Predictive performance of ten equations for estimating creatinine clearance in cardiac patients

OBJECTIVE: The predictive performance of 10 equations used to estimate creatinine clearance (Cl-cr) was assessed retrospectively from data collected on 420 patients. DESIGN: This study is a retrospective data analysis of information collected on hemodynamically stable patients awaiting coronary angiography during the Iohexol Cooperative Study. SETTING: The Iohexol Cooperative Study was a multicenter study that compared nephrotoxicity of high- and low-osmolar contrast media in patients undergoing coronary angiography. Data used for this analysis were preangiography 24-hour urine collections that were primarily collected in hospitalized patients. PATIENTS: Patients selected from the Iohexol Cooperative Study database for analysis were participants categorized into one or more of six subgroups: elderly (n = 222), hypoalbuminemic (n = 25), chronic renal insufficiency (n = 128), low serum creatinine (n = 115), obese (n = 208), and diabetic (n = 191) who had baseline urine collections of at least 24 hours. OUTCOME MEASURES: Predictive performance was assessed using bias, precision, slopes, and y-intercepts. RESULTS: The Salazar-Corcoran equation was unbiased in the entire group as well as in five of the subgroups. The Cockcroft-Gault equation was unbiased in three of the subgroups. All other equations were biased in predicting Cl-cr in the entire group as well as in at least four of the subgroups. Precision was generally poor. All slopes were significantly different than one and all y-intercepts were significantly different than zero (p < 0.01). Correlation coefficients were between 0.63 and 0.79 with the exceptions of the low serum creatinine subgroup (r values 0.35-0.64) and the Davis-Chandler equation (r values 0.35-0.71 across groups). CONCLUSIONS: Of the equations studied, Salazar-Corcoran and Cockcroft-Gault appear to be the best for predicting Cl-cr.