Evaluating surrogate measures of renal dysfunction after cardiac surgery

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) (DeltaCr(72h)) and the percentage 72-h change in calculated (Cockcroft-Gault equation) Cr clearance (%DeltaCrCl(72h)) 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. DeltaCr(72h) was skewed to the right, whereas %DeltaCrCl(72h) was normally distributed. ROC curve areas showed DeltaCr(72h) to be a good predictor of dialysis (0.98), death (0.83), and prolonged hospitalization (0.74). %DeltaCrCl(72h) 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 DeltaCr(72h) to %DeltaCrCl(72h), results in a normal distribution that is amenable to parametric statistical tests. DeltaCr(72h) and %DeltaCrCl(72h) may be used as surrogate outcomes in future trials.