Comparing outcomes related to dialysis modality is complicated by selection bias introduced by patients and physicians. To address the impact of selection bias, this study compared mortality by initial dialysis modality among patients who had ESRD and were placed on the transplant waiting list. This study was a historical prospective cohort of 12,568 patients in the United States who initiated dialysis between May 1, 1995, and October 31, 1998, and were placed on the transplant waiting list before dialysis initiation. Two-year mortality was compared using Kaplan-Meier curves and Cox proportional hazards models that analyzed patients primarily using an intention-to-treat approach and separately censored patients on a modality switch. At 2 yr, the unadjusted mortality rate was 6.6% among peritoneal dialysis (PD) patients compared with 6.9% among hemodialysis (HD) patients (hazard ratio [HR] 1.01; 95% confidence interval [CI] 0.82 to 1.23). After controlling for differences in baseline characteristics, comorbidities, and laboratory variables, the selection of PD versus HD remained associated with a similar 2-yr mortality risk (HR 1.03; 95% CI 0.83 to 1.28). In separate models, 2-yr mortality associated with PD versus HD was significant among patients with body mass index (BMI) > or = 26 kg/m2 (HR 1.37; 95% CI 1.01 to 1.83) but not among patients with BMI < 26 kg/m2 (HR 0.81; 95% CI 0.61 to 1.07). Results were similar after censoring on a modality switch. In conclusion, although choice of initial dialysis modality seems to be associated with equivalent outcomes among patients who have ESRD and are placed on the transplant waiting list, patients with BMI > or = 26 kg/m2 have increased 2-yr mortality associated with the selection of PD versus HD. Because the interpretation of observational data is highly affected by residual confounding and selection bias, further efforts should focus on the formation and testing of hypotheses to improve dialysis delivery.