The relationship between laboratory-based outcome measures and mortality in end-stage renal disease: a systematic review.

Despite data that traditional laboratory-based outcome measures in dialysis are improving over time, population-based data indicate that mortality rates are not improving in parallel. With increased focus on performance measures based on laboratory-based outcomes (e.g., hematocrit, albumin, and parathyroid hormone), less emphasis has been placed on other markers, some of which may be stronger predictors of mortality. We performed a systematic review to interpret the predictive value of laboratory-based outcome measures in dialysis. We identified studies with data regarding the predictive value of laboratory-based outcomes for mortality in dialysis. We calculated the sample size-weighted pooled relative risk of death with dichotomized "high" vs. "low" levels of each measure. We rank-ordered predictors by scaling the pooled relative risk of each measure by its pooled standard deviation. There were 5171 titles, of which 128 (representing 44 laboratory-based outcomes) were selected. Nine were significantly associated with mortality, in order of decreasing scaled effect size: (1) tumor necrosis factor-alpha, (2) hematocrit, (3) interleukin-6, (4) troponin T, (5) Kt/V(urea), (6) prealbumin, (7) urea reduction ratio, (8) serum albumin, and (9) C-reactive protein. Other oft-cited measures such as calcium phosphate product and parathyroid hormone were not significantly associated with mortality in pooled analysis. Quality improvement efforts to improve traditional laboratory-based outcomes in end-stage renal disease are necessary, but likely insufficient, to improve overall mortality in dialysis. Renewed consideration of cardiovascular, inflammatory, and nutritional markers that are especially strong predictors of mortality may have important implications for risk stratification and targeted therapeutic interventions.