David W Allen1, Bryan Ma2, Kelvin C Leung2, Michelle M Graham3, Neesh Pannu3, Mouhieddin Traboulsi4, David Goodhart4, Merril L Knudtson4, Matthew T James5. 1. Department of Cardiac Sciences, Cumming School of Medicine, University of Calgary, Calgary, Alberta, Canada. 2. Department of Medicine, Cumming School of Medicine, University of Calgary, Calgary, Alberta, Canada. 3. Department of Medicine, Faculty of Medicine, University of Alberta, Edmonton, Alberta, Canada. 4. Department of Cardiac Sciences, Cumming School of Medicine, University of Calgary, Calgary, Alberta, Canada; Department of Medicine, Cumming School of Medicine, University of Calgary, Calgary, Alberta, Canada. 5. Department of Medicine, Cumming School of Medicine, University of Calgary, Calgary, Alberta, Canada; Libin Cardiovascular Institute of Alberta, Cumming School of Medicine, University of Calgary, Edmonton, Alberta, Canada; Department of Community Health Sciences, Cumming School of Medicine, University of Calgary, Edmonton, Alberta, Canada. Electronic address: mjames@ucalgary.ca.
Abstract
BACKGROUND: Identification of patients at risk of contrast-induced acute kidney injury (CI-AKI) is valuable for targeted prevention strategies accompanying cardiac catheterization. METHODS: We searched MedLine and EMBASE for articles that developed or validated a clinical prediction model for CI-AKI or dialysis after angiography or percutaneous coronary intervention. Random effects meta-analysis was used to pool c-statistics of models. Heterogeneity was explored using stratified analyses and meta-regression. RESULTS: We identified 75 articles describing 74 models predicting CI-AKI, 10 predicting CI-AKI and dialysis, and 1 predicting dialysis. Sixty-three developed a new risk model whereas 20 articles reported external validation of previously developed models. Thirty models included sufficient information to obtain individual patient risk estimates; 9 using only preprocedure variables whereas 21 included preprocedural and postprocedure variables. There was heterogeneity in the discrimination of CI-AKI prediction models (median [total range] in c-statistic 0.78 [0.57-0.95]; I2 = 95.8%, Cochran Q-statistic P < 0.001). However, there was no difference in the discrimination of models using only preprocedure variables compared with models that included postprocedural variables (P = 0.868). Models predicting dialysis had good discrimination without heterogeneity (median [total range] c-statistic: 0.88 [0.87-0.89]; I2 = 0.0%, Cochran Q-statistic P = 0.981). Seven prediction models were externally validated; however, 2 of these models showed heterogeneous discriminative performance and 2 others lacked information on calibration in external cohorts. CONCLUSIONS: Three published models were identified that produced generalizable risk estimates for predicting CI-AKI. Further research is needed to evaluate the effect of their implementation in clinical care.
BACKGROUND: Identification of patients at risk of contrast-induced acute kidney injury (CI-AKI) is valuable for targeted prevention strategies accompanying cardiac catheterization. METHODS: We searched MedLine and EMBASE for articles that developed or validated a clinical prediction model for CI-AKI or dialysis after angiography or percutaneous coronary intervention. Random effects meta-analysis was used to pool c-statistics of models. Heterogeneity was explored using stratified analyses and meta-regression. RESULTS: We identified 75 articles describing 74 models predicting CI-AKI, 10 predicting CI-AKI and dialysis, and 1 predicting dialysis. Sixty-three developed a new risk model whereas 20 articles reported external validation of previously developed models. Thirty models included sufficient information to obtain individual patient risk estimates; 9 using only preprocedure variables whereas 21 included preprocedural and postprocedure variables. There was heterogeneity in the discrimination of CI-AKI prediction models (median [total range] in c-statistic 0.78 [0.57-0.95]; I2 = 95.8%, Cochran Q-statistic P < 0.001). However, there was no difference in the discrimination of models using only preprocedure variables compared with models that included postprocedural variables (P = 0.868). Models predicting dialysis had good discrimination without heterogeneity (median [total range] c-statistic: 0.88 [0.87-0.89]; I2 = 0.0%, Cochran Q-statistic P = 0.981). Seven prediction models were externally validated; however, 2 of these models showed heterogeneous discriminative performance and 2 others lacked information on calibration in external cohorts. CONCLUSIONS: Three published models were identified that produced generalizable risk estimates for predicting CI-AKI. Further research is needed to evaluate the effect of their implementation in clinical care.
Authors: John A Dodson; Alexandra Hajduk; Jeptha Curtis; Mary Geda; Harlan M Krumholz; Xuemei Song; Sui Tsang; Caroline Blaum; Paula Miller; Chirag R Parikh; Sarwat I Chaudhry Journal: Am J Med Date: 2019-06-04 Impact factor: 4.965