Estimated glomerular filtration rate better predicts 30-day mortality after non-cardiac surgery than serum creatinine: a retrospective analysis of 92,888 patients

Abstract

BACKGROUND: Serum creatinine is the most commonly used indicator of renal function, but its derivative, estimated glomerular filtration rate (eGFR), has been shown to be superior in non-surgical settings. It remains unknown if eGFR better predicts postoperative mortality in non-cardiac surgical patients. We thus tested the hypothesis that eGFR predicts 30-day mortality after non-cardiac surgery better than serum creatinine. METHODS: We evaluated patients who had inpatient non-cardiac surgery of at least one hour duration during January 2006 to December 2011 at the Cleveland Clinic Main Campus and whose preoperative serum creatinine was measured within 30 days before surgery. The eGFR was calculated using the Chronic Kidney Disease-Epidemiology Collaboration equation. Preoperative eGFR was compared in a multivariable analysis with preoperative serum creatinine (both assessed as continuous variables) on the ability to predict 30-day mortality in all patients. Secondarily, the comparison was made within subgroups based on amount of blood loss, blood transfusion, and sex. RESULTS: There were 92,888 patients included in the final analysis. The eGFR was a modestly better discriminator of 30-day mortality than serum creatinine, with an estimated c-statistic (95% confidence interval) of 0.67 (0.65 to 0.68) for eGFR vs 0.61 (0.59 to 0.63) for serum creatinine (P < 0.001). Furthermore, the eGFR was consistently a better discriminator of 30-day mortality across blood loss, transfusion, and sex groups. Reclassification analyses suggested improved individual predictions of 30-day mortality using eGFR compared with serum creatinine. Nevertheless, a multivariable combination of baseline characteristics of American Society of Anesthesiologists physical status, age, and body mass index (all P < 0.001) discriminated 30-day mortality with a c-statistic of 0.850. Adding eGFR to the model improved the c-statistic to only 0.851, while separately adding serum creatinine did not change the c-statistic. CONCLUSION: The eGFR is a modestly better predictor of 30-day mortality than serum creatinine in patients having inpatient non-cardiac surgery. Given that eGFR is often reported by clinical laboratories and is otherwise easy to calculate, it should generally be used in preference to creatinine alone.