J-015 DEVELOPMENT OF A STROKE RISK PREDICTION MODEL FOR JAPANESE

Background Currently, available risk prediction tools of stroke are derived mainly from western populations and few risk prediction tools are developed for the Japanese general population. Methods A total of 25,054 Japanese men and women (aged 40–89 years) without a history of cardiovascular events from 11 community-based cohorts were followed up for incident total stroke and stroke subtypes. A multivariable stroke risk prediction model was developed that incorporated age, sex, blood pressure, body mass index, diabetes and smoking. Scoring systems were constructed based on multivariable Poisson regression models for predicting a 5-year probability of developing stroke. Results In an average 8.2-year follow-up, 509 cerebral infarction and 148 cerebral hemorrhage were observed. In multivariable models, hazard ratios of total stroke increased with elevating blood pressure categories of normal, high normal, stage 1 HT and stage 2 HT groups were 1.00, 1.82 (95%CI, 1.42–2.33), 2.26 (1.68–3.05), and 3.49 (2.30–5.29) for men, those for women 1.00, 1.70 (95%CI, 1.35–2.13), 2.04 (1.52–2.4), and 3.86 (2.58–5.78), respectively. Scoring systems were constructed based on multivariable Poisson regression models for predicting a 5-year probability of developing stroke. Area under the receiver operating curve (AUC) of cerebral infarction and cerebral hemorrhage were 0.809 and 0.768, respectively. Conclusion We developed a new stroke risk prediction model in a general population of Japanese. The risk prediction model would provide a useful guide to estimate absolute risk of cardiovascular disease and to treat individual risk factors.