Clinical utility of different estimates of renal function for predicting mortality in chronic heart failure

BackgroundThe clinical utility of different renal function (RF) measures for risk stratification in chronic heart failure (CHF) and the incremental discriminative value of renal dysfunction have not been investigated thoroughly.MethodsWe studied 802 patients with systolic CHF. The primary outcome was all-cause mortality. The association of candidate variables and RF measures [serum creatinine (SCr), serum urea nitrogen (SUN), estimated creatinine clearance adjusted for body-surface area (eCrClBSA), and estimated glomerular filtration rate (eGFR)] with mortality was evaluated using Cox proportional-hazards analyses. Recommended metrics of goodness-of-fit and discrimination were calculated.ResultsAt follow-up (median: 1269 days), there were 301 deaths. Age (p < 0.001), ischemic etiology (p = 0.009), NYHA class (p < 0.001), anemia (p < 0.001), and left ventricular ejection fraction (p < 0.001) independently predicted mortality (reference risk model). On multivariable analysis incorporating one of the measures of RF at a time, each had an independent value for predicting mortality (p < 0.001). The addition of each RF measures to the reference model significantly increased the likelihood-ratio χ2 and the models incorporating eCrClBSA or SUN demonstrated the highest probability of being the best. Although changes in C statistic and net reclassification were not significant, the Integrated Discrimination Index was significantly improved by the addition of eCrClBSA. Calibration was improved by all measures of RF expect SUN. The model incorporating eCrClBSA demonstrated both the best goodness-of-fit and discrimination.ConclusionsOur data suggest that renal dysfunction significantly improves risk stratification in a context of established risk factors. eCrClBSA appears to be the most performing measure of RF for this purpose.