Relationship between dialysis modality and mortality

Abstract

Mortality differences between peritoneal dialysis (PD) and hemodialysis (HD) are widely debated. In this study, mortality was compared between patients treated with PD and HD (including home HD) using data from 27,015 patients in the Australia and New Zealand Dialysis and Transplant Registry, 25,287 of whom were still receiving PD or HD 90 d after entry into the registry. Overall mortality rates were significantly lower during the 90- to 365-d period among those being treated with PD at day 90 (adjusted hazard ratio [HR] 0.89; 95% confidence interval [CI] 0.81 to 0.99]; P < 0.001). This effect, however, varied in direction and size with the presence of comorbidities: Younger patients without comorbidities had a mortality advantage with PD treatment, but other groups did not. After 12 mo, the use of PD at day 90 was associated with significantly increased mortality (adjusted HR 1.33; 95% CI 1.24 to 1.42; P < 0.001). In a supplementary as-treated analysis, PD treatment was associated with lower mortality during the first 90 d (adjusted HR 0.67; 95% CI 0.56 to 0.81; P < 0.001). These data suggest that the effect of dialysis modality on survival for an individual depends on time, age, and presence of comorbidities. Treatment with PD may be advantageous initially but may be associated with higher mortality after 12 mo.

Figure 1.

Kaplan-Meier survival curves, censored at transplantation, by RRT modality at 90 d (left) and as-treated modality, from 90 d. In the as-treated analyses, patients moved between groups at the date of treatment modality change. We then calculated the survivor function using the risk group at the point of each failure event.

Figure 2.

HR for mortality during PD versus during HD, by modality at 90 d, stratified by BMI at RRT start. ▴, Risk (95% CI) for death within 12 mo of RRT start; •, risk for death ≥12 mo after dialysis start. Also shown is the proportion of the cohort in each BMI category.

Figure 3.

HR for PD versus HD by year of starting dialysis. ▴, Risk (95% CI) for death within 12 mo of RRT start; •, risk for death ≥12 mo after dialysis start. The numbers above the x axis indicate the number in each subcohort.

Figure 4.

Risk of PD compared with HD stratified by age and the presence of any comorbidity. ▴, Risk (95% CI) for death within 12 mo of RRT start; •, risk for death ≥12 mo after dialysis start. There is statistically significant variation in the risk for death in the 90- to 365-d group (P = 0.005 for interaction) but not for the >1-yr group (P = 0.5 for interaction).

Figure 5.

Kaplan-Meier graph by modality of treatment at 90 d, by presence of any comorbidity and age group at RRT start. ▴, Risk (95% CI) for death within 12 mo of RRT start; •, risk for death ≥12 mo after dialysis start. Graphs are adjusted for vintage (to year 2000) and gender (to equal male and female numbers). Note that the y axes differ between panels.

Figure 6.

HR for mortality for patients treated with PD compared with HD, treating modality as a time-dependent covariate (i.e., analyses of modality as treated). HR are adjusted for all measured covariates and derived from shared frailty Cox model.

Figure 7.

HR for mortality associated with PD treatment compared with HD, by quartile of predicted probability of PD treatment at 90 d. ▴, Risk (95% CI) for death within 12 mo of RRT start; •, risk for death ≥12 mo after dialysis start. The bars indicate the observed proportion of people who received PD at 90 d in each quartile.