Previous models for contrast-induced acute kidney injury (CI-AKI) after percutaneous coronary intervention (PCI) include procedure-related variables in addition to pre-procedural variables. We sought to develop a risk model for CI-AKI based on pre-procedural variables and compare its predictability with a conventional risk model and also to develop an integer score system based on selected variables. A total of 5,936 consecutive PCIs registered in the Japanese Cardiovascular Database were analyzed (derivation cohort, n = 3,957; validation cohort, n = 1,979). CI-AKI was defined as an increase in serum creatinine of 50% or 0.3 mg/dl compared with baseline. From the derivation cohort, 2 different CI-AKI risk models were generated using logistic regression analyses: a pre-procedural model and a conventional model including both pre-procedural and procedure-related variables. The predictabilities of the models were compared by c-statistics. An integer score was assigned to each variable in proportion to each estimated regression coefficient for the final model. In our derivation cohort, the proportion of CI-AKI was 9.0% (n = 358). Predictors for CI-AKI included older age, heart failure, diabetes, previous PCI, hypertension, higher baseline creatinine level, and acute coronary syndrome. Presence of procedure-related complications and insertion of intra-aortic balloon pumping were included as procedure-related variables in the conventional model. Both the conventional model (c-statistics 0.789) and the pre-procedural model (c-statistics 0.799) demonstrated reasonable discrimination. The integer risk-scoring method demonstrated good agreement between the expected and observed risks of CI-AKI in the validation cohort. In conclusion, the pre-procedural risk model for CI-AKI had acceptable discrimination compared with the conventional model and may aid in risk stratification of CI-AKI before PCI.
Copyright © 2015 Elsevier Inc. All rights reserved.