Abstract 18019: Race and Sex Sensitive Stroke Risk Prediction: Analysis of Pooled Data From Atherosclerosis Risk in Communities (ARIC) and Cardiovascular Health Study (CHS) Cohorts

Abstract

Context: A risk assessment tool predicting increased risk of first stroke helps to identify at-risk individuals and aid in development of intervention strategies leading to the primary prevention of stroke. Most relied-upon Framingham Heart Study (FHS) stroke risk profile (FRS) does not include race due to lack of information on race other than white in FHS. Objective: To develop a race and sex sensitive risk score for incident stroke. Methods: Risk prediction functions were estimated using pooled data from the ARIC and CHS studies on 16,379 participants (55% women, 23% black) with 1,363 (8.3%) incident stroke during 263,295 person-years with median follow-up 18.3 years. Study sample included individual ages 45-84 years and free of stroke and cancer at the initial examination. Study sample did not include stroke events or individuals lost due to other events or follow-up during the first year as well as individuals with missing data on factors of interest. Cox proportional hazards regression analysis was performed after confirming the proportionality of hazards. Akaike (AIC) and Bayesian information criterion (BIC) was used to develop the best fitting model. Results: The risk equation included age, sex, race (non-black/black), left ventricular hypertrophy, peripheral arterial disease, current smoking, prior myocardial infarction, atrial fibrillation (AF), waist to hip ratio, systolic blood pressure (SBP), and diastolic blood pressure, use of antihypertensive therapy (HTNMED), fasting blood sugar and serum creatinine. The risk equation also included interactions between age x race, age x sex, race x AF, and SBP x HTNMED. All predictors included were significantly associated at p <0.05 except for race x AF (p = 0.078). The risk equation demonstrated good discrimination (Harrell’s C = 0.7877) with receiver operator (ROC) curve statistics for 10-year stroke risk of 0.7521 for non-black and 0.7012 for black. The ROC curve statistics for 10-year stroke risk estimated from the FRS were 0.7099 for non-black and 0.6552 for black. Conclusion: This improved stroke risk predictions derived from a large population based data provide race and sex sensitive risk prediction of first stroke based on predictors readily available during routine clinical examination.