Quantity of Research Available
A total of 587 citations were identified in the literature search. Following screening of titles and abstracts, 562 citations were excluded and 25 potentially relevant reports from the electronic search were retrieved for full-text review. Seven potentially relevant publications were retrieved from the grey literature search. Of the 32 potentially relevant articles, 28 publications were excluded for various reasons, while four publications met the inclusion criteria and were included in this report. Appendix 1 describes the PRISMA flowchart of the study selection.
Summary of Study Characteristics
Study Design
All of the included studies2,9–11 had retrospective observational designs. Two studies2,9 used propensity-matched analyses to correct for case-mix differences between patients who initiated peritoneal dialysis (PD) and hemodialysis (HD) as first renal replacement therapy. Another study10 used marginal structural models to analyze outcomes in a cohort of patients who initiated dialysis using either the PD or HD modality. One single-center study11 with a non-matched control group, in which univariate and multivariate regression were used for analysis.
Country of Origin
One study each was conducted in Norway2 and Germany,11 while two studies9,10 were conducted in the United States of America (USA). The Norwegian study2 and one study from the USA9 used national databases of their respective countries. Another study from the USA10 used data from a healthcare organization with a kidney care division operating a large number of facilities (2,225 outpatient facilities as at September 2015) across the USA. The German study11 analyzed dialysis registry data of patients from a single nephrology centre. Both univariate and multiple logistic regression analyses were used to assess the impact of potential confounder on results.
Patient Population
One study2 analyzed data from 692 matched pairs of adult end-stage renal disease (ESRD) patients who initiated dialysis as first renal replacement therapy from 2005 to 2012. The mean age per modality was 64.6 years and 65.2 years for PD and HD, respectively. The mean body mass index (BMI) was 25.4 kg/m2 for PD and 25.5 kg/m2 for HD, while the mean estimated glomerular filtration rates (eGFR) were 8.8 ml/min/1.73 m2 and 9.0 ml/min/1.73 m2 for the PD and HD modalities, respectively. Another study9 analyzed data from 1003 matched pairs of ESRD patients who initiated PD or HD as their first dialysis modality between 1 January 2001 and 30 June 2013. The mean age per modality was 57.4 years for PD, and 58.4 years for HD. The study did not report BMI and eGFR data. One study10 analyzed data from 23,718 incident dialysis patients who initiated dialysis from July of 2001 to June of 2004. Depending on whether patients remained on their initial modality, switched to another modality, or received transplantation, the mean age in the PD group ranged from 47 years to 62 years, while the mean age in the HD group ranged from 44 years to 64 years. By the same designation, mean BMI ranged from 25.5 to 27.8 kg/m2 in the PD group and 25.1 to 26.3 kg/m2 in the HD group. Data on eGRF not reported in this study.10 Another study11 involved patients who started PD or HD in an urgent fashion due to late referral or unexpected deterioration of residual renal function with manifestation of uraemic syndrome or over-hydration. The patients were mainly elderly with mean ages of 72.6 years and 74.1 years for the PD and HD modalities, respectively. Most of the patients (PD, 95%; HD, 89%) were already hospitalized before initiation of dialysis. The study did not report BMI and eGFR data.11
Interventions and Comparators
In the studies that applied the propensity-matched model2,9 PD patients were matched in a 1:1 fashion with HD patients, creating 692 pairs of patients with comparable baseline variables in one study,2 and 1003 matched pair in the other study.9 In one study,2 the median follow-up time was 13.0 months in the PD group and 10.0 months in the HD group, while another study9 had a nine year follow-up period for each modality. In another study10 1,358 patients initiated dialysis using PD, while 22,360 initiated dialysis using HD, with a 24 months follow-up period for each modality. In one study,11 patients received acute unplanned therapy with either PD or HD. Unplanned dialysis initiation was defined as beginning dialysis urgently due to late referral or unexpected deterioration of residual renal function with uraemic syndrome or over-hydration in patients without functional fistula for dialysis and therefore needing a central venous dialysis catheter11. The duration of follow-up was six months.
Outcomes
Survival was the outcome measure of interest in all the included studies.2,9–11 In two studies2,11 both cardiovascular and all-cause mortality were reported. In one study2 cardiovascular cause of death was defined in accordance with the European Renal Association–European Dialysis and Transplant Association (ERA-EDTA) cause of dead (COD) codes as a composite of COD group I–IV including myocardial ischemia and infarction, heart failure, cardiac arrest/sudden death and cerebrovascular accident. The other study11 did not specify what constituted cardiovascular mortality. This study11 also reported the risk of infections and the mortality due to infections for PD and HD. In two studies,2,9 survival analyses were conducted both in terms of the as-treated and the intention-to-treat (ITT) populations. In the as-treated analysis, survival was analyzed on the basis of the modality a patient was on at the time of death, regardless of whether the patient began on a different modality. The ITT analysis attributed death to the initial dialysis modality of the patient regardless of whether the patient switched modalities over the course of the study. In both methods of analysis, patients were censored for the earliest of the following: renal transplantation, dialysis cessation, death, renal recovery, loss to follow-up or study end. In addition, the as-treated analyses censored patients for change in dialysis modality. One study10 reported modality change, differential transplantation rates, and detailed time-varying laboratory measurements in addition to survival outcomes.
Summary of Critical Appraisal
Two studies2,9 used propensity-matched designs and matching patients in terms of baseline disease burden and demographics to adjust for the case-mix differences between HD and PD patients to resolve the bias due to confounding by indication associated with regular observational studies. They each matched patients on a larger number of variables to ensure improved reliability of the study results. Furthermore, each study2,12 analyzed survival outcomes on the basis of both ITT and as-treated population, with both methods yielding similar results to indicate rigorous findings of the studies. Three studies2,9,10 analyzed data from large, diverse cohorts of patients, thus increasing the potential for their respective study populations to be representative of the general renal failure population requiring renal replacement therapy (RRT), while one study11 retrospectively analyzed data from a nephrology center. The authors of three studies2,9,11 declared no competing interests. In one study,10 an author had received grant support and/or honoraria from a firm that deals in products for renal care including dialysis. However, this is unlikely to bias the outcomes since the company caters for both PD and HD needs.
All of the included studies2,9–11 were nonrandomized nature, making them susceptible to uncontrolled confounding, measurement errors, and selection bias. Thus, despite improvement over traditional observational studies to adjust for confounding by indication, propensity matched analyses used by two studies2,9 are unable to properly correct for non-measured potentially confounding variables such as pre-dialysis care, residual renal function, and the kind of physiological dialysis solutions used. For the study that utilized marginal structural models,10 liability to these limitations and confounding by indication cannot be ruled out. One study11 had a non-matched control group. Therefore, the potential for biased study results due to imbalances and confounders was high.
The propensity-matched approach, require that only HD patients with similar baseline characteristics as PD patients were included in the study. Therefore, it is unknown if the results from these two studies2,9 will be generalizable in all patients who require RRT with dialysis. One study9 included patients who utilized either an arteriovenous fistula or a graft during the first 90 days of study, excluding those who utilized a central venous catheter as vascular access at any time during the first 90 days of dialysis, including patients who initiated dialysis urgently with a catheter. The motivation for the exclusion criterion was to reduce case-mix bias. However, the potential of excluding “crash dialysis patients” from this study9 cannot be overlooked. In one study,11 there were significant differences between PD and HD patients at baseline with respect to heart failure (NYHA Stage III–IV), serum creatinine and glomerular filtration rate, and prevalence of malignancy, which could influence the results. Although regression analyses found that these had no significant impact on overall mortality, the possibility that the study was not sufficiently powered to make such detection cannot be ruled out. The study population consisted mainly of elderly patients, and multivariate analysis showed that age at initiation of dialysis was significantly associated with overall mortality risk.11 Therefore, it is uncertain whether the reported findings will be reproducible in a younger population. All the studies were conducted in countries other than Canada, with one study2 solely analyzing patient data from the Norwegian Renal Registry, while two9,10 used analyzed data from the USA, and another study11 analyzed data from a single nephrology center in Germany.11 Therefore, the generalizability of the reported findings from these studies to the Canadian context is unknown.
One study9 had a nine-year follow-up period while two other studies had five-year2 and two-year10 follow-up periods, and another study11 had a six-months of follow-up. It is unknown if these follow-up durations were long enough to predict late complications of any initial dialysis modality.
Summary of Findings
What is the comparative clinical effectiveness of peritoneal dialysis versus hemodialysis for the “first initiative” treatment of renal failure in adult patients?
Survival
One study2 found that the initial dialysis modality did not affect all-cause mortality as determined by as-treated analyses, or ITT analyses. The two year hazard ratios (HR) and 95% confidence interval (CI) for the two modalities (PD versus HD) were 0.87 (0.67 to 1.12) for as-treated and 0.93 (0.73 to 1.18) for ITT analyses. Similar results were observed in analyses performed at 5 years with HR (95 % CI) of 0.95 (0.77 to 1.17) for as-treated and 0.99 (0.82 to 1.19) for ITT analyses. Another study11 found that after six months of therapy, all-cause mortality among incident PD and HD patients with acutely initiated unplanned PD or HD was 30.3% and 42.1% respectively. The difference was not statistically significant (P = 0.19). Among PD patients, the proportion of mortality attributed to unknown causes 30% compared to 16.7% among HD patients. The study11 also found that there was no statistically significant difference (P = 1.00) in cardiovascular mortality at the end of six months between incident PD and HD patients with acutely initiated unplanned PD or HD. It is unknown whether the smaller sample size (n=123) with a relatively small incidence of cardiovascular death (six in each group) or the short duration of the study contributed to these observations.
However, two studies9,10 found that initiating dialysis with PD modality was associated with improved survival outcomes compared with HD. In one study,9 both the as-treated analyses and ITT analyses at one year resulted in more than twice survival advantage in the PD group compared with HD with the cumulative hazard ratio (CHR) for death (HD versus PD) of 2.38 (95% CI: 1.68 to 3.40; P<0.0001) for the as-treated analysis and 2.10 (95% CI: 1.50 to 2.94; P<0.0001) for the ITT analysis. At 2 years the CHR was 1.39 (95% CI: 1.06 to 1.82; P = 0.017) in the as-treated analysis and 1.26 (95% CI: 0.98 to 1.63; P = 0.070) in the ITT analysis. A trend of cumulative risk of death favoring PD was observed for nearly 3 years of follow-up using as-treated analysis and for 2 years using ITT analysis, with no significant differences in outcomes between the modalities thereafter. Among patients initiated on PD modality, as treated analysis found that adjusted survival rates were 95% and 87%, for years 1 and 2, respectively compared with 89% and 83% among patients who initiated HD in the same time period. Similar results were reported from ITT analysis with survival rates of 94% and 86% for PD patients compared with 89% and 83% for HD patients, for years 1 and 2 respectively. Another study10 found that PD was associated with persistently greater survival compared with HD. Patients who initiated dialysis using the PD modality had 48% lower mortality (i.e., a death hazard ratio [HR] of 0.52; 95% CI: 0.34 to 0.80).
Subgroup outcomes
One study2 stratified analysis by age and reported that for both two- and five-year follow-up PD was associated with reduced all-cause mortality in patients aged below 65 years. This was not observed in patients older than 65 years. At year-two, the as–treated analysis found that the HR (PD versus HD) was 0.39 (95 % CI: 0.19 to 0.81; P = 0.011 for interaction; and HR (95% CI) for the ITT analysis was 0.47 (0.26 to 0.85; P = 0.009). Corresponding HR (95% CI) results for analysis at year-five were 0.49 (0.27 to 0.89; P = 0.010) for as–treated analysis and 0.58 (0.36 to 0.93; P = 0.009) for ITT analysis. The study2 also found that cardiovascular disease was the cause of death in 49.1 % at year-two, and 48.2 % at year-five as-treated follow-up, with patients younger than 65 years who had initial PD modality associated with lower cardiovascular mortality compared with those who had initial HD modality (HR 0.38, 95 % CI 0.15 to 0.96).
There were no interactions between gender or diabetes mellitus and initial dialysis modality found concerning mortality. PD was not associated with impaired prognosis in any of the pre-specified subgroups compared to HD. However, multivariate analysis in one study11 showed that age at initiation of dialysis was significantly associated with overall mortality risk.
Other outcomes
One study10 found that 6% of those with initial HD compared with 57% of those with initial PD switched modality at least once during their first two years of dialysis, and 6% of the initial PD patients compared with 18% of the initial HD had renal transplantation within the same period. Another study11 reported that the incidence of bacteremia in incident PD and HD patients treated with acutely initiated unplanned PD or HD was significantly higher in the HD group compared with PD group over the course of the six months (21.1% versus 3.0%, respectively, P < 0.01); with the RR (95% CI) for bacteremia (HD vs. PD) estimated to be 0.16 (0.05 to 0.57; P = 0.005). However, the overall incidence of mortality due to infection was not significant different between the two groups (17.5% versus 9.1%, respectively, P = 0.19).
What are the evidence-based guidelines regarding optimal first line dialysis for the treatment of renal failure in adult patients?
The literature search for this review did not find any guidelines regarding optimal first line dialysis modality for the treatment of renal failure in adult patients.
What are the evidence-based guidelines regarding optimal peritoneal dialysis care for the treatment of renal failure in adult patients?
The literature search for this review did not find any guidelines regarding optimal peritoneal dialysis care for the treatment of renal failure in adult patients.
Limitations
One limitation is that three included studies2,9,10 retrospectively analyzed data from ESRD patients from the time they initiated dialysis for the first time, and the proportion of such patients who were given “crash start” dialysis is unknown. For the purpose of this review, “crash start” refers to the situation where the dialysis modality is applied in patients who have renal failure but are naïve to any dialysis (see Table 1). Although one study11 investigated unplanned acute dialysis which must have involved crash start patients, it was a single-center study with uncertainty about sufficient power to detect differences in outcomes between the treatment groups; and because the patients were elderly, it is unknown whether the study results will reproducible in a younger population. In addition, the non-randomized nature of the included studies2,9,10 makes them liable to biases not accounted for by the various analytical methods. For instance none of the two propensity-matched analysis2,9 adjusted for potentially confounding variables such as pre-dialysis care, and the kind of physiological dialysis solutions that was used for the selected modality; and one study9 could not adjust for confounding due to residual renal function and key baseline laboratory values because data regarding these potential confounders were not available at the time dialysis was initiated. In the marginal structural model study10 uncontrolled confounding (especially confounding by indication), measurement errors, and selection bias were limitations. The potential for such confounders to skew the reported findings in favor of one initial dialysis modality cannot be excluded. The potential for bias due to imbalances and confounders was high for the study11 in which incident patients had acutely initiated unplanned PD or HD therapy because it was an observational and retrospective study with non-matched control group.
