Presentation (CTI)

OBJECTIVES: Cost-effectiveness models (CEMs) for cardiovascular therapies typically require long-term predictions derived from clinical trial data. To address evidence gaps, CEMs incorporate risk equations for specific clinical outcomes. The aim of this study was to develop such risk equations while balancing statistical rigor with clinical relevance.
METHODS: Using a prospective cohort design, we estimated risk equations from SELECT trial data for acute coronary syndrome (ACS) and stroke. Expert elicitation identified and prioritized risk factors for the SELECT population (overweight and cardiovascular disease). Clinical experts ranked potential risk factors by importance, followed by discussions to incorporate their insights. We used a least absolute shrinkage and selection operator (LASSO) approach combined with clinical input to identify predictors based on both clinical and statistical relevance.
RESULTS: Clinical experts identified key predictors for ACS (including CKD; body weight; blood pressure; lipid levels; smoking status; cardiovascular history; and hsCRP) and stroke (including age; CKD; prior stroke/TIA; body weight; blood pressure; sleep apnea; lipid medications; smoking status; hsCRP; atrial fibrillation; and DVT). These predictors were included in a preliminary set of risk equations constructed using interaction testing and LASSO penalization. The preliminary risk equations were then reviewed by clinical experts who evaluated the clinical validity and suggested adjustments. Based on their feedback, several predictors were modified, such as substituting LDL cholesterol with the total/HDL cholesterol ratio in the ACS model. The final risk equations incorporated refined clinical factors validated through expert input, achieving good predictive capabilities and clinical applicability.
CONCLUSIONS: We developed robust risk equations for cardiovascular events from the SELECT trial by integrating clinical expertise throughout the process. This collaborative approach aligns with CEM requirements in HTA settings while emphasizing clinically relevant predictors. To enhance the validity and transparency of future risk equations that rely on clinical expertise, it is beneficial to consistently report findings on factors of interest.