Dialysis patients exhibit an inverse, L- or U-shaped association between blood pressure and mortality risk, in contrast to the linear association in the general population. We prospectively studied 9333 hemodialysis patients in France, aiming to analyze associations between predialysis systolic, diastolic, and pulse pressure with all-cause mortality, cardiovascular mortality, and nonfatal cardiovascular endpoints for a median follow-up of 548 days. Blood pressure components were tested against outcomes in time-varying covariate linear and fractional polynomial Cox models. Changes throughout follow-up were analyzed with a joint model including both the time-varying covariate of sequential blood pressure and its slope over time. A U-shaped association of systolic blood pressure was found with all-cause mortality and of both systolic and diastolic blood pressure with cardiovascular mortality. There was an L-shaped association of diastolic blood pressure with all-cause mortality. The lowest hazard ratio of all-cause mortality was observed for a systolic blood pressure of 165 mm Hg, and of cardiovascular mortality for systolic/diastolic pressures of 157/90 mm Hg, substantially higher than currently recommended values for the general population. The 95% lower confidence interval was approximately 135/70 mm Hg. We found no significant correlation for either systolic, diastolic, or pulse pressure with myocardial infarction or nontraumatic amputations, but there were significant positive associations between systolic and pulse pressure with stroke (per 10-mm Hg increase: hazard ratios 1.15, 95% confidence interval 1.07 and 1.23; and 1.20, 1.11 and 1.31, respectively). Thus, whereas high pre-dialysis blood pressure is associated with stroke risk, low pre-dialysis blood pressure may be both harmful and a proxy for comorbid conditions leading to premature death.