Mortality in end-stage renal disease: a reassessment of differences between patients treated with hemodialysis and peritoneal dialysis.

Recent registry studies comparing mortality between peritoneal dialysis (PD) and hemodialysis (HD) patients show conflicting results. The purpose of this study is to determine whether previously published results showing higher mortality for patients treated with PD versus HD in the United States continue to hold true over the period 1987-1993. National mortality rates for PD and HD were extracted from the U.S. Renal Data System (USRDS) annual reports for the cohort periods: 1987-1989, 1988-1990, 1989-1991, 1990-1992, and 1991-1993. Using Poisson regression, death rates per 100 patient years were compared between PD and HD for each cohort period controlling for age, gender, race, and cause of end-stage renal disease (diabetes versus nondiabetes). When incident patients and patients with a prior transplant were included in the analysis, starting with the 1989-1991 cohort, we found little or no difference in the relative risk (RR PD:HD) of death between PD and HD (1987-1989: RR = 1.17, P < 0.001; 1988-1990: RR = 1.12, P < 0.001; 1989-1991: RR = 1.06, P = NS; 1990-1992: RR = 1.06, P = NS; 1991-1993: RR = 1.08, P = 0.043). After a test for goodness-of-fit, separate analyses for diabetic patients and nondiabetic patients were done to examine unexplained variation in death rates. For nondiabetic patients, there was less than a 1% difference in the adjusted 1-yr survival between PD and HD from 1989-1993 (1989-1991: RR = 1.05, P = NS; 1990-1992: RR = 1.04, P = NS; 1991-1993: RR = 1.07, P < 0.01). Among diabetic patients, the PD:HD death rate ratio varied significantly according to gender and age. For the average male diabetic patient, there was little or no difference in risk between PD and HD from 1989-1993 (1989-1991: RR = 1.02, P = NS; 1990-1992: RR = 1.05, P = NS; 1991-1993: RR = 1.08, P < 0.01). For diabetic patients under the age of 50, those treated with PD had a significantly lower risk of death than those treated with HD (1989-1993: 0.84 < or = RR < or = 0.89, P < 0.005). Over the same period, female diabetic patients treated with PD had a higher risk, on average, than HD (1.18 < or = RR < or = 1.19, P < 0.001) as did diabetic patients over the age 50 (1.28 < or = RR < or = 1.30, P < 0.001). Unlike previously published results that were restricted to prevalent-only patients, this national study of both prevalent and incident patients found little or no difference in overall mortality between PD and HD. The recent trends in mortality likely reflect the inclusion of incident patients, but they may also reflect changes in case-mix differences and/or improved PD practice. Additional incident-based studies that allow for additional case-mix adjustments are needed to better compare outcomes between HD and PD.