Improved cardiovascular disease risk prediction by implementing sex-specific cut-offs for intervention and new risk markers

• Published in: European heart journal Vol. 43; no. Supplement_2
• Main Authors: Bye, A, Wiig-Fisketjon, A, Dalen, H, Langaas, M
• Format: Journal Article
• Language: English
• Published: 03.10.2022
• ISSN: 0195-668X; 1522-9645
• DOI: 10.1093/eurheartj/ehac544.2283

Abstract Background More than 17 million people die from cardiovascular disease (CVD) annually. CVDs are preventable and several risk prediction models are available for determining the 10-year risk of CVD, including the Norwegian NORRISK2 model. However, the available risk prediction models only explain a modest proportion of the incidence. For myocardial infarction (MI), it is estimated that 15–20% of the patients have none of the traditional risk factors and would be classified as “low risk”. Purpose Aim was to develop improved models for predicting the 10-year risk of MI. Methods We included 31.946 participants from the third wave of the Trøndelag Health Study (HUNT3) with no previous CVDs. HUNT data included 101 variables from interviews, clinical measurements, and biological samples on each participant. Totally, 11% of the men and 6% of the women experienced an MI between the HUNT3 and HUNT4 (10-year follow-up). The dataset was split 80/20 into a training set and a test set. XGBoost and logistic regression (LR) were used to fit two models for each sex predicting MI including variables from HUNT3. The models were evaluated by the area under the Receiver-Operating-Characteristic (ROC) curve and the Precision-Recall (PR) curve, both for the full test set and the test set divided into age groups. Thresholds for classification were suggested by maximizing different performance measures through 10-fold cross-validation on the training set. We then explored age- and sex-specific thresholds for intervention with a reasonable trade-off between sensitivity and specificity. All results were compared with NORRISK 2, which was implemented and applied to the same test set for exact comparison. Results For men, the XGBoost model improved risk prediction compared to NORRISK 2 for all age groups (AUC-ROC for XGBoost and NORRISK 2, respectively, 0.72 and 0.65 (age 45–54), 0.63 and 0.62 (age 55–64), 0.69 and 0.62 (age 65–74)). The liver-related enzyme alkaline phosphatase (ALP) was among the new predictors for MI in men. For women, NORRISK 2 performed best when evaluated by ROC curves, however, when evaluated by PR curves, the XGBoost models indicated improved prediction compared to NORRSIK 2 in the women 55–64 years (AUC-PR for XGBoost and NORRISK 2, respectively, 0.20 and 0.12). The thyroid stimulation hormone (TSH) was among the new predictors for MI in women. Regarding NORRISK 2, our results indicated that the thresholds for intervention should be increased for all age groups in men and decreased for all age groups in women for improved balance between sensitivity and specificity. Conclusion New risk factors should be considered implemented in CVD risk prediction algorithms, for improved identification of individuals at increased risk of MI. In addition, implementing sex-specific thresholds for intervention could be a useful step towards improved prevention of CVD for both men and women. Funding Acknowledgement Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Regional Health Authorities