Abstract WP214: Differential Impact of Stroke Risk Factors Across the Age Spectrum: The Reasons for Geographic and Racial Differences in Stroke (REGARDS) Study

Abstract

Introduction: With the “graying of America” an increasing proportion of stroke events occur in the very old; however, risk models to predict stroke were developed in studies with few elderly participants. We hypothesized that, among those aged 45+, predictive factors for stroke differ in older than younger people. Methods: REGARDS enrolled 30,231 black and white adults 45 and older in 2003-7. We considered 3 age strata (45-65, 66-73 and 74+) with approximately equal number of stroke events, and hence equal statistical power to detect risk factor associations. Age specific race-sex adjusted associations of risk factors were assessed using proportional hazards models. We used bootstrap techniques to test the likelihood of specific risk factors being selected for a prediction model in each of the 3 age groups. This was based on the percentage of times each risk factor was selected for inclusion by backwards elimination in 1000 bootstrap model. Results: 27,123 stroke-free participants were followed for 9.4 years, during which 1,114 ischemic strokes occurred (379, 364 and 371 across age groups). Among those age 65 and younger, risk factors with the strongest association with incident stroke were heart disease, hypertension, atrial fibrillation, diabetes, and chronic kidney disease (HRs >2; see table). Hypertension, diabetes, heart disease, smoking, atrial fibrillation, and frailty were selected in the multivariable model for those age 45-65. The impact of some factors decreased at older age 74+, where atrial fibrillation, LVH, smoking, hypertension and frailty emerged as most important. Conclusions: Many “traditional” stroke risk factors were defined in middle-aged populations (and were confirmed here); however, a different milieu of risk factors are important for stroke risk prediction in the elderly.