Longevity & AgingResearch PaperPaywall

GDF-15 fails to predict bleeding risk in patients on anticoagulation for blood clots

A 383-patient cohort finds the aging-linked biomarker GDF-15 did not predict bleeding during VTE anticoagulation or improve an established risk score.

Thursday, November 12, 2026 0 views
Published in J Thromb Haemost
Glowing GDF-15 protein molecule beside a blood vessel with a small clot and flowing red cells, clinical blue tones

Summary

GDF-15, a stress-response protein that rises with age and disease, has been proposed as a marker for bleeding risk on blood thinners. Researchers in Vienna tested this in 383 adults with acute pulmonary embolism or deep vein thrombosis, excluding people with active cancer and pregnant or peripartum patients. They measured GDF-15 at baseline and 3 months and followed participants for up to 2 years. Forty-nine patients had a first clinically relevant bleed, a 1-year cumulative incidence of 11.9%. Baseline GDF-15 was not associated with bleeding, even after adjusting for age, sex, and bleeding history. Adding it to the VTE-PREDICT bleeding score did not meaningfully improve prediction. The ABC-bleeding score, which includes GDF-15, showed only modest discrimination. In this younger, cancer-free population, GDF-15 does not appear useful for bleeding risk assessment.

0:00--:--

Detailed Summary

Anticoagulation prevents recurrent blood clots but carries a bleeding risk, and clinicians lack reliable tools to identify who is most vulnerable. Growth differentiation factor-15 (GDF-15), a stress-responsive protein that rises with aging and chronic disease, has been proposed as a bleeding biomarker and is included in the ABC-bleeding score used in atrial fibrillation.

This analysis was part of BACH-VTE, a cohort of patients with acute symptomatic pulmonary embolism or deep vein thrombosis followed for 2 years. Patients with active cancer and pregnant or peripartum patients were excluded. GDF-15 was measured at baseline and at 3 months. The authors assessed clinically relevant bleeding (major plus clinically relevant non-major bleeding) using models that accounted for competing risks of death and stopping anticoagulation. They also tested whether GDF-15 improved the VTE-PREDICT score and how the ABC-bleeding score performed.

Among 383 patients (median age 55, 39.2% female), 49 had a first clinically relevant bleed, a 1-year cumulative incidence of 11.9%. Median GDF-15 was 968 pg/mL at baseline and 836 pg/mL at 3 months. Baseline GDF-15 showed no association with bleeding (adjusted subdistribution hazard ratio 0.96 per doubling, 95% CI 0.70-1.30). It did not meaningfully improve VTE-PREDICT, and ABC-bleeding discriminated only modestly.

These findings suggest GDF-15 should not be used on its own, or added to existing scores, to estimate bleeding risk in cancer-free VTE patients on anticoagulation.

Caveats include a modest sample and event count, a relatively young cohort, and exclusion of cancer patients, who carry high bleeding risk. Results may not generalize to older patients or other indications. Only the abstract was available for this summary, so details of score performance are limited.

Key Findings

  • Baseline GDF-15 was not associated with clinically relevant bleeding (adjusted subdistribution HR 0.96 per doubling, 95% CI 0.70-1.30).
  • Of 383 VTE patients, 49 had a first clinically relevant bleed, a 1-year cumulative incidence of 11.9%.
  • Adding GDF-15 did not meaningfully improve the predictive performance of the VTE-PREDICT bleeding score.
  • The ABC-bleeding score, which includes GDF-15, showed only modest discrimination for bleeding in this cohort.
  • Median GDF-15 was 968 pg/mL at baseline and 836 pg/mL at 3 months.

Methodology

Prospective BACH-VTE cohort of patients with acute symptomatic pulmonary embolism or deep vein thrombosis, followed for 2 years, excluding active cancer and pregnancy/peripartum. GDF-15 was measured at baseline and 3 months, and competing-risk regression (death, anticoagulation stop) was used. Score performance was assessed by discrimination, calibration, and overall performance.

Study Limitations

Only 49 bleeding events among 383 patients limit statistical power and precision. The cohort was relatively young (median age 55) and excluded patients with cancer, limiting generalizability. This summary is based only on the abstract, so detailed score metrics were unavailable.

Enjoyed this summary?

Get the latest longevity research delivered to your inbox every week.

Enter your email to subscribe: