. The reference lists of the eligible publications were manually reviewed for relevant studies. Articles published in languages other than English or those with no available full text were excluded.

We included all trials and studies that provided data on one or more of our target outcomes for both the treated group and control group: CRBSIs and ESIs. Two investigators (CHC and YMC) independently reviewed potential trials and studies for inclusion. Disagreements were resolved by consensus. We also tried to contact the corresponding authors of selected papers to provide clarifications and missing data where needed.

#### *2.2. Definition of Study Outcomes*

Based on the original studies, the treated group comprised of patients undergoing hemodialysis using citrate as the locking solution for central venous catheters; for the control group heparin was used as the locking solution (Table 1). The outcomes of the original studies were included in this meta-analysis. The primary outcomes included (1) CRBSI, defined as bacteremia caused by an intravenous catheter, and (2) ESI, defined as the development of a purulent redness around the exit site that did not result from residual stitches. The secondary outcomes included (1), the need to remove the catheter due to catheter malfunction; and (2) the need for thrombolytic treatments due to catheter malfunction; and (3) all-cause mortality at any timeframe. Incidence was presented as the number of episodes per catheter or per patient depending on the available data.

#### *2.3. Data Extraction and Quality Assessment*

Two reviewers examined all retrieved articles and extracted data using a pre-determined form, recording the name of the first author, year of publication, country where the study was conducted, study design (RCT or observational studies), demographic and disease characteristics of participants, number of participants enrolled, and quality assessment of each study. Each reviewer independently evaluated the quality of the eligible studies, using Jadad scoring [25] for the RCTs and the Newcastle-Ottawa quality assessment scale [26] for the comparative experimental studies.

#### *2.4. Data Synthesis and Analysis*

The outcomes were measured by determining the odds ratios (ORs). A random effects model was used to pool individual ORs. Analyses were performed with the Comprehensive Meta-Analysis software version 3.0 (Biostat, Englewood, NJ, USA). Between-trial heterogeneity was determined using *I* <sup>2</sup> tests; values > 50% were regarded as considerable heterogeneity [27]. Funnel plots and Egger's test were used to examine potential publication bias [27]. Statistical significance was defined as *p* < 0.05, except for the determination of publication bias where *p* < 0.10 was considered significant. This study was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Table S1) [28].

In trial sequential analyses, the inconsistence of heterogeneity (*I* 2) adjusted by determining the required information size. The required information size was calculated with an intervention effect of a 10% relative risk reduction, an overall 5% risk of a type I error, and a 20% risk of a type II error. All trial sequential analyses were performed using TSA version 0.9 Beta (www.ctu.dk/tsa/, Copenhagen Trial Unit, Copenhagen, Denmark).


**Table 1.** Summary of the retrieved trials investigating experimental group and control group.


**Table 1.** *Cont.*

Abbreviations: AU, Australia; Au, Austria; BZ, Brazil; CA, Canada; CH, China; CiT, citrate; EDTA, tetra-sodium ethylenediaminetetraacetic acid; EtOH, ethanol; FR, France; GM, gentamicin; GR, Greece; HpR, heparin; IR, Iran; MMP, 0.15% methylene blue + 0.15% methylparaben + 0.015% propylparaben; NaCl, sodium chloride; N, number; NL, Netherlands; PL, Poland; QA, quality assessment; RCT, randomized controlled trial; SI, Slovenia; TMP/SMX, cotrimoxazole (=trimethoprim/sulfamethoxazole); TRD, taurolidine ; UFH, unfractionated heparin; UK, United Kingdom; US, United States. #, the study was evaluated using Jadad scale. \*, the study was assessed using the Newcastle-Ottawa scale.

#### **3. Results**

#### *3.1. Eligible Studies*

The literature search yielded 458 potentially eligible articles. By screening the abstracts, we removed 350 irrelevant articles. The remaining 100 articles were assessed further by full-text reading, of which 79 were excluded (Figure 1). Thus, 21 selected articles comparing citrate with heparin lockings for central venous catheters used in hemodialysis were included in this meta-analysis [4,6,8,12–14,17–24,29–35].

**Figure 1.** Preferred reporting items for systematic reviews and meta-analyses (PRISMA) flow diagram for the search and identification of the included studies.

The studies published in the selected articles were conducted from the earliest record to July 2018, with a total of 4832 patients and 318,769 total days of catheter use. Six studies compared citrate alone with heparin lockings; 14 studies tested regimens of citrate and other antimicrobials (gentamicin, taurolidine, methylparaben, methylene blue, and propylparaben) with heparin lockings; and two studies compared ethanol or combination solution (citrate, heparin and taurolidine) with non-heparin locking. Studies were conducted in North America (5 studies), South America (1), Europe (12), and Asia (3). A variety of end points were used in these studies. Most studies reported on CRBSI (17 studies [4,6,8,12–14,17,19,21–24,30–34]), followed by ESI (11 studies [4,8,12,14,17–19,24,30,31,33]), catheter removal for poor flow (9 studies [6,8,12,14,18,24,29,31,33]), thrombolytic treatment (8 studies [4,8,14,17,18,32,33]), and mortality (5 studies [6,14,19,32,33]). The characteristics of the studies fulfilling the inclusion criteria are listed in Table 1. Thirteen studies were identified as RCT, and 6 studies were not double-blinded (Table 1).

#### *3.2. Pooled Odds for Primary Outcomes and Subgroup Analysis*

#### 3.2.1. Catheter-Related Bloodstream Infection (CRBSI)

Seventeen studies (1731 patients; 217,128 catheter days) reported on CRBSI. The incidence of CRBSI was significantly lower in the treated group compared with the control group (OR, 0.424; 95% CI, 0.267–0.673; *p* < 0.001) (Figure 2). CRBSI subgroup analysis showed that the OR appeared to have a tendency to favor the treatment groups with either the combined regimen (OR, 0.206; 95% CI, 0.058–0.730; *p* = 0.027), the single regimen (OR, 0.289; 95% CI, 0.083–0.365; *p* = 0.037), a regimen containing antibiotics (OR, 0.136; 95% CI, 0.051–0.365; *p* = 0.002), or a low concentration of a major regimen (OR, 0.421; 95% CI, 0.186–0.956; *p* = 0.039; Table 2).

**Figure 2.** Forest plot of the overall odds ratios for catheter-related bloodstream infection in the treated group versus the control group. The random model of overall odds ratio showed a significant overall effect of interventions in reducing the risk for developing catheter-related bloodstream infections as compared with the control condition (OR, 0.424; 95% CI, 0.267–0.673; *p* < 0.001).

#### 3.2.2. Exit-Site Infection (ESI)

A total of 11 RCTs (2,425 patients; 231,086 catheter days) described ESI. The incidence of ESI was significantly lower in the treated group compared with the control group (OR, 0.627; 95% CI, 0.441–0.893; *p* = 0.001; Figure 3). Further focusing at exit-site infection (Table 3), the subgroup analysis (combined regimen, regimen containing antibiotic, and concentration of regimen for exit-site infection) disclosed no significant differences between any groups except for combined regimen.


**Table 2.** Subgroup analysis of odds ratio based on study designs, combined regimen, regimen containing antibiotic, and concentration of regimen for CRBSI.

Abbreviation: RCT, randomized controlled trial.

**Table 3.** Subgroup analysis of odds ratio based on study designs, combined regimen, regimen containing antibiotic, and concentration of regimen for exit site infection.


Abbreviation: RCT, randomized controlled trial.

**Figure 3.** Forest plot of the overall odds ratios for exit-site infection in treated group versus the control group. The random model of overall odds ratio for exit-site infection showed a significant overall effect of interventions in reducing the risk for developing exit-site infection as compared with the control condition (OR, 0.627; 95% CI, 0.441–0.893; *p* = 0.001).

#### *3.3. Pooled Odds for Secondary Outcomes and Subgroup Analysis*

#### 3.3.1. Catheter Withdrawal Due to Malfunction

Nine studies (1826 patients; 205,163 catheter days) reported catheters being removed for poor blood flow. As shown in Figure 4, no difference was identified between the two groups (OR, 0.696; 95% CI, 0.397–1.223; *p* = 0.208). Further subgroup analysis (combined regimen, regimen containing antibiotic, and concentration of regimen for catheter removal due to catheter malfunction) failed to reveal any differences between any groups (Table 4).

**Figure 4.** Forest plot of the overall odds ratios for catheter removal due to catheter malfunction in the treated group vs. the control group. The random model of overall odds ratio for the need to remove the catheter for malfunction showed a significant overall effect of the interventions in reducing the risk for catheter removal compared with the control condition (OR, 0.696; 95% CI, 0.397–1.223; *p* = 0.208).


**Table 4.** Subgroup analysis of odds ratio based on study designs, combined regimen, regimen containing antibiotic, and concentration of regimen for catheter removal due to catheter malfunction.

Abbreviation: RCT, randomized controlled trial.

#### 3.3.2. Thrombolytic Treatment Due to Catheter Malfunction

Overall, in eight RCTs (2092 patients; 220,460 catheter days) included in this meta-analysis the patients underwent thrombolytic treatment [4,8,14,17,18,32,33]. The incidence of thrombolytic treatment was not significantly lower in the treated group compared with the control group using the random-effects model (OR, 1.105; 95% CI, 0.655–1.573; *p* = 0.946; Figure 5). Thrombolytic treatment subgroup analysis showed no differences in the OR between the two groups (Table 5).

**Figure 5.** Forest plot of the overall odds ratios for thrombolytic treatments for catheter malfunction in the treated group versus the control group. The random model of overall odds ratio for the need to administer thrombolytic treatment for catheter malfunction showed a significant overall reduced risk for receiving thrombolytic treatments with interventions as compared with the control condition (OR, 1.105; 95% CI, 0.655–1.573; *p* = 0.946).


**Table 5.** Subgroup analysis of odds ratio based on study designs, combined regimen, regimen containing antibiotic, and concentration of regimen for the need of thrombolytic treatment for catheter malfunction.

Abbreviation: RCT, randomized controlled trial.

#### 3.3.3. All-Cause Mortality

The meta-analysis included five RCTs (2,327 patients) comparing all-cause mortality rate between the two groups; no significant difference was identified (OR, 0.909; 95% CI, 0.580–1.423; *p* = 0.676; Figure 6). The corresponding subgroup analysis (combined regimen, regimen containing antibiotic, and concentration of regimen for all-cause mortality) showed no apparent differences between the two groups (Table 6).

**Figure 6.** Forest plot of the overall odds ratios for all-cause mortality in the treated group versus the control group. The random model of overall odds ratio for all-cause mortality rate showed a significant overall effect of the interventions in reducing mortality rate as compared with the control condition (OR, 0.909; 95% CI, 0.580–1.423; *p* = 0.676).


**Table 6.** Subgroup analysis of odds ratio based on study designs, combined regimen, regimen containing antibiotic, and concentration of regimen for all-cause mortality.

Abbreviation: RCT, randomized controlled trial.

#### *3.4. Pooled Odds for Outcomes in Trial Sequential Analysis*

In trial sequential analysis between the treated and control groups, the overall OR of CRBSI was 0.439 (95% CI, 0.290–0.668; *p* < 0.001; Figure 7a), the OR of ESI was 0.644 (95% CI, 0.469–0.883; *p* = 0.006; Figure 7b), the OR of the need to remove the catheter for catheter malfunction was 0.746 (95% CI, 0.431–1.293; *p* = 0.151; Figure 7c), the OR of the need to receive thrombolytic treatment for catheter malfunction was 1.015 (95% CI, 0.655–1.573; *p* = 0.461; Figure 7d), and the OR of all-cause mortality was 0.976 (95% CI, 0.663–1.439; *p* = 0.296; Figure 7e).

#### (**b**) **Trial sequential analysis of exit-site infection**

**Figure 7.** *Cont*.

#### (**e**) **Trial sequential analysis of all-cause mortality**

**Figure 7.** Trial sequential analysis of the odds ratio for evaluation event: (**a**) Trial sequential analysis of catheter-related bloodstream infection. Trial sequential analysis of 17 studies with a lower risk of bias in reporting catheter-related bloodstream infection, with a control event proportion of 17%, diversity of 45%, type I error of 5%, power of 80%, and relative risk reduction of 30%. The required information size of 630,022 was not reached and none of the boundaries for benefit, harm, or futility were crossed, leaving the meta-analysis inconclusive at a 30% relative risk reduction. The overall OR of CRBSI was 0.439 (95% CI, 0.290–1.668; *p* < 0.001); (**b**) trial sequential analysis of exit-site infection. Trial sequential analysis of eleven studies with low risk of bias reporting exit-site infection, with a control event proportion of 17%, diversity of 30%, type I error of 5%, power of 80%, and relative risk reduction of 30%. The required information size of 336,863 was not reached and none of the boundaries for benefit, harm, or futility were crossed, leaving the meta-analysis inconclusive at a 30% relative risk reduction. The OR of ESI was 0.644 (95% CI, 0.469–0.883; *p* = 0.006); (**c**) trial sequential analysis of nine studies with a lower risk of bias reporting the need to remove the catheter for catheter malfunction, with a control event proportion of 17%, diversity of 71%, type I error of 5%, power of 80%, and relative risk reduction of 30%. The required information size of 625,306 were not reached and none of the boundaries for benefit, harm, or futility were crossed, leaving the meta-analysis inconclusive at a 30% relative risk reduction. The OR of the need to remove the catheter for catheter malfunction was 0.746 (95% CI, 0.431–1.293; *p* = 0.151); (**d**) trial sequential analysis of thrombolytic treatments for catheter malfunction. Trial sequential analysis of nine studies with low risk of bias reporting the need to receive thrombolytic treatment for catheter malfunction, with a control event proportion of 17%, diversity of 91%, type I error of 5%, power of 80%, and relative risk reduction of 30%. The required information size of 615,306 were not reached and none of the boundaries for benefit, harm, or futility were crossed, leaving the meta-analysis inconclusive at a 30% relative risk reduction. The OR of the need to receive thrombolytic treatment for catheter malfunction was 1.015 (95% CI, 0.655–1.573; *p* = 0.461); (**e**) trial sequential analysis of all-cause mortality. Trial sequential analysis of five studies with a lower risk of bias reporting all-cause mortality, with a control event proportion of 17%, diversity of 78%, type I error of 5%, power of 80%, and relative risk reduction of 30%. The required information size of 8419were not reached and none of the boundaries for benefit, harm, or futility were crossed, leaving the meta-analysis inconclusive at a 30% relative risk reduction. The OR of all-cause mortality was 0.976 (95% CI, 0.663–1.439; *p* = 0.296). Notes: The solid blue line is the cumulative Z-curve. The vertical black dashed line is required information size. The green dashed lines represent the trial sequential monitoring boundaries and the futility boundaries.

#### *3.5. Funnel Plot for the Overall OR of the Included Studies among Four Outcomes*

We examined possible sources of underlying heterogeneity across studies. With regards to OR heterogeneity, the *I* <sup>2</sup> value was calculated in both the overall studies included. In the funnel plot of the OR for evaluation event, the *I* <sup>2</sup> value of CRBSI was 70.1% (*p* = 0.303, Figure 8a), ESI was 28.0% (*p* = 0.010; Figure 8b), the need to remove the catheter for catheter malfunction was 55.9% (*p* = 0.208; Figure 8c), the need to receive thrombolytic treatment for catheter malfunction was 88.69% (*p* = 0.946; Figure 8d), and all-cause mortality was 88.6% (*p* = 0.804; Figure 8e).

**Figure 8.** Funnel plot of the odds ratio for evaluation event: (**a**) Funnel plot of the odds ratio of catheter-related bloodstream infection. *I* <sup>2</sup> value, 70.1%; *p* = 0.303; (**b**) funnel plot of the odds ratio of exit-site infection. *I* <sup>2</sup> value, 28.0%; *p* = 0.010; (**c**) funnel plot of the odds ratio of catheter removal for catheter malfunction. *I* <sup>2</sup> value, 55.9%; *p* = 0.208; (**d**) funnel plot of the odds ratio of thrombolytic treatments for catheter malfunction. *I* <sup>2</sup> value, 88.69%; *p* = 0.946; (**e**) funnel plot of the odds ratio of all-cause mortality. *I* <sup>2</sup> value, 88.6%; *p* = 0.804. Regarding odds ratio heterogeneity, the *I* <sup>2</sup> value in both the overall studies included is indicated for each case. Egger's test revealed the existence of significant publication bias regarding the overall odds ratios, *p*-value is indicated for each case.

#### **4. Discussion**

Our meta-analysis and trial sequential analysis shows that routine locking solutions for hemodialysis catheters could effectively reduce the incidence of CRBSI and ESI. Our current meta-analysis, based on 21 selected studies with a total of 6118 participants, showed that the incidence of CRBSI and ESI significantly decreased in the treated group relative to the control group, that is less infections when using citrate or citrate mixtures versus heparin. Moreover, we found no significant difference in preserving catheter function, including in the need for catheter withdrawal or for thrombolytic treatment due to catheter malfunction, between the treated and control groups. We found no significant alteration in all-cause mortality between the two groups. The lack of statistical significance may not only be due to the heterogeneity and underlying variance in the outcomes of each regimen, but also due to inadequate required information sizes, as revealed by the trial sequential analysis. Regular locking care with citrate is standard practice for patients undergoing hemodialysis in many healthcare institutes, but not in some countries including Taiwan. Our updated review suggests that the role of routine locking solutions in preventing CRBSI and ESI in hemodialysis patients is robust. However, it does not show a benefit in preserving catheter function in hemodialysis patients, including in the need to remove catheters or in the need for thrombolytic treatment for catheter malfunction.

The current study shows that the incidence of CRBSI significantly decreased in the treated group relative to the control group, which is consistent with previous studies [36,37]. Subgroup analyses based on the type of locking solutions for hemodialysis catheters revealed that the usage of citrate-base regimens was associated with a lower incidence of CRBSI [4,14]. Our subgroup analysis for the concentration of citrate used showed that the incidence of CRBSI was similar in treated group, although the American Society of Diagnostic and Interventional Nephrology and the European Renal Best Practice recommend 4% citrate to be used as a catheter locking solution [38,39]. In some countries, including Taiwan, 4% citrate is still not routinely used in locking solutions for hemodialysis catheters. The current meta-analysis emphasizes that 4% citrate shows a benefit and could be routinely used as a locking solution for hemodialysis catheters.

Our current study shows that the incidence of ESI is significantly decreased in the treated group compared with the control group. Our result is in agreement with previous studies [14,19]. In some studies, patients received additional antibiotic ointments at the exit site during dressing changes, which could reduce the incidence of ESI [8,14,40]. After subgroup categorization, there is no significant difference between two groups except for combined regimen, which could result from the heterogeneity of the included studies and inadequate information size.

We found no significant difference in preserving catheter function between the treated and control groups, including the need to remove catheters or the need for thrombolytic treatment. However, Yahav et al. reported that citrate reduced catheter removals [41]. This incongruity may arise from the following: (1) Variation in enrollment criteria and definitions for the spectrum of catheter removal and (2) the number of cases is still limited because the meta-analysis information size does not meet the required information size. Concerning thrombolytic treatments and thrombosis episodes, our report is similar to previous studies [41,42]. Focusing on the need to remove catheters and to receive thrombolytic treatment for catheter malfunction, further large-scale RCTs are necessary to elucidate this issue for preserving catheter function.

The possible association between the two groups and all-cause mortality was not statistically significant in the current study (OR, 0.909; 95% CI, 0.580–1.423; *p* = 0.676). Subgroup analysis showed no difference in all-cause mortality. Mortality due to CRBSIs or ESIs account around one-tenth of all hemodialysis patient deaths [1–4]. Protective strategies with locking solutions to prevent CRBSIs and ESIs in hemodialysis patients still cannot decrease the mortality rate. Further large-scale RCTs are necessary to elucidate modifiable risk factor for decreasing morality in hemodialysis patients.

Guidelines for the Prevention of Intravascular Catheter-Related Infections has been published by the Center for Disease Control and Prevention [15], which recommends using prophylactic antimicrobial locking solution in patients undergoing hemodialysis who have a history of multiple CRBSI, despite optimal maximal adherence to aseptic techniques (Category II). This recommendation has been embraced by some dialysis centers due to the low execution rate of locking solutions in preventing CRBSI in hemodialysis patients. In fact, many challenges persist in managing daily care in dialysis centers, such as a lack of safety locking solutions for hemodialysis catheters, lack of a designated health-care workers to perform locking care, limited training on catheter care among health-care workers of dialysis centers, potential hemodialysis patients' noncompliance due to discomfort, as well as health-care workers being unable to maintain high adherence rates in conducting care procedures.

The current study has several limitations. Firstly, the enrolled trials and studies included in the primary analysis dealt with different indications for outcome measures by randomizing a variety of patient groups in different clinical settings. Thus, there is the risk of introducing potentially heterogeneity. Additionally, it is difficult to perform a subgroup analysis based on conditions, such as catheter type, heparin dosage, and other differences in individual unit practices. Secondly, differences in the study individuals, disease severity, setting, and type of infections between individual studies made the study population highly heterogeneous. The *I* <sup>2</sup> value for OR heterogeneity ranged from 25% to 50%, and this heterogeneity would impact the findings of this meta-analysis. Thus, the influence of measurement precision was considered when reporting treatment effectiveness using ORs. Due to the lack of adjusted data in our selected trials, we compiled the unadjusted ORs. We therefore suggest that future similar trials should record serial changes in catheter function and infection status to provide a more accurate indication of clinical effectiveness. Regardless of aforementioned limitations, we have minimized bias throughout the process by our methods of study identification, data selection, and statistical analysis, as well as in our control of publication bias. These steps should strengthen the stability and accuracy of the meta-analysis. Our findings of this meta-analysis are reliable to provide suggestions for improving clinical care.

#### **5. Conclusions**

In conclusion, our study demonstrated that routine locking solutions for hemodialysis catheters could effectively reduce the incidence of CRBSI and ESI. Our findings showed no benefit of routine locking solutions for hemodialysis catheters in decreasing all-cause mortality as well as preserving catheter function, including in the need to remove catheters and in the need to receive thrombolytic treatment, both due to catheter malfunction. The latter results lack statistical significance and the comparisons are limited due to the heterogeneity of the included trials and inadequate information size. Therefore, further well-conducted observational studies and randomized controlled trials are urgently needed to conclusively evaluate the impact of routine locking solutions on preserving catheter function and improving the rates of all-cause mortality.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/8/3/412/s1, Table S1: PRISMA 2009 Checklist.

**Author Contributions:** Conceptualization, C.-H.C., Y.-M.C., Y.Y., Y.-J.C., L.-J.Y. and H.-C.Y.; Methodology, C.-H.C., Y.-M.C. and Y.-J.C.; Software, C.-H.C., Y.-M.C. and Y.-J.C.; Validation, C.-H.C., Y.-M.C., Y.Y. and Y.-J.C.; Formal Analysis, C.-H.C., Y.-M.C. and Y.-J.C.; Investigation, X.X.; Data Curation, C.-H.C.; Writing—Original Draft Preparation, C.-H.C.; Writing—Review and Editing, C.-H.C., Y.-M.C., Y.Y. and Y.-J.C.; Writing—Approval, C.-H.C., Y.-M.C., Y.Y., Y.-J.C., L.-J.Y. and H.-C.Y.; Visualization, C.-H.C. and Y.-M.C.; Project Administration, C.-H.C.; Funding Acquisition, C.-H.C.

**Funding:** The present work was partially supported by a grant obtained from the Changhua Christian Hospital (106-CCHIRP-097).

**Acknowledgments:** The authors thank Ping-Tao Tseng for technical analysis. The authors thank the staff at the Epidemiology and Biostatistics Center, the Department of Pharmacology, and the Department of Computer at Changhua Christian Hospital for the literature findings. This research project would not have been possible without the support of many people. The authors wish to express their gratitude to the staffs of the Department of Critical Care, Epidemiology and Biostatistics Center, the Division of Nephrology, the Division of Infectious Diseases, the Department of Pharmacology, the Department of Nursing, and the Department of Healthcare Quality at Changhua Christian Hospital who were extremely helpful and provided invaluable assistance and support.

**Conflicts of Interest:** All authors declare that they have no competing interests. The sponsors had no role in the design, execution, interpretation, or writing of the study.

### **Appendix A**

Supplement Search strategy in PubMed

#1 lock

#2 filling solution

#3 #1 or #2

#4 End-Stage Kidney Disease or Disease, End-Stage Kidney, End Stage Kidney Disease, End-Stage Chronic Kidney Failure, End-Stage Renal Disease, End-Stage Renal Disease, Chronic Chronic Renal Failure, or ESRD

#5 Renal Dialyses, Renal Dialysis, Hemodialyses, Extracorporeal Dialysis or Renal replacement therapy #6 Catheter Related Infections, Catheter-Related Infection

#7 #4 or #5 and #6

#8 #3 and #7

### **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/).
