Background

Patients with active cancer who develop venous thromboembolism (VTE) face elevated risks of both recurrence and anticoagulant-related bleeding complications compared with non-cancer populations [1]. Prior observational studies and clinical experience indicated that risk is heterogeneous and influenced by cancer-related factors (tumor site and stage), patient-related (e.g. performance status, prior VTE or bleeding), and treatment-related (e.g. type of chemotherapy). Existing clinical scores (e.g., the Ottawa score for recurrence, CAT-BLEED and B CAT for bleeding) provided only moderate discrimination with variable external validation [2–4]. Consequently, individualized anticoagulation decisions have relied on integrating non-modifiable patient and cancer characteristics with treatment goals and bleeding tolerance rather than on a single robust prediction model.

Main assumptions of meta-analysis

Khan et al. in the systematic review and meta-analysis published in European Heart Journal synthesizes evidence from 33 adjusted-effect studies including 96,753 adults with cancer-associated VTE to identify prognostic factors for two competing risks: recurrent VTE and clinically relevant anticoagulant-related bleeding (CRB). The definitions of active cancer and outcomes followed International Society on Thrombosis and Haemostasis criteria [5,6].

Recurrent venous thromboembolism risk

For recurrent VTE several factors showed high-certainty associations with increased risk [1]. A previous history of VTE was associated with higher recurrence (aHR 1.50, 95% CI 1.08–2.09). Functional status was highly important. Eastern Cooperative Oncology Group (ECOG) performance status above 0 increased risk (aHR 1.81, 95% CI 1.34–2.46), and ECOG above 1 conferred an even greater risk (aHR 2.44, 95% CI 1.55–3.84). Advanced cancer, including metastatic disease or stage IV, also raised recurrence risk (aHR 1.38, 95% CI 1.15–1.65). Cancer site was a powerful determinant. Pancreatic cancer had the highest pooled estimate, followed by hepatobiliary, lung, and genitourinary cancers. Moderate-certainty evidence suggested that an index event of isolated deep vein thrombosis probably increases recurrence compared with other VTE presentations (aHR 1.49, 95% CI 1.06–2.08). In contrast, recent surgery within three months, breast cancer diagnosis, female sex and increasing age were associated with a lower recurrence risk [1].

Recurrent bleeding risk

Similarly multiple factors demonstrated high-certainty associations with increased CRB risk [1]. A history of bleeding more than doubled risk (aHR 2.41, 95% CI 1.50–3.88). Poor functional status (ECOG ≥2) was similarly impactful (aHR 2.10, 95% CI 1.48–2.99). Advanced cancer diagnosis increased bleeding risk as well (aHR 1.60, 95% CI 1.29–1.97). Bleeding risk varied by cancer site. Brain, gastrointestinal, genitourinary and prostate cancers all conferred higher risk. Moderate-certainty evidence indicated that anemia and liver dysfunction contribute to bleeding risk, and that ongoing cancer treatment modestly increases risk. Notably, age likely has no material effect on CRB on a per-year basis (aHR 1.00, 95% CI 0.97–1.02) [1].

Summary

Clinically the results provided by Khan et al. [1] support a risk-adapted approach to anticoagulation in cancer-associated VTE. These decisions should be made through shared decision-making, acknowledging that most influential predictors are non-modifiable but meaningfully affect risk balance [1]. From a research and implementation perspective the findings highlight limitations of existing prediction tools. The Ottawa score for recurrence and the CAT-BLEED and B CAT scores for bleeding have only moderate performance and limited external validation [2–4]. Incorporating high-certainty predictors identified by Khan et al. [1] may improve discrimination and calibration of future models. To sum up, these robust associations provide a practical background for individualized anticoagulation decisions and should guide the next generation of validated prediction models.