Study Inclusion and Exclusion Criteria

Studies that met all the following basic criteria were included in our meta-analysis: (1) a randomized controlled trial (RCT) for patients on continuous ambulatory peritoneal dialysis (CAPD) or automated peritoneal dialysis (APD) as the treatment of ESRD; (2) LS was compared with CS. The crossover randomized trials or RCTs that did not assess RRF were excluded.

Search Strategy

We identified eligible RCTs by searching the PubMed, Embase, Wiley, Scopus, Ovid databases and abstracts presented at the annual meetings of the American Society of Nephrology (ASN), the National Kidney Foundation (NKF), and the European Renal Association (ERA), from inception to July 2014, using appropriate Medical Subject Headings (MeSH) and text words: peritoneal dialysis, glucose degradation products, biocompatible solution, low-GDP, APD, CAPD in combination with “residual renal function”. Further, the reference lists of retrieved articles were then searched for additional relevant studies. No language restrictions were imposed.

Study Selection

We included RCTs examining the effect of LSs on RRF in PD patients >18 years old compared with CSs. PD modality was restricted as either CAPD or APD. The outcomes of RCTs should include the RRF value, which is measured as the arithmetic means of residual renal clearances of urea and creatinine by collecting 24-hour urine volume. Other endpoints for the evaluation may include small solute clearance, peritoneal solute transport rate (PSTR), nutritional status, and all-cause mortality of PD patients. The study had at least 12 months of duration of follow-up without restriction on sample size. Two investigators (NZ and JW), independently, screened titles and abstracts of all electronic citations to select studies that met the inclusion criteria for further analysis. All articles identified by the investigators were retained.

Study Validity Assessment

We used the Cochrane Collaboration’s bias tool and Jadad score for assessing the risk of bias for the included studies. The first approach incorporates assessment of randomization (sequence generation and allocation sequence concealment), blinding (participants, personnel, and outcome assessors), completeness of outcome data, selection of outcomes reported, and other sources of bias. The items were scored with “yes,” “no,” and “unclear” ¹⁸. The Jadad scale score ranged from 0 to 5 points about the randomization, double-blinding, and withdrawals and dropouts ¹⁹.

Data Extraction

Two investigators extracted the useful data independently and reached a consensus on all eligible data. Relevant information was obtained by contacting the corresponding authors of the respective studies.

Study characteristics were extracted from all included trials with respect to year of publication, the study sample, baseline characteristics of the trials, follow-up, and the following reported outcomes of different follow-up months (baseline, 6, 12, and 24 months): (1) RRF (mL/min) (2) total weekly urea clearance (total Kt/V) and peritoneal urea clearance (peritoneal Kt/V), (3) total creatinine clearance (total CrCl) (L/week/1.73m²), and peritoneal creatinine clearance (peritoneal CrCl) (L/week/1.73m²), (4) daily urine volume (UV) (mL), daily peritoneal ultrafiltration (UF) (mL) and daily glucose exposure (g), (5) dialysate-to-plasma ratio of creatinine at 4 hours of peritoneal equilibration test (PET) (D/Pcr) and D/D0 glucose at 4 hours (D/D0 glucose), (6) blood pressure (mmHg) including systolic blood pressure (SBP) and diastolic blood pressure (DBP), (7) nutritional data, including serum albumin (g/dL), subjective global assessment (SGA) and normalized protein nitrogen appearance (nPNA) (g/kg/day), (8) all-cause mortality.

Data Synthesis and Analysis

Continuous outcomes results were presented as the mean difference (MD) and its 95% confidence intervals (CIs). Dichotomous outcomes were reported as the risk ratio (RR) and 95% CIs. Statistical pooling was performed with a random-effect model, via generic inverse variance weighting. All the statistical analyses in this meta-analysis were performed using Review Manager 5 software (RevMan 2012) for the meta-analysis.

Hypothesis testing was set at the twotailed and results were considered statistically significant at 0.05 level. The I² statistic was calculated as a measure of statistical heterogeneity, and I2 values of 25%, 50%, and 75% corresponded to low, medium, and high levels of heterogeneity. When heterogeneity was found (I²>25%), sensitivity analysis was performed in an attempt to explain the findings. When doing a pool for some outcome assessment, we excluded the study which has the significant difference at baseline to keep two groups in all studies have the consistent outcome at the baseline. For each parameter estimate, an integrated analysis was given, finally.

The meta-analysis was performed in accordance with the recommendations by Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) workgroup ²⁰.

Study Characteristics

A total of 223 potentially relevant citations were identified and screened, of which 197 were selectively excluded from the study because they were not clinical RCTs or did not expose the outcome of interest. Twenty-six articles were retrieved for detailed evaluation. Overall, seven RCTs were included with a combined total of 632 patients ^{3,15,17,21-24} (Fig. 1).

Fig 1 Flow chart showing the number of citations retrieved by individual searches and the number of trials included in the review. 

The details of the characteristics and the demographic data of the RCTs included in our analysis were summarized in Table 1. These studies varied in sample size, and follow-up duration differed from 12 to 24 months, spanning nearly 10 years. The mean age of the populations ranged from 51~62 years and the mean of body mass index (BMI) ranged from 23~28.4 kg/m². The prevalence of diabetes in the patients was from 11%~56%. More than half of the patients in both groups used angiotensin converting-enzyme inhibitors (ACEI) or angiotensin II receptor blockers (ARB) and half of the patients in both groups used diuretics in two studies ^3,23. All trials evaluated the LS (Balance: Fresenius Medical Care) compared with a CS (Stay•Safe: Fresenius Medical Care). Almost all studies included incident CAPD patients except the Choi et al.²¹ study, and patients with CAPD modality except the balANZ Trial ³.

Table 1 Characteristics of the included RCTs in this analysis. 

Note: data are presented as mean or median (range). NA, not available. L/C, neutral pH and low-GDP PDSs/conventional PDSs; DM, diabetes mellitus; BMI, body mass index; ACEi, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker.

Baseline of outcomes in these included studies were shown in Table 2. Kim et al.²² demonstrated that there were no significant differences of all outcomes between the two groups except CrCl (LS group, 95.5±5.0 vs. CS group, 78.6±11.8 L/week/1.73m², p<0.05) and nPNA (LS group, 0.85±0.07 vs. CS group, 1.06±0.11 g/kg/day, p<0.05). The D/Pcr at the baseline was higher in the LS group than in the CS group in the two trials studied by Kim et al. ²³ and Park et al.¹⁷. Moreover in the study by Park et al. ¹⁷ peritoneal CrCl and was higher in the LS group, peritoneal UF volume was lower in the LS group at baseline in keeping with higher peritoneal transport characteristics in this group. Szeto et al.¹⁵ showed that at baseline, the CS group had a better nutritional status than the LS group (serum albumin, p=0.004 and SGA, p=0.023), but the difference disappeared in 12 months.

Table 2 The baseline of outcomes in the included RCTs. 

Note: data are presented as mean±SD or median (range). NA, not available. Bold indicates the parameters have significant differences between the two groups and asterisk (*) indicates p<0.05 versus CS group. RRF, mean of creatinine clearance (Ccr) and urea clearance (Curea); Kt/V, total weekly urea clearance; peritoneal Kt/V, weekly peritoneal urea clearance; CrCl, total creatinine clearance; peritoneal CCr, peritoneal creatinine clearance; peritoneal UF, peritoneal ultrafiltration; D/Pcr, dialyzate-to-plasma creatinine ratio at 4 hours of peritoneal equilibration test (PET); D/D0 glucose, D/D0 glucose at 4 hours of PET; SBP, systolic blood pressure; DBP, diastolic blood pressure; SGA, subjective global assessment; nPNA, normalized protein nitrogen appearance.

Quality Assessment

Two investigators assessed the quality of the included studies independently. All RCTs were considered fair to good quality (Fig. 2). Allocation methods and concealment were generally, incompletely reported and therefore difficult to assess. Allocation concealment was adequate in four studies (43%). Six studies (86%) were classified as low risk of performance bias and only one study was unclearly reported. However, no information about the blinding of outcome assessment (detection bias) of the studies was provided. Completeness of outcome reporting and intention-to-treat analysis methodology was applied in 29% of included studies. Selective reporting was observed in six studies (86%). No other significant biases were identified in these seven studies, except an unclear description of participant details in four studies. The Jadad score was 3 or higher (Table 1), even though the method of random sequence generation, blinding of participants and allocation concealment were not mentioned in most studies.

Fig 2 Risk of bias graph: each risk of bias item is presented as percentages across all included studies. 

Outcome Measurement

PD patients in these different studies were followed up for different periods, which may have influenced the effectiveness of the outcomes of this analysis. Therefore, subgroup analysis was used to decrease clinical heterogeneity according to the follow-up periods.

Residual Renal Function

Two studies ^17,23 of seven RCTs were undertaken to calculate the RRF of 226 patients after 6 months of follow-up, and indicated that LS group was beneficial for preserving RRF compared with the control group (MD 1.28 mL/min, 95% CI 0.52 to 2.03, p=0.0009; I²=0%). Similar results were obtained after 12 months of follow-up in all studies including 520 patients (MD 0.60 mL/min, 95% CI 0.18 to 1.02, p=0.005; I²=11%). The balANZ Trial ³ followed up 24 months and RRF was measured at baseline, 12 and 24 months, as well as the study by Bajo et al.²⁴, and the pooled data indicated no difference between the two groups (p=0.76). As the studies duration continued from 6 to 24 months, the difference of RRF between the two groups was reduced gradually. This should be commented in the abstract and/or conclusions. Considering the heterogeneity, exclusion of the study ²⁴ with a small sample size did not materially change the results of the meta-analysis or the subgroup analyses Overall, the use of LS induced a reduction in RRF decline compared with the control group (MD 0.66 mL/min, 95% CI 0.34 to 0.99; p<0.0001; I2=4%; Fig. 3).

Fig 3 Effect of low-GDP PD solution on RRF (mL/min). 

Daily Urine Volume

Three studies ^3,17,23 with a total of 377 patients and five studies ^{3,15,17 21,23} with a total of 462 patients showed the 24h urine volume separately at 6 and 12 months. The 24h urine volume in the LS group was higher than that in the CS group (MD 155.42 mL/d, 95% CI 37.84 to 273.00; p=0.01) at 6 months. A total of 238 patients were followed up in the LS groups and 224 patients were followed up in the CS groups after 1 year’s study. Patients with the LS had more daily urine volume than the CS group (MD 158.93 mL/d, 95% CI 83.22 to 234.64; p<0.0001). Only the balANZ Trial ³ reported the urine volume at 24 months follow-up, and there was no significant difference between the two groups. As the study duration continued from 12 to 24 months, the MD of the residual urine volume decreased from 158.93 mL/d to 115.00 mL/d. The pooled urine volume in patients using LS was greater than using CS (MD 153.15 mL/d, 95% CI 96.62 to 209.68; p<0.00001; I2=0%; Table 2). Overall, our meta-analysis indicated that the LS had a significant effect on RRF with an increase in daily urine output compared with the CS group.

Small solute clearance

At 6 months, Kim et al.²³ and Park et al.¹⁷ published the data of total Kt/V and peritoneal Kt/V showing that there was no statistical difference between the two groups (p=0.99; p=0.18). After one year follow up, five studies involving 360 patients reported the effect of LS on total Kt/V in PD patients ^15,17,21-23. Compared to the CS group, the LS group showed significantly increased Kt/V (MD 0.13, 95% CI 0.06 to 0.20; p=0.0002). Overall, we found that patients with LS had higher total Kt/V than with CS (MD 0.11, 95% CI 0.05 to 0.17; p=0.0007; I²=0%) (Fig. 4) and the CS group had a higher peritoneal Kt/V than the LS group (MD -0.10, 95% CI -0.20 to -0.01; p=0.03; I²=0%) (Table 2).

Fig. 4 Effect of low-GDP PD solution on total Kt/V. 

Our subgroup analyses showed no statistical differences of total CrCl and peritoneal CrCl between the LS and CS groups at 6 and 12 months (Table 2). We excluded the total CrCl data of the follow-up period from the study performed by Kim et al.²² who reported the significant difference between the two groups at baseline but no statistical difference observed at 12 months. The study by Park et al.¹⁷ was excluded because this study published that peritoneal CrCl was higher in the LS group at baseline and there was no significant difference after 6 months.

Peritoneal Ultrafiltration and Glucose Load

Five studies ^{3,15,17,21,23} published the daily peritoneal UF volume in the followup period. Park et al.¹⁷ indicated that the CS group had higher UF than the LS group at baseline and 6 months. After exclusion of this study, we pooled the data at 6 months, showing the higher UF in the CS group (MD -261.97 mL/d, 95% CI -427.73 to -96.21; p=0.002). In the subgroup analyses of 12 months, Choi et al.²¹ who included all prevalent PD patients with more than half number of anuric, revealed the outcome that UF was significantly higher in the LS group than in the CS group at all follow-up visits. The exclusion of this study did materially change the results of the meta-analysis or the subgroup analyses. Table 3 showed that patients with the LS had less daily peritoneal UF volume than the CS (MD -193.45 mL/d, 95% CI -315.36 to -71.54; p=0.002; I²=36%). The subgroup analyses of glucose load suggested that there was no statistically significant difference between patients using the LS and CS at 6 and 12 months.

Blood Pressure

The balANZ Trial ³ and the study by Park et al.¹⁷ followed up the blood pressure of the two groups. There was no significant difference between the two -groups in controlling blood pressure during 1 year of follow-up (SBP, p=0.91; DBP, p=0.59) (Table 3).

Peritoneal Solute Transport Rate

Five studies ^3,17,21-23 published the D/Pcr. In the study by Kim et al. ²³, the D/Pcr was higher in the LS group than in the CS group, and this difference persisted throughout the treatment period. Similar results were obtained from Park et al.¹⁷, but after 6 months, the D/Pcr showed no difference between the two groups. The patients of two groups in these three included studies ^3,21,22 had high average transport status. Overall, there was no statistically significant difference in the D/Pcr between the two groups (MD 0.00, 95% CI -0.02 to 0.02; p=1.00; I²=0%) (Table 3).

However, two studies ^22,23 with a difference on D/D0 glucose were small sample trials. The pooled analysis suggested that the CS had a higher D/D0 than the LS (MD -0.03, 95% CI -0.05 to -0.01; p=0.01; I²=17%) (Table 3).

Overall, the D/Pcr and D/D0 glucose of all patients included in this subgroup analysis indicated that both the LS group and the CS group had high-average transport characteristics of peritoneal membrane ⁶.

Nutritional Status

Our meta-analysis indicated that patients using the LS had lower serum albumin than the CS (MD -0.14 g/dL, 95% CI -0.23 to -0.05; p=0.002; I²=45%) (Table 3).

In the meta-analysis of five studies ^{3,15,17,21,23}, we found there was no significant difference in nPNA between the two groups (MD -0.02 g/kg/d, 95% CI -0.05 to 0.00; p=0.10;I²=0%) (Table 3).

Only two studies ^17,21, publishing the data of SGA were small sample trials. We found that the LS group had a better SGA score than the CS group (MD 0.33, 95% CI 0.08 to 0.57; p=0.009; I²=21%) (Table 3).

All-cause Mortality

All seven studies ^{3,15,17,21-24} published the effect of LS on patients’ survival. No patient died in the two groups at 6-month follow-up. At 12 months five studies ^15,17,21-23 involving 417 patients and at 24 months two studies involving 215 patients were included in the subgroup analysis, suggesting that there was no significant difference between the two groups, respectively (Table 3).