A number of studies have found a relationship of lower all-cause mortality risk for ESRD patients treated with increasing dose of dialysis. The objective of this study was to determine the relationship of delivered dose of dialysis with cause-specific mortality. Data from the USRDS Case Mix Adequacy Study, which includes a national random sample of hemodialysis patients, were utilized. To minimize the contribution of unmeasured residual renal function, the sample used in this analysis (N = 2479) included only patients on dialysis for one year or more. Cox proportional hazards models, stratified for diabetes, were used to analyze the effect of delivered dose of dialysis (measured and reported by both Kt/V and URR) on major causes of death and withdrawal from dialysis, adjusting for other covariates including demographics, comorbid diseases present at start of study, functional status, laboratory values and other dialysis parameters. Patient follow-up for mortality was censored at the earliest of time of transplantation, 60 days after a switch to peritoneal dialysis or at the time of data abstraction. For each 0.1 higher Kt/V, the adjusted relative risk of death due to coronary artery disease was 9% lower (RR = 0.91, P < 0.05), due to other cardiac causes was 12% lower (RR = 0.88, P < 0.01), due to cerebrovascular disease (CVD) was 14% lower (RR = 0.86, P < 0.05), due to infection was 9% lower (RR = 0.91, P = 0.05), and due to other known causes was 6% lower (RR = 0.94, P < 0.05). There was no statistically significant relationship of Kt/V and risk of death among patients who died of malignancy (RR = 0.84, P = 0.10) or among patients whose death cause was missing (RR = 0.95, P = 0.41). The risk of withdrawal from dialysis prior to death due to any cause was 9% lower (RR = 0.91, P < 0.05) for each 0.1 higher Kt/V. The relationships of delivered dose of dialysis, as measured by URR, and cause-specific mortality were essentially similar in relative magnitude and statistical significance as the relationships observed using Kt/V as the measurement of dialysis dose, with the exception that the relationship was less significant for cerebrovascular disease and withdrawal from dialysis. The relationship of dialysis dose with risk of death due to each cause of death category except other cardiac causes and "other" causes appeared to be of greater magnitude and of greater statistical significance among diabetics than non-diabetics. These results indicate that low dose of dialysis is not associated with mortality due to just one isolated cause of death, but rather is due to a number of the major causes of death in this population. This study is consistent with hypotheses that low doses of dialysis may promote atherogenesis, infection, malnutrition and failure to thrive through a variety of pathophysiologic mechanisms. Further study is necessary to confirm these results and to test hypotheses that are developed.