The development of these ASCVD risk prediction models for CKD patients used data obtained from the Chronic Renal Insufficiency Cohort (CRIC) Study, which was a longitudinal cohort study of patients with mild-to-moderate CKD.¹ A total of 2,604 patients (52% male) at 21-74 years of age (mean age=55.8 years) were recruited from 7 clinical centers across the United States.¹ The mean [standard deviation(SD)] estimated glomerular filtration rate (eGFR) was 56.0 (24.7)mL/minute per 1.73m².¹ About 252 incidents of ASCVD events occurred during the first 10 years of follow-up from baseline.¹ ASCVD was defined as the first incidence of stroke (fatal or non-fatal) or myocardial infarction.¹

In this study, 2 prediction models were created and evaluated.¹ The CRIC clinical model included candidate predictor variables, which are outlined in the American College of Cardiology/American Heart Association pooled cohort equations (ACC/AHA PCEs), such as age, sex, race, total cholesterol, high-density lipoprotein (HDL) cholesterol, systolic blood pressure (BP), use of BP-lowering medications, history of diabetes, and current smoking status.¹ The CRIC enriched model included the aforementioned candidate predictor variables, in addition to variables that were specific to CKD patients at high risk for ASCVD, including metabolic factors, kidney disease, lipid metabolism, mineral metabolism, inflammation factors, and cardiac biomarkers.¹

The application of the published ACC/AHA PCEs coefficients to the CRIC sample resulted in an area under the receiver operating characteristic curve (AUC) of 0.730, and a model with coefficients estimated within the CRIC sample having higher discrimination (p=0.03), resulting in an AUC of 0.736 (95% CI: 0.649-0.826).¹ The CRIC clinical model had an AUC of 0.760 (95% CI: 0.678-0.851).¹ The CRIC enriched model, which included novel biomarkers, had an AUC of 0.771 (95% CI: 0.674-0.853), and was significantly higher than the clinical model (p=0.001).¹

Both the clinical and enriched models were well-calibrated and improved the reclassification of non-events as compared with the PCEs (6.6%; 95% CI: 3.7%-9.6% and 10.0%; 95% CI: 6.8%-13.3%, respectively).¹

In conclusion, the 10-year ASCVD risk prediction models created for the CKD population, which included novel kidney and cardiac biomarkers, outperformed the equations created for the general population which used only the conventional risk factors.¹ The results advocated the utilization of CKD-specific prediction tools to estimate the risk of ASCVD in CKD patients who possess different risk profiles as compared with the general population.¹ Additionally, these important prediction tools not only assist the CKD population who are at high risk for ASCVD, but also aid in the identification of those who are at low risk with which invasive procedures are not beneficial.4 The significance of this study may improve shared decision-making between patients and clinicians for preventive therapies to reduce ASCVD events and improve clinical outcomes.