Preoperative estimated glomerular filtration rate to predict cardiac events in major noncardiac surgery: a secondary analysis of two large international studies

BACKGROUND: Optimised use of kidney function information might improve cardiac risk prediction in noncardiac surgery. METHODS: In 35,815 patients from the VISION cohort study and 9219 patients from the POISE-2 trial who were ≥45 yr old and underwent nonurgent inpatient noncardiac surgery, we examined (by age and sex) the association between continuous nonlinear preoperative estimated glomerular filtration rate (eGFR) and the composite of myocardial injury after noncardiac surgery, nonfatal cardiac arrest, or death owing to a cardiac cause within 30 days after surgery. We estimated contributions of predictive information, C-statistic, and net benefit from eGFR and other common patient and surgical characteristics to large multivariable models. RESULTS: The primary composite occurred in 4725 (13.2%) patients in VISION and 1903 (20.6%) in POISE-2; in both studies cardiac events had a strong, graded association with lower preoperative eGFR that was attenuated by older age (P_interaction<0.001 for VISION; P_interaction=0.008 for POISE-2). For eGFR of 30 compared with 90 ml min^-1 1.73 m^-2, relative risk was 1.49 (95% confidence interval 1.26-1.78) at age 80 yr but 4.50 (2.84-7.13) at age 50 yr in female patients in VISION. This differed modestly (but not meaningfully) in men in VISION (P_interaction=0.02) but not in POISE-2 (P_interaction=0.79). eGFR contributed the most predictive information and mean net benefit of all predictors in both studies, most C-statistic in VISION, and third most C-statistic in POISE-2. CONCLUSIONS: Continuous preoperative eGFR is among the best cardiac risk predictors in noncardiac surgery of the large set examined. Along with its interaction with age, preoperative eGFR would improve risk calculators. CLINICAL TRIAL REGISTRATION: ClinicalTrials.gov NCT00512109 (VISION) and NCT01082874 (POISE-2).