v2.0Biomarker-Enhanced Bleeding Risk Validation (COMBINE AF)
Laboratory Inputs
ng/L
Key predictor (~400-800 normal)
ng/L
High-sensitivity Troponin T
g/dL
Lower levels increase risk
years
Clinical Details
Risk Prediction
Requires complete biomarker profile (GDF-15, hs-cTnT, Hemoglobin) to calculate accurate 1-year major bleeding probability.
Guidelines & Evidence
Clinical Details
Section 1
When to Use
When to Use
Quantitative bleeding risk in AF patients on oral anticoagulation (DOAC or warfarin)
To refine anticoagulation decisions where HAS-BLED or ORBIT has been insufficient
When GDF-15, hs-cTnT, and haemoglobin are available from routine laboratory platforms
Used alongside the ABC-AF Stroke Score for net clinical benefit assessment in individual patients
ESC 2020 and 2024 guidelines support biomarker-enhanced bleeding risk assessment in AF
Patient Population
Derived and validated in patients with non-valvular AF on anticoagulation. Performs consistently across subgroups including age, sex, renal function, heart failure, diabetes, and prior bleeding. Version 2.0 (2026) incorporates OAC type for treatment-specific risk estimates.
When Not to Rely on This Score Alone
GDF-15 unavailable — the score cannot be calculated without it
Valvular AF or mechanical heart valves — score not derived in these populations
Active haemorrhage — this is a chronic risk stratification tool, not an acute triage instrument
Do not withhold anticoagulation from high-stroke-risk patients based on bleeding score alone without formal net benefit analysis
Section 2
Formula & Logic
Scoring Variables
Five inputs via Cox regression (ARISTOTLE, n = 14,537): age, prior bleeding history, GDF-15 (ng/L), hs-cTnT (ng/L), and haemoglobin (g/dL). The score is continuous — it produces a direct 1-year major bleeding risk estimate as a percentage. It is not point-based and requires a nomogram, Excel calculator, or the validated online tool at abc-score.com.
Biomarker Roles
01
GDF-15 (Growth Differentiation Factor-15): The dominant predictor. A TGF-β superfamily cytokine induced by oxidative stress and cardiomyocyte injury. Strongly and independently associated with major bleeding across all major DOAC trials. Likely reflects systemic frailty, vascular fragility, and subclinical end-organ damage rather than coagulation directly.
02
hs-cTnT: Reflects myocardial injury and cardiac fibrosis. Moderate independent predictor — likely capturing vascular senescence pathways that increase haemorrhagic risk.
03
Haemoglobin: Low haemoglobin signals both subclinical chronic blood loss and impaired haemostatic reserve. The strongest objective bedside indicator in the score.
Risk Thresholds
01
< 1%/year: Low risk — standard anticoagulation well-tolerated.
02
1–2%/year: Moderate risk — anticoagulation appropriate; address modifiable factors.
03
> 2%/year: High risk — anticoagulation may still be appropriate if stroke risk is proportionate; address all modifiable factors; consider LAAO if long-term OAC is not feasible.
Version 2.0 — OAC Type Incorporated
The 2026 update (COMBINE AF, n = 25,962) adds OAC type (DOAC vs warfarin) to correct a calibration issue: warfarin patients were slightly underestimated for risk; DOAC patients slightly overestimated. C-index 0.69 major bleeding, 0.72 GI bleeding, 0.66 intracranial bleeding — superior to HAS-BLED (0.61), DOAC score (0.65), and ORBIT (0.66) across all subgroups.
Section 3
Pearls/Pitfalls
Why This Outperforms HAS-BLED
HAS-BLED was designed in 2010 from clinical variables only. Its C-index for major bleeding in DOAC-era cohorts is consistently 0.58–0.64 — barely above chance. It also contains stroke risk factors (hypertension, prior stroke) as bleeding predictors, creating circular reasoning in the risk-benefit trade-off. The ABC-AF score captures pathophysiological signals independent of and additive to clinical factors, producing C-indices of 0.68–0.72 across diverse international populations.
The GDF-15 Barrier
GDF-15 is the dominant predictor but also the main adoption barrier. Measurable on Roche Cobas Elecsys platforms alongside hs-cTnT — but not yet part of routine AF workup in most settings outside Scandinavia and trial centres. Where GDF-15 is unavailable, ORBIT is the next-best clinical alternative — it outperforms HAS-BLED without biomarkers.
RCT Evidence — A Caution
The ABC-AF RCT (Oldgren et al., Circulation 2025, n = 3933) tested whether score-guided treatment tailoring reduced stroke or death vs. guideline-based care. Result: HR 1.19 (95% CI 0.96–1.48), p = 0.12 — no benefit. The study was prematurely terminated and underpowered. The null result reflects the difficulty of improving outcomes in a population already well-managed by guidelines, not a failure of the score's discrimination. The ABC-AF scores remain validated prognostic tools; their role in treatment decision support is still under study.
Modifiable Bleeding Risk Factors
Control blood pressure — uncontrolled hypertension is the most common modifiable risk, especially for intracranial haemorrhage
Stop aspirin unless there is a clear ongoing indication (ACS, recent stenting)
Avoid NSAIDs — strongly associated with GI bleeding in anticoagulated patients
Correct anaemia — low haemoglobin in the score may reflect treatable iron deficiency
Adjust DOAC dose for CKD — reduces accumulation-related bleeding without sacrificing efficacy
Address falls risk in elderly patients with high GDF-15 — intracranial bleeding from falls is the dominant concern
Section 4
Next Steps
Interpreting the Result
01
< 1%/year: Anticoagulation strongly favoured. No specific additional monitoring beyond standard AF care.
02
1–2%/year: Anticoagulation appropriate for most patients with CHA2DS2-VASc ≥ 2(M) or ≥ 3(F). Prefer DOAC — lower intracranial bleeding vs warfarin. Address modifiable factors.
03
> 2%/year: Anticoagulation still appropriate if stroke risk proportionate. Address all modifiable factors. Consider LAAO if contraindication to long-term OAC. Do not withhold OAC on bleeding score alone.
Net Clinical Benefit
The ABC-AF Stroke and Bleeding scores are designed to be used together. The abc-score.com calculator displays both simultaneously, allowing clinicians to visualise the individual trade-off — particularly valuable for patients in the 1–3%/year bleeding zone where the decision is genuinely balanced.
Complementary Tools
ABC-AF Stroke Score
HAS-BLED Score
ORBIT Score
CHA2DS2-VASc Score
Section 5
Evidence Appraisal
Original Derivation and Validation
The novel biomarker-based ABC (age, biomarkers, clinical history)-bleeding risk score for patients with atrial fibrillation: a derivation and validation study.
Hijazi Z et al. • Lancet.. 2016;387(10035):2302–2311. Derived n = 14,537 (ARISTOTLE); validated n = 8,468 (RE-LY). C-index 0.68 vs 0.61 HAS-BLED (p < 0.0001).
View SourceENGAGE AF-TIMI 48 Validation
Performance of the ABC scores for assessing the risk of stroke or systemic embolism and bleeding in patients with atrial fibrillation in ENGAGE AF-TIMI 48.
Berg DD et al. • Circulation.. 2019;139:760–771. Independent validation in the edoxaban trial cohort. Confirmed superior discrimination over HAS-BLED and ORBIT.
View SourceABC-AF Bleeding Score 2.0 — COMBINE AF
Evaluation of the updated ABC-AF-bleeding score 2.0 in patients with atrial fibrillation treated with a direct oral anticoagulant or warfarin.
Hijazi Z et al. • Journal of Thrombosis and Haemostasis.. 2026;24(2):399–407. n = 25,962 (COMBINE AF). OAC type incorporated. C-index 0.69 major bleeding, 0.72 GI, 0.66 intracranial. Superior to all comparator scores across all subgroups.
View SourceABC-AF RCT
Biomarker-Based ABC-AF Risk Scores for Personalized Treatment to Reduce Stroke or Death in Atrial Fibrillation.
Oldgren J et al. • Circulation.. 2025;152(21). n = 3933. Score-guided treatment vs standard care: HR 1.19 (95% CI 0.96–1.48), p = 0.12. No improvement in clinical outcomes. Study prematurely terminated.
View SourceClick to Read
Section 6
Literature
Dr. Ziad Hijazi and the Uppsala Group
Developed by Ziad Hijazi and Lars Wallentin at Uppsala Clinical Research Center, Sweden — the same group responsible for major DOAC trial biomarker research over two decades. The bleeding score emerged alongside a companion ABC-AF Stroke Score (2016), both derived from ARISTOTLE, creating the first integrated biomarker-based framework for simultaneous stroke and bleeding risk quantification in AF.
Score Development Timeline
01
2010: HAS-BLED published — first widely adopted clinical bleeding score in AF. C-index 0.61–0.64.
02
2014: Wallentin, Hijazi et al. demonstrate GDF-15 as a strong independent bleeding predictor in ARISTOTLE.
03
2016 (Lancet): ABC-AF Bleeding Score derived (n = 14,537) and validated (n = 8,468). Outperforms HAS-BLED and ORBIT.
04
2019 (Circulation): Independent validation in ENGAGE AF-TIMI 48.
05
2020: ESC AF Guidelines recommend biomarker-enhanced bleeding risk assessment; ABC-AF explicitly referenced.
06
2025 (Circulation): ABC-AF RCT — null result; study underpowered. Score discrimination validated; treatment tailoring utility under study.
07
2026 (J Thromb Haemost): ABC-AF Bleeding Score 2.0 published (COMBINE AF, n = 25,962). OAC type incorporated.
Last Comprehensive Review: 2026