In the randomized Hemodialysis (HEMO) Study, chronic high-flux dialysis, as defined by higher beta-2 microglobulin (beta(2)M) clearance, compared with low-flux dialysis did not significantly alter all-cause mortality in the entire cohort but was associated with lower mortality in long-term dialysis patients. This analysis examined the determinants of serum beta(2)M levels and the associations of serum beta(2)M levels or dialyzer beta(2)M clearance with mortality. In a multivariable regression model that examined 1704 patients, baseline residual kidney urea clearance and dialyzer beta(2)M clearance were strong predictors of predialysis serum beta(2)M levels at 1 mo of follow-up, with regression coefficients of -7.21 (+/-0.69 SE) mg/L per ml/min per 35 L urea volume (P < 0.0001) and -1.94 (+/-0.30) mg/L per ml/min (P < 0.0001),respectively. In addition, black race and baseline years on dialysis correlated positively whereas age, diabetes, serum albumin, and body mass index correlated negatively with serum beta(2)M levels (P < 0.05). In time-dependent Cox regression models, mean cumulative predialysis serum beta(2)M levels but not dialyzer beta(2)M clearance were associated with all-cause mortality (relative risk = 1.11 per 10-mg/L increase in beta(2)M level; 95% confidence interval 1.05 to 1.19; P = 0.001), after adjustment for residual kidney urea clearance and number of prestudy years on dialysis. This association is supportive of the potential value of beta(2)M as a marker to guide chronic hemodialysis therapy.