Development and validation of Cardiovascular disease risk prediction model for patients with Chronic kidney disease stage 3-5 within 5 years

Abstract Background and aims: Cardiovascular disease (CVD) is the leading reason for death in patients who are with chronic kidney disease (CKD). However, with progress of CKD, the ability of traditional CVD risk factors to predict clinical outcomes weakens, and non-traditional risk factors play a key role in the pathogenesis of CVD. Previous prediction models based only on traditional CVD risk show limitations and inaccuracies. Our study aims to develop and validate a prediction model for CVD risk including traditional and non-traditional risk factors in stage 3-5 CKD patients within 5 years in China. Methods: 301 patients with CKD stage 3-5 were recruited from January 2010 to January 2022 and followed up till July 2022. Lasso regression and multivariate logistic regression were used to identify baseline predictors for model development, regression modeling was performed using logistic regression and internally validated using tenfold cross-validation. Discrimination and calibration of resulting prediction models were assessed using c-statistic and P-value of the Hosmer-Lemeshow test. Decision curve analysis was performed to assess clinical effectiveness. Results: During follow-up, 169 developed first CVD events within 5 years. The median time of occurrence was 10 months. Of 29 clinical parameters, 11 variables were finally identified as significant predictors and included in the prediction model. 4 prediction models were created in a derivation cohort: original, inflammation, imaging and full model. Full model had the lowest AIC of 311.531 and P-value of 0.3319 of the Hosmer-Lemeshow test. Conclusions: A nomogram was constructed to predict the risk of CVD for CKD patients.