Comparison of clinical effects between icodextrin and glucose solutions on outcomes of peritoneal dialysis: systematic review and meta-analysis of randomized controlled trials

Icodextrin enhances peritoneal filtration for patients on peritoneal dialysis (PD). However, clinically important outcomes have not yet been analyzed using authentic, objective statistical methods. The present systematic review aimed to determine the risks and benefits of icodextrin compared with a glucose-based solution with respect to clinically important and patient-centered outcomes. We systematically investigated only randomized controlled trials (RCTs) by adopting the Cochrane Database of Systematic Review (2014) and searched the CENTRAL, MEDLINE, and EMBASE databases for eligible studies reported in the literature. The quality of the evidence was assessed using the GRADE approach. We finally evaluated important outcomes in 13 RCTs. Icodextrin significantly decreased the number of reported episodes of uncontrolled fluid overload in four RCTs that involved 236 patients (relative risk [RR], 0.31; 95% confidence interval [CI], 0.12 to 0.82; moderate certainty evidence). However, the inclusion of icodextrin for peritoneal ultrafiltration did not significantly differ in six RCTs involving 252 patients (mean difference [MD], 186.76 mL; 95% CI, − 47.08 to 420.59; low certainty evidence). Regarding other clinically important outcomes, all-cause mortality in 10 RCTs involving 1106 patients (RR, 0.75; 95% CI, 0.33 to 1.71; low certainty evidence) and technical survival in five RCTs involving 470 patients (RR, 0.57; 95%CI, 0.29 to 1.12; low certainty evidence) were not significant. Urine volume in four RCTs involving 136 patients, residual renal function in five RCTs involving 181 patients and peritoneal function measured as the ratio of solute concentration in dialysate and plasma (D/P ratio) in two RCTs involving 105 patients were not specifically affected by icodextrin, and the results for adverse events were similar between icodextrin and glucose PD solutions. Icodextrin could relieve uncontrolled fluid overload without adding risk. However, a significant effect on clinically relevant outcomes such as technical survival and overall patient survival was not suggested. More trials are required to increase the statistical power and to verify the value of icodextrin in clinical practice. PROSPERO, CRD42018104360


Background
Guidelines need to be updated to reflect the current status of patients on peritoneal dialysis (PD) in Japan. Therefore, reliable revised guidelines should be formulated that contain graded evaluations and recommendations to improve patient prognosis. To ensure the reliability of medical information, the Japan Society for Dialysis Therapy (JSDT) revised the clinical practice guideline (CPG) in 2016 based on the GRADE (Grading of Recommendations Assessment, Development and Evaluation) system [1], which is the current global standard method of creating guidelines. In the meantime, icodextrin has become prevalent in Japanese routine clinical practice during the past two decades, and it now occupies an extremely important position based on distinctive characteristics. Ultrafiltration of the peritoneal membrane is essentially important as it is associated with the management of fluid status and it is a determinant of whether or not to continue with PD therapy [2,3]. Icodextrin is a high-molecular-weight, watersoluble glucose polymer in a sustained colloid osmotic gradient that helps to control excess fluid in patients treated by PD [4]. Icodextrin facilitates prolonged and stable peritoneal filtration regardless of the peritoneal function of patients and has thus improved the global quality of clinical peritoneal dialysis. Considerable evidence has recently accumulated about the role of icodextrin in clinical practice [5,6]. However, icodextrin as a peritoneal solution has not been fully evaluated at the level of patient-centered outcomes. Furthermore, patient values have not been adopted to evaluate its effect. Thus, the risks of benefits of icodextrin remain obscure. A systematic review should compare clinical effects between icodextrin and glucose solutions.

Inclusion and exclusion criteria
We included RCTs that compared icodextrin with glucose solutions among patients on PD. The exclusion criteria comprised PD with a neutral-buffered solution, peritonitis associated with PD, and studies other than RCTs. Patients who received PD and HD combination therapy were not included in the RCTs adopted in the present systematic review (SR).

Searches
We created a SR that complied with Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines [7] (Additional file 1). Before starting the present SR, we registered the review protocol (PROSPERO: CRD42018104360) and adopted the extant a reliable SR, the Cochrane Database of Systematic Reviews (2014) for an initial systematic review [8]. That search was in accordance with the Cochrane Renal Group's Specialized Register that comprises studies identified from the CENTRAL, MEDLINE, and EMBASE databases. We then searched MEDLINE and Ichushi-Web to identify further investigation to date. Four reviewers (AK, TF, EF, and KW) independently screened all titles, abstracts and the full texts of articles.

Data extraction
Four reviewers (AK, TF, EF, and KW) independently extracted data using a pre-specified format in advance and integrated the results. Disagreements about data collection were resolved by consultation with YT. One reviewer (TF) verified the results, and another (KW) stored the data extracted from each study in a specific format suitable for analysis.

Risk of bias (quality) assessment
Two independent reviewers (AK, TF) used the standard Cochrane Collaboration risk of bias tool to assess bias in all eligible studies evaluate it in the present SR. Discrepancies were resolved by group discussion to reach a consensus. The following sources of bias were assessed: random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective outcome reporting (reporting bias), and other sources. We evaluated the certainty of evidence for the main outcomes using the GRADE approach [9].

Strategy for data synthesis
Data were integrated using a random-effects model. Results are expressed as RRs with 95% CIs for dichotomous outcomes, and the MD was used for continuous outcomes. Heterogeneity was examined using χ 2 on N-1 degrees of freedom with an alpha level of 0.05 for statistical significance, and I 2 tests. Data were analyzed using Review Manager (RevMan Version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). P values of < 0.05 were defined as statistically significant. Figure 1 shows a PRISMA flow diagram of the present SR. We extracted and assessed 106 studies of SRs and meta-analysis published between 1968 and 2017. Among them, four SRs and one CPG met the eligibility criteria and the antecedent Cochrane review was adopted. We subsequently selected 10 RCTs that compared icodextrin and glucose solutions. While searching RCTs published after the initial SR, we added three RCTs that satisfied our inclusion conditions among 260 studies, and finally, used 13 RCTs as core studies.

Characteristics of included studies
We analyzed data of 1275 patients that included 677 who were treated by PD using icodextrin . Among 13 studies, five, three, and one were of patients treated with continuous ambulatory peritoneal dialysis (CAPD) , automated peritoneal dialysis (APD) [11,13,14], and continuous cycling peritoneal dialysis (CCPD) [15], respectively, and the remainder comprised combinations of the modalities. Two studies involved patients only with diabetes [16,17] and one of the other studies was limited to non-diabetic patients [18]. Three studies allowed a choice of dialysates with appropriate glucose concentrations to achieve desirable control of edema or blood pressure [10,14,19]. Three studies were restricted to patients with high or high-average peritoneal solute transport as the principal inclusion criterion [13,17,20]. Two studies excluded patients with uncontrolled volume status at the time of entry [10,21]. Table 1 shows the characteristics and outcomes of the included studies. We also evaluated each outcomes according to the GRADE approach ( Table 2).   The Glucose group showed a significant decline in urine volume at 6 and 12 months from 0 month, but the icodextrin group did not. Figure 2 shows the risk of bias in the included studies. Risk of bias was generally low for random sequence generation and allocation concealment, but most studies were considered unclear regarding blinding of outcome assessments and outcome data were incomplete in five studies.

All-cause mortality
The effects of icodextrin and glucose solutions on patient survival did not significant differ in 10 RCTs of 1106 patients (RR, 0.75; 95% CI, 0.33 to 1.71; P = 0.49, I 2 = 0%; low certainty evidence; Fig. 3). However, the point estimate was better for icodextrin than glucose solutions by 7 patients' reductions among 1000 patients.

Peritoneal ultrafiltration
Icodextrin solution did not lead to a significant increase in peritoneal ultrafiltration compared with glucose solutions in six RCTs of 252 patients (MD, 186.76 mL/day; 95% CI, − 47.08 to 420.59; P = 0.12, I 2 = 64%; low certainty evidence; Fig. 6). Moderate to severe heterogeneity might have been derived from the study design. The design of five studies was openlabel and dropout rates were high in all of them. These results might also be biased and attenuated in relation to icodextrin because hypertonic 3.86% and 4.25% glucose PD solutions served as controls without restrictions in two trials [15,19].

Residual renal function
Residual renal function determined from glomerular filtration rates or renal creatinine clearance was similar between icodextrin and glucose PD solution in five RCTs of 181 patients (MD, 0.56 mL/min; 95% CI − 0.37 to 1.49; P = 0.24, I 2 = 0%; moderate certainty evidence; Additional file 2: Figure S2). Among the included studies, one study used the averages of creatinine and urea clearance to indicate residual renal function [14], and this was adopted and evaluated in the preceding SR [8].

Peritoneal function
We adopted dialysate-to-plasma creatinine ratio (D/P Cr) as a marker of peritoneal function. However, the amount of clinical research was insufficient to evaluate outcomes, and icodextrin did not affect the D/P Cr with a moderate level of heterogeneity in two RCTs that included 105 patients (MD, 0.001; 95% CI, − 0.07 to 0.07; P = 0.97, I 2 = 65%; very low certainty evidence; Additional file 2: Figure S3).

Rash
The occurrence of rash elicited by icodextrin and glucose PD solutions did not significantly differ in four RCTs of 855 patients (RR, 1.84; 95% CI, 0.48 to 7.09; P = 0.35, I 2 = 46%; low certainty evidence; Additional file 2: Figure S5). Most reports are based on comparative trials that processed at the time when icodextrin entered the market.

Discussion
The present review found that the quality of evidence supporting icodextrin was moderate and significantly associated with a decreased frequency of uncontrolled fluid overload compared with glucose solutions. However, icodextrin did not contribute to improved allcause mortality and technical survival. In addition, icodextrin was not associated with increased urine volumes, or the D/P creatinine ratio as an indicator of peritoneal creatinine clearance. Recent results have been consistent with those of an earlier systematic review, which concluded that icodextrin alleviated uncontrolled fluid overload better than glucose solutions [5]. Uncontrolled fluid overload was not strictly defined in the studies included herein, and only one study found excessive volumes lead to technical failure [16]. On the other hand, icodextrin was not associated with a significant increase in peritoneal filtration compared to glucose solutions in the present SR. The lack of significance might be derived, at least in part, from the following. The present SR included prospective RCTs that allowed the choice of dialysates with glucose concentrations up to 4.25% [15,19]. These studies had considerable weight in the overall evaluation of peritoneal ultrafiltration outcomes. Additionally, the present SR defined peritoneal ultrafiltration as overall daily output, and not as a longer dwell period. Therefore, some studies did not meet our inclusion criteria. The volume of peritoneal ultrafiltration was significantly greater in the control than the icodextrin group in one study, which could be explained by the dialysate comprising > 2.5% glucose [19]. The study protocol required that the control group underwent four exchanges of dialysate with glucose, whereas icodextrin was applied once for the long dwell and the glucose dialysate was exchanged twice in the experimental group. Another explanation might be associated with a need to convert the standard error or 95%CI into standard deviation, which contributed heterogeneity [18,21]. These factors might have diminished the advantage of icodextrin in terms of peritoneal ultrafiltration compared with glucose solutions.
One study recommended that APD and icodextrin should be considered for patients with high or highaverage solute transport and that issues related to high transport could be avoided by using icodextrin for the long exchange to prevent fluid reabsorption [22]. Patients with fluid reabsorption in the long dwell should be identified to maximize the benefits of icodextrin and consider its appropriate indication in clinical practice.
The occurrence of rash in icodextrin was marginally significant compared with glucose solution. Previous studies have concluded that the incidence of rash is not significantly higher for icodextrin than glucose solutions, and notable, rash was most frequently reported when icodextrin first entered the market [13,23]. However, the prevalence of rash has remarkably decreased since then [10,11]. The present SR included maculopapular reactions as rash, unlike previous studies. Consequently, the incidence of rash did not significantly differ between icodextrin and glucose solutions.
Here, we determined which outcomes were important to clinical practice after discussion at two consecutive panel meetings. Differences in their opinions were addressed by voting among committees and the process of creating an SR was developed. Four independent investigators collected the studies for review and analyzed the data. A standardized method of peritoneal function assessment has not been established and the numbers of RCT were insufficient to evaluate the reliability of this parameter as an indicator. Therefore, the importance of peritoneal function, determined as D/P creatinine, was not as critical as all-cause mortality and technical survival. Finally, the last panel meeting concluded that peritoneal function as an outcome should be modified and downgraded.
The present review had some limitations. Firstly, the degree of heterogeneity was high in eligible studies with respect to the glucose concentrations of control PD dialysates. Secondly, some outcomes had no definite criteria or indicators, which might have resulted in selection bias derived from the extracted studies. As to technical survival, we defined it as the number of patients who were forced to discontinue PD and transited to HD treatment, while death, transplantation, recovery of renal function, and loss to follow-up were not counted. Consequently, the difference in the definition of technical survival resulted in a discrepancy of the number of events reported in each study. Thirdly, we use icodextrin for patients with excessive fluid tendencies in clinical practice. However, the patients in the reviewed studies did not necessarily have issues with fluid management. This discrepancy might lead to a  concern regarding routine icodextrin treatment for more efficient water removal among patients on PD without a fluid excess. Finally, subgroups were not analyzed when analytical findings found no overall effect. Therefore, patients that might benefit from icodextrin could not be confirmed for each outcome.

Conclusions
The present SR suggests that icodextrin could reduce the frequency of uncontrolled fluid overload and ameliorate impaired peritoneal ultrafiltration among patients with PD. However, all-cause mortality, technical survival and peritoneal ultrafiltration did not significantly differ between icodextrin and PD solutions. The present SR reflects an important relationship between icodextrin dialysate and patient-centered outcomes among PD patients.
Additional file 1: The PRISMA checklist of items to include when reporting a systematic review.
Additional file 2: Figure S1. Effects of icodextrin on urine volume. Figure S2. Effects of icodextrin on residual renal function. Figure S3.
Effects of icodextrin on peritoneal function. Figure S4. Effects of icodextrin on peritonitis. Figure S5. Effects of icodextrin on rash.