Development of Sex-Stratified Prediction Models for Recurrent Venous Thromboembolism: A Danish Nationwide Cohort Study

Abstract

To optimize decision making for anticoagulant treatment duration after incident venous thromboembolism, we derived and internally validated two clinically applicable sex-specific prediction models for venous thromboembolism recurrence, discarding the traditional categorization of provoked and unprovoked venous thromboembolism. This study was based on data from Danish nationwide registries. We identified all routine care in- and outpatients with completed anticoagulant treatment for incident venous thromboembolism from 2012 through 2017. The outcome was recurrent venous thromboembolism within 2 years. Risk scores were derived using Cox regression analysis and a backward selection process on a set of 24 potential predictors. Performance was assessed through calibration and discrimination using bootstrap techniques to internally validate the scores. The study included 11,519 patients. Risk scores under the joint acronym AIM-SHA-RP were developed. Age, Incident pulmonary embolism, and recent Major surgery were predictors for both sexes; Statin treatment, Heart disease and Antiplatelet treatment were predictors specifically for men, while chronic Renal disease and recent Pneumonia or sepsis were predictors specifically for women. The risk scores were well calibrated and identified a low- (< 5%), intermediate- (5-10%), and high-risk (> 10%) group for both sexes. Generally, discriminative capacities, as measured by the c-statistic, were limited. We developed two clinically applicable risk scores to estimate the risk of recurrent venous thromboembolism after completed anticoagulant treatment. The risk scores can potentially guide treatment duration of anticoagulation after incident venous thromboembolism but require further external validation before implemented in clinical practice.