Prediction of primary venous thromboembolism based on clinical and genetic factors within the U.K. Biobank

Figure 1. Adjusted hazard ratios for primary venous thromboembolism for established risk factors. Shown are the adjusted hazard ratios for primary venous thromboembolism for nine established risk factors. For any individual risk factor, models were adjusted for all other established risk factors and the first four principal components of ancestry. The I bars represent 95% confidence intervals. Body mass index (*) was measured in kg/m2.

Figure 2. Adjusted hazard ratios for primary venous thromboembolism for common medications and non-cancer illnesses. Shown are the adjusted hazard ratios for primary venous thromboembolic events for two factors of interest: common medications (A) and non-cancer illnesses (B). All estimates were adjusted for all nine established risk factors and the first four principal components of ancestry. The I bars represent 95% confidence intervals.

Figure 3. Model performance of a genetic, clinical, and combined score. (A) Shows the density of the polygenic risk score stratified by the 50th, 75th, 90th, 95th, 99th and 99.5th percentiles. (B) Shows the odds ratio for each polygenic risk score percentile. (C) Shows the concordance values, derived from Fine-Gray models, for the De Haan, genetic, clinical, and combined scores. The concordance of the combined model is significantly higher than the concordance of each of the other models (P < 0.001). The I bars represent 95% confidence intervals.

Figure 4. Prediction of venous thromboembolism. (A) and (B) Show the rates of venous thromboembolism derived from Fine-Gray models for the genetic and clinical score, respectively. (C) Shows 10-year event rates for venous thromboembolism, stratified by both the clinical score and genetic score. The I bars represent 95% confidence intervals.