Sex-specific risk prediction models for aneurysmal subarachnoid haemorrhage - a UK Biobank study

Abstract

Background: We recently developed and validated the SMA2SH2ERS risk prediction model for aneurysmal subarachnoid hemorrhage (ASAH) in the general population (c-statistic 0.62; 95% confidence interval [CI] 0.60–0.64). Given that women have higher ASAH incidence than men, and that predictors for ASAH have different effect sizes between sexes, we developed sex-specific risk prediction models. Methods: Data from the prospective UK Biobank Study were used for model development. Participants with ASAH (per hospital-based ICD codes) before baseline or with missing predictor data were excluded. We developed multivariable Cox proportional hazards models for women and men separately to study the association between earlier recognized SMA2SH2ERS predictors and incident ASAH. Predictive performances were assessed with c-statistics and calibration plots and corrected for overfitting using bootstrapping. Results: A total of 246,771 women and 210,085 men were included with median follow-up of 12 years. ASAH incidence rate per 100 000 person years was 16.1 in women, and 10.7 in men. The women-specific model had a c-statistic of 0.63 (95% CI 0.60–0.65) and the mean predicted absolute 10-year ASAH risk was 0.15%. Independent predictors for women were higher age, family history of stroke, former and current smoking, alcohol consumption, and intermediate education. The men-specific model c-statistic was 0.57 (95% CI 0.53–0.60) and the mean 10-year risk 0.10%. Independent predictors for men were higher age, hypertension, and smoking status. Conclusion: The sex-specific models did not perform better than the general SMA2SH2ERS model in women or in men. Further validation studies are needed before clinical use can be recommended.