External validation of the Leiden Thrombosis Recurrence Risk Prediction models (L-TRRiP) for the prediction of recurrence after a first venous thrombosis in the Heart and Vascular Health study

Background

Long-term outcome after a first venous thromboembolism (VTE) might be optimized by tailoring anticoagulant treatment duration on individual risks of recurrence and major bleeding. The L-TRRiP models (A–D) were previously developed in data from the Dutch Multiple Environment and Genetic Assessment of Risk Factors for Venous thrombosis study to predict VTE recurrence.

Objectives

We aimed to externally validate models C and D using data from the United States Heart and Vascular Health (HVH) study.

Methods

Data from participants with a first VTE who discontinued initial anticoagulant therapy were used to determine model performance. Missing data were imputed, and results were pooled according to Rubin’s rules. To determine discrimination, Harrell’s C-statistic was calculated. To assess calibration, the observed/expected (O/E) ratio was estimated, and calibration plots were created, in which we accounted for the competing risk of death. A stratified analysis based on age <70 or >70 years was performed.

Results

Of 1430 participants from the HVH study, 187 experienced an unprovoked VTE recurrence during follow-up. The C-statistics of L-TRRIP models C and D were 0.62 (95% CI, 0.56-0.67) and 0.61 (95% CI, 0.55-0.67), respectively. The O/E ratio (1.00; 95% CI, 0.84-1.17 and 1.09; 95% CI, 0.91-1.27, respectively) and calibration plots indicated good calibration. The discrimination was similar between participants <70 or >70 years, whereas overall calibration was lower in participants <70 years.

Conclusion

The L-TRRiP models showed moderate discrimination and good calibration in a different population and can be used to guide clinical decision making. To assess the added value in daily clinical practice, a management study is needed.