Development and external validation of an international score to predict cancer 1 year after venous thromboembolism.

Abstract

Routine cancer screening is not recommended after venous thromboembolism (VTE), but improved risk stratification may support tailored strategies. We conducted a prospective, multicentre cohort study (2001-2023) including 48,595 patients from the RIETE registry (207 centres, 31 countries). Patients were allocated to derivation (n = 24,670) and internal validation (n = 10,572) cohorts. External validation (n = 4042) included patients from the COMMAND-VTE-II registry and the SOME trial. The primary outcome was cancer diagnosis within 12 months after VTE. Model performance was assessed using discrimination (AUC), calibration, Brier score, and positive predictive value (PPV). Cancer was diagnosed in 1158 patients (3.29%, 95%CI: 3.10-3.48%) within 12-months after VTE. Predictors included male sex, age ≥70 years, chronic lung disease, anemia, platelet count >350 × 10⁹/L, and clinically relevant bleeding occurring during follow-up. The model showed an AUC of 0.63, 0.63 and 0.60 in the derivation, internal validation and external validation cohort, respectively. High-risk classification occurred in 38.4% and 46.7% of patients in the derivation and external validation cohort. The positive predictive value for cancer diagnosis within 1 year was 4.9% (95% CI 4.5-5.3%) in the derivation cohort and 3.3% (95% CI 2.6-4.2%) in the external validation cohort. A nomogram was developed to facilitate individualized risk estimation. We derived and externally validated a simple clinical score to stratify the risk of cancer diagnosis within one year after VTE. Although its discriminative performance was modest its high negative predictive value may help identify patients at lower probability of occult cancer and support risk-adapted follow-up strategies.