PREDICTORS OF MORTALITY IN HEMODIALYSIS-PATIENTS

Serum biochemical measures suggestive of undernutrition have been reported to correlate with 1-yr mortality risk in prevalent groups of hemodialysis patients. The predictive power of these variables has not been reported in newly diagnosed patients or in patients whose dialysis prescription is guided by urea kinetics. The relationship of these predictors to mortality over periods of longer than 1 yr is also unreported. Therefore, the survival of 184 hemodialysis patients was examined for up to 44 months (1987 to 1991) with the Cox proportional hazards model. Baseline demographic, clinical, and biochemical parameters were used as independent variables. To adjust for bias in patient selection, the survival of patients with 12 months or less of prior dialysis at the time of enrollment (-new cases-) was analyzed separately from that of patients with more than 1 yr of prior treatment (''long-standing cases''). Serum albumin was less than 3.5 g/dL in 31% of new cases and in 12% of long-standing cases. Adjusting for the other variables, low serum albumin was the strongest mortality risk predictor in both new and long-standing cases. Low serum cholesterol was an independent risk predictor in both groups. Diabetes and race were not significant predictors. Mean age at enrollment was nearly a decade higher for nonsurvivors than for survivors, in both new and long-standing groups. Yet, age was not an independent risk predictor in the Cox model for the new group because of an unexpectedly high death rate among young black men. Female gender, which was confounded by increased age, took the place of age in the model for the new group. For each model, there was good agreement between observed and predicted mortality for up to 24 months. To assess the influence of dialysis treatment time and dose (measured as pre-to-post treatment urea ratio) on risk, survival was examined in a subset of 139 patients monitored for up to 22 months, from 1989 to 1991, a period when the urea ratio was used routinely. Adjusting for the other variables, low serum albumin and cholesterol again independently increased risk. The urea ratio was also a significant independent predictor. The pattern of mortality by urea ratio was U shaped, with minimum risk for values between 2.5 and 3.4. Treatment time did not influence risk. It was concluded that baseline serum values of albumin and cholesterol strongly influence survival for up to 2 yr in new and long-standing hemodialysis patients. Because albumin and cholesterol are indices of visceral protein status, these findings suggest that visceral protein depletion is a major risk factor for mortality in hemodialysis patients, even when they are adequately dialyzed.