Thrombo-vera: a new thrombosis risk model for polycythemia vera using modern variable selection methods.

Abstract

Background: Thrombosis is a major complication in polycythemia vera (PV), contributing to significant morbidity and mortality. This retrospective study aimed to develop a predictive model for thrombosis risk in PV patients using advanced statistical techniques.

Research design and methods: The study included 817 consecutive PV patients, with a median follow-up of 59 months. A Bayesian logistic regression model with sparsity-inducing R2D2 priors was used to predict thrombosis.

Results: Thrombotic events occurred in 13.2% of patients. The thrombosis group had significantly higher median neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), splenomegaly, cardiovascular risk factors, microvascular symptoms, pruritus, previous thrombosis, and Charlson Comorbidity Index (CCI) compared to the no-thrombosis group. Both groups were comparable in age. Multivariate regression analysis identified CCI, PLR, splenomegaly, and microvascular symptoms as key predictors of thrombosis. A clinical score, ThromboVera CS, was developed based on these predictors, classifying patients into low, moderate, or high-risk groups. In the low-risk group, 6.94% experienced thrombosis, compared to 15.76% in moderate-risk group and 48.78% in the high-risk group.

Conclusions: The ThromboVera CS score is a reliable, easy-to-use tool for predicting thrombosis in PV patients. It can help clinicians identify those at high risk, enabling early intervention that could significantly improve patient outcomes by targeting nearly 50% of high-risk patients.