Evaluating Surrogate Measures of Renal Dysfunction After Cardiac Surgery

Publisher Website
Google Scholar
Abstract
Renal insufficiency after cardiac surgery is associated with increased mortality, morbidity, and length of intensive care unit stay. A convenient surrogate measure would facilitate the evaluation of renal-protective therapies. We evaluated two measures: the 72-h change in serum creatinine (Cr) (ΔCr72h) and the percentage 72-h change in calculated (Cockcroft-Gault equation) Cr clearance (%ΔCrCl72h). We randomly selected 2000 individuals who underwent aortocoronary bypass, valve surgery, or both at the Toronto General Hospital between May 1999 and August 2000. The variables were analyzed with frequency histograms and normal probability plots. Their association with dialysis, mortality, and prolonged intensive care unit stay was determined by using receiver operating characteristic (ROC) curves. ΔCr72h was skewed to the right, whereas %ΔCrCl72h was normally distributed. ROC curve areas showed ΔCr72h to be a good predictor of dialysis (0.98), death (0.83), and prolonged hospitalization (0.74). %ΔCrCl72h had similar ROC curve areas for predicting dialysis (0.97), death (0.82), and prolonged hospitalization (0.74). ROC curve areas did not differ significantly with respect to mortality (P = 0.89), dialysis (P = 0.49), or prolonged hospitalization (P = 0.85). Both variables were correlated with patient-relevant outcomes. Mathematical transformation of ΔCr72h to %ΔCrCl72h results in a normal distribution that is amenable to parametric statistical tests. ΔCr72h and %ΔCrCl72h may be used as surrogate outcomes in future trials.