Differential Outcomes Between Dialysis Modalities

Cardiovascular morbidity and mortality among patients receiving lifesaving dialysis far exceeds that of the age-adjusted general population. Over the last several decades, there have been ongoing debates and numerous investigations seeking to identify the ideal dialysis modality for individual end-stage renal disease (ESRD) patients. Despite a paucity of evidence to support the use of one modality over another, the prevalence rates of peritoneal dialysis (PD) have declined in the United States to a current level of 7%.1 In the era of bundled dialysis payments, health care reform, and pay-for-performance, there is a heightened interest in identifying the most cost-effective interventions associated with the best clinical outcomes. Home dialysis modalities are cost-effective, but do they improve clinical outcomes? While the annual payer costs for patients treated with peritoneal dialysis are lower compared with hemodialysis, it remains unknown which modality is associated with the best clinical outcomes. Several recent studies have identified an early survival advantage associated with the use of PD versus hemodialysis (HD) among incident ESRD cohorts, while others have either failed to identify a survival difference or have identified poorer outcomes associated with PD versus HD.2–6 However, considering the known selection bias introduced by patients and physicians in choosing the appropriate dialysis modality, comparisons of outcomes between dialysis modalities inevitably is limited by unmeasured prognostic differences between groups at baseline. While a randomized controlled trial is necessary to solve the differential findings in observational studies, an initial attempt at a trial was not successful.7 Thus, observational studies with robust statistical methods are the only feasible alternative. In this issue of JASN, Perl et al. compare outcomes between PD and HD among a Canadian cohort of incident ESRD patients, with the HD patients stratified by dialysis access.8 The authors speculate that patients initiating HD with an arteriovenous fistula or graft (AVF/AVG) versus a catheter would be more similar to patients starting dialysis with a PD catheter, thus minimizing selection bias. They also hypothesize that the previously identified early survival advantage associated with PD versus HD would be attenuated in a comparison that controlled for vascular access. This study included 38,512 patients incident to dialysis between 2001 and 2008 who were registered in the Canadian Organ Replacement Register. Approximately 19% of patients started PD, and, of those who started on HD, only 21.4% initiated dialysis with an AVF or AVG. Similar to prior studies, PD patients in this cohort had lower overall comorbidity compared with HD patients, including a lower prevalence of diabetes mellitus, coronary artery disease, peripheral vascular disease, malignancy, and pulmonary disease. Overall, 1-yr adjusted mortality was higher with HD compared with PD; however, when patients were stratified by dialysis access type, the increased mortality was limited to HD patients with a catheter. Among patients who started HD with an AVF/AVG versus PD, 1-yr adjusted mortality was equivalent between HD and PD (hazard ratio 0.9, 95% CI 0.8 to 1.1). However, 5-yr adjusted mortality was lower among HD patients with an AVF/AVG relative to PD patients (HR 0.80, 95% CI 0.80 to 0.90). Among patients who started HD with a catheter versus PD, 1-yr and 5-yr adjusted mortality was higher among HD-catheter patients versus PD patients. However, after the first year, HD-catheter patients had similar mortality risk to PD patients. In further sensitivity analyses including the use of marginal structures models, which adjusted for propensity scores for selection of dialysis modality and probability of renal transplantation, the results were robust. In prespecified subgroup analyses, nearly all subgroups had improved 5-yr survival associated with HD with an AVG/AVF versus PD, while there was decreased survival among those with HD with a catheter versus PD. These findings provide new insights into outcomes associated with different dialysis modalities. Only patients who have been under the care of a nephrologist and have planned for dialysis initiation will have dialysis access in place (whether a PD catheter or AVG/AVF), which makes comparisons between these groups reasonable. Furthermore, the use of a catheter versus AVF/AVG is known to be associated with higher infectious complications and higher mortality, and may bias toward worse early outcomes with HD, as suggested by prior observational studies. Thus, the analysis by Perl et al. appears to be a reasonable approach to compare outcomes between dialysis modalities while minimizing the inherent selection bias between groups. However, cautious interpretation and application of these findings should be considered. This is a secondary analysis in which patients were allocated to one dialysis modality for a reason, whether it was patient or physician selection or other possibly unmeasured risk factors. While this study attempts to minimize these differences by minimizing pre-ESRD care as a confounder, residual unmeasured confounding likely exists. Despite over 50 years of renal replacement therapy being federally funded in the United States, the ideal way to deliver dialysis and the best modality to use for individual patients remains uncertain. However, in the era of bundled payments, it has been proposed that the prevalence rates of PD will rise, as it remains a lower-cost therapeutic option. This study by Perl et al. supports the individualized use of PD as an equivalent dialysis modality to HD and as a preferred modality if the alternative is to start hemodialysis with a catheter. DISCLOSURES JKI receives support from the NIH (K23 HL092297) and investigator initiated grant support from Genzyme. RDT receives support from the NIH (K24DK002818) and grant support from Novartis and Reata Pharma.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.