Studies of outcomes associated with dialysis therapies have yielded conflicting results. Bloembergen et al showed that prevalent patients on continuous ambulatory peritoneal dialysis (CAPD) or continuous cycling peritoneal dialysis (CCPD) had a 19% higher mortality risk than hemodialysis patients, and Fenton et al, analyzing Canadian incident patients, found a 27% lower risk. Attempting to reconcile these differences, we evaluated incident Medicare patients (99,048 on hemodialysis, 18,110 on CAPD/CCPD) from 1994 through 1996, following up to June 30, 1997. Patients were followed to transplantation, death, loss to follow-up, 60 days after modality change, or end of the study period. For each 3-month survival period, we used an interval Poisson regression to compare death rates, adjusting for age, gender, race, and primary renal diagnosis. A Cox regression was used to evaluate cause-specific mortality, and proportionality was addressed in both regressions by separating diabetic and nondiabetic patients. The Poisson regressions showed CAPD/CCPD to have outcomes comparable with or significantly better than hemodialysis, although results varied over time. The Cox regression found a lower mortality risk in nondiabetic CAPD/CCPD patients (women younger than 55 years: risk ratio [RR] = 0. 61; Cl, 0.59 to 0.66; women age 55 years or older: RR = 0.87; Cl, 0. 84 to 0.91; men younger than 55 years: RR = 0.72; Cl, 0.67 to 0.77; men age 55 years or older: RR = 0.87; Cl, 0.83 to 0.92) and in diabetic CAPD/CCPD patients younger than 55 (women: RR = 0.88; Cl, 0. 82 to 0.94; men: RR = 0.86; Cl, 0.81 to 0.92). The risk of all-cause death for female diabetics 55 years of age and older, in contrast, was 1.21 (Cl, 1.17 to 1.24) for CAPD/CCPD, and in cause-specific analyses, these patients had a significantly higher risk of infectious death. We conclude that, overall, within the first 2 years of therapy, short-term CAPD/CCPD appears to be associated with superior outcomes compared with hemodialysis. It also appears that patients on the two therapies have different mortality patterns over time, a nonproportionality that makes survival analyses vulnerable to the length of follow-up. Further investigation is needed to evaluate both the potential explanations for these findings and the use of more advanced statistical methods in the analysis of mortality rates associated with these dialytic therapies.