11-LB: A Prediction Model on Incident CKD among Individuals with T2D in the United States

Abstract

CKD is highly prevalent and costly in individuals with diabetes. Clinicians need to have a tool for early identification and prediction of incident CKD in individuals with diabetes. This study aimed to predict incident CKD among individuals with diabetes. A time-varying Cox model was derived from the ACCORD clinical trial data to predict the risk of incident CKD, defined as eGFR<60 mL/min/1.73 m2 or UACR>30 mg/g lasting for 3 months. A list of candidate variables was chosen based on literature reviews and experts’ consultation, including demographic characteristics, physical exam, laboratory, medical history, drug use, and healthcare utilization. Predictors were selected from candidate variables using a stepwise algorithm. Data were split into training and validation set. Model performance was evaluated by Brier score (BS) and C statistics. Confidence intervals (CI) was calculated using a bootstrap method. Decomposition analysis was conducted to assess the predictor contribution. External validation was performed on patient-level data of HARMONY Outcome clinical trial. A total of 6,006 diabetes patients free of CKD at baseline were used for model development, with a median follow-up of 3 years and 2,257 new onset CKD events. The CKD risk model identified age at T2D diagnosed, smoking status, BMI, HDL, VLDL, ALT, eGFR, UACR, hypoglycemia event, retinopathy event, CHF event, CHD history, antihyperlipidemic drug use, antihypertensive drug use, and hospitalization. The model demonstrated good discrimination (C-statistics 0.828 [95% CI 0.812-0.845]) and calibration (BS 0.0552 [95% CI 0.0474-0.0593]) performance. UACR, eGFR, and CHF event were the top 3 factors that contributed most to the prediction. A total of 8,221 patients free of CKD was selected from HARMONY Outcome trial with 1,288 CKD events and median follow-up of 2 years. The model demonstrated acceptable discrimination (C-statistics: 0.772 [95% CI 0.767-0.805]) and calibration (BS: 0.0504 [95% CI 0.0477-0.0531]) in external data. A risk prediction of incident CKD among T2D patients was developed and validated for use in decision support of CKD prevention. Disclosure Y. Lin: None. H. Shao: Board Member; BRAVO4HEALTH, LLC. A. H. Anderson: None. V. Fonseca: Consultant; Abbott, Asahi Kasei Corporation, Bayer AG, Novo Nordisk, Sanofi, Research Support; Fractyl Health, Inc., Jaguar Gene Therapy, Stock/Shareholder; Abbott, Amgen Inc., BRAVO4Health, Mellitus Health. V. Batuman: None. L. Shi: None.