The objectives of this study were to identify predictors of survival on hemodialysis in patients with diabetic end-stage renal disease (ESRD) and to explain ethnic differences in survival among non-Hispanic whites, African-Americans, and Mexican-Americans. The study design was a survival analysis of an inception cohort and was conducted in dialysis centers in two urban counties in Texas. A population-based, tri-ethnic cohort of 638 adult patients with diabetic ESRD were studied. Follow-up was completed in 96% of the cohort, with a median length of follow-up of 3.8 years. Survival length on center hemodialysis was the main outcome measure. In a combined model of types I and II diabetes, Mexican-Americans (relative hazard [RH], 0.666; 95% confidence interval [CI], 0.457 to 0.944) and African-Americans (RH, 0.598; 95% CI, 0.414 to 0.864) showed a better survival than non-Hispanic whites. Other predictors independently associated with survival were age (RH, 1.015 per 10 years of age; 95% CI, 1.001 to 1.028), high self-reported physical disability (RH, 1.770; 95% CI, 1.213 to 2.583), coronary artery disease (RH, 1.445; 95% CI, 1.044 to 2.012), lower extremity amputations (RH, 2.049; 95% CI, 1.438 to 2.920), and average blood glucose levels prior to ESRD (RH, 1.002 per 1 mg/dL increment; 95% CI, 1.003 to 1.004). Non-Hispanic whites had a significantly higher rate of type I diabetes, but did not have a greater burden of any of the other predictors. In separate type I and II models, ethnicity was still a significant predictor of survival among type I but not among type II. In conclusion, we have reconfirmed the survival advantage on dialysis of African-Americans and Mexican-Americans over non-Hispanic whites with diabetic ESRD. However, among type II patients, this minority survival advantage disappears. Self-reported physical disability is an important predictor of survival among both diabetes types. Functional status at baseline is an important predictor of survival and should be assessed as an adjunct to measurement of co-morbidities. Macrovascular disease is important for type II, while educational status is important for type I. While amputation may be a marker for the severity of systemic illness, it could be a marker for quality of primary care provided to diabetic patients, since a majority of diabetic lower extremity amputations are thought to be preventable.