Risk prediction model predicts VTE recurrence after initial thrombotic event

Reference

Timp JF, Braekkan SK, Lijfering WM, et al. Prediction of recurrent venous thrombosis in all patients with a first venous thrombotic event: The Leiden Thrombosis Recurrence Risk Prediction model (L-TRRiP). PLoS Med 2019;16(10):e1002883.

Synopsis

These authors evaluated patients from a population-based case-control study that identified nearly 5000 adults with an objectively diagnosed first VTE (deep vein thrombosis of the leg or pulmonary embolism). The researchers gathered extensive information about possible VTE risk factors, collected blood samples after completion of anticoagulation therapy, and followed up patients for a median of 5.7 years to determine if there was a subsequent VTE. All the data they collected were then subjected to a whirlwind of statistical jujitsu to develop prediction models to develop and validate internally, and then validate externally on a different cohort of adults. They developed 4 different models that used a variety of variables, including: sex, type and location of the first VTE, surgery, pregnancy, hormone use, use of cast, immobilization in bed, cardiovascular disease, blood group (non-O vs O), Factor V Leiden mutation, and several other laboratory parameters. During the follow-up period 507 patients experienced another VTE for an incidence rate of 26.4 per 1,000 person-years. Each of the models performed similarly, so the researchers validated the 2 most parsimonious models on another cohort of patients. The authors don't report the sensitivity or specificity of the models, just that the models performed similarly well in each of the validation exercises, which is disappointing given the whirlwind of statistics. The authors don't compare their models with any of the other existing models.