Biomarker-Based Risk Score Better Predicts Stroke in AF Than Clinical Tools

A risk-assessment tool incorporating NT-proBNP and high-sensitivity troponin performs better than clinical scores for predicting stroke among patients with atrial fibrillation (AF) who are taking oral anticoagulation, researchers report.

The ABC-AF-stroke risk score (for total stroke) and the ABC-AF-istroke risk score (for ischemic stroke) had improved discrimination when compared with either the ATRIA or CHA₂DS₂-VASc scores in a cohort that included patients from three pivotal direct oral anticoagulant (DOAC) trials, according to investigators led by Lars Wallentin, MD, PhD (Uppsala University, Sweden).

The findings, published in the March 25, 2025, issue of the Journal of the American College of Cardiology, suggest the ABC-AF scores could be used to help identify patients with a persistently high risk of stroke despite anticoagulation therapy, they indicate.

“We know that today a patient treated with a direct oral anticoagulant will still have a risk of stroke, and this risk of stroke is maybe around 0.3% up to 0.9% per year,” Wallentin told TCTMD. “There is still variability in the risk of stroke during oral anticoagulant treatment, so the real vision for the future was that if we had patients treated with a direct oral anticoagulant that still had a high risk of stroke, these patients might benefit from additional treatment like left atrial appendage occlusion [LAAO] devices, or maybe you have a more liberal indication for ablation in order to eliminate A-fib or reduce the burden of A-fib as a risk indicator for stroke.”

Elsayed Soliman, MD (Wake Forest University School of Medicine, Winston-Salem, NC), said “the study offers additional evidence supporting the potential role of biomarkers in improving stroke risk stratification for patients with AF.”

However, the improvement in discrimination observed with the ABC-AF score over the others “is unlikely to translate into meaningful clinical utility,” he said in an email. “Furthermore, the higher costs associated with incorporating biomarkers and the complexity of calculating the ABC-AF-stroke score make it an impractical replacement for the CHA₂DS₂-VASc score in routine clinical risk assessment, at least for now.”

Refining Stroke Risk Prediction

In recent years, research has shown that assessing levels of biomarkers like NT-proBNP and high-sensitivity troponin enhances prediction of the risk of stroke or systemic embolic events (SEE). These findings supported the creation of the ABC-AF risk score—incorporating age, biomarkers, and clinical history of stroke—based on data from the ARISTOTLE trial; it was later validated using data from the RE-LY and ENGAGE AF-TIMI 48 trials.

“In all three cohorts treated with oral anticoagulation, the ABC-AF risk score provided superior discrimination and better calibration of stroke/systemic embolic events [SEE] than risk scores based only on clinical factors,” Wallentin et al write, noting that a similar improvement in discrimination has been observed in patients with AF who were not receiving oral anticoagulation.

In the current analysis, the investigators updated the ABC-AF risk score to focus specifically on ischemic rather than total strokes, creating the ABC-AF-istroke risk score. “When DOAC is the preferred treatment for prevention of stroke in AF, it seems preferable that risk scores for decision support are based on estimates of the risk of ischemic stroke, as the risk of hemorrhagic stroke is substantially lower during DOAC treatment,” they explain.

Using the COMBINE-AF database, Wallentin et al examined 26,452 patients (median age 71 years; 37% women) who were randomized to standard-dose DOAC therapy or warfarin as part of ARISTOTLE, ENGAGE AF-TIMI 48, or RE-LY and had biomarker measurements available.

The median ATRIA score was 6, and the median CHA₂DS₂-VASc score was 4. The median ABC-AF-stroke risk score indicated a risk of total stroke/SEE of 1.2% per year, whereas the median ABC-AF-istroke score indicated a risk of ischemic stroke/SEE of 0.9% per year.

During a median follow-up of 25 months, there were 756 cases of stroke/SEE, including 534 cases of ischemic stroke/SEE.

The ABC-AF-stroke score had significantly better discrimination for total stroke compared with either ATRIA or CHA₂DS₂-VASc (C-index 0.667 vs 0.632 and 0.614; P < 0.001 for both comparisons). Findings were similar for the ABC-AF-istroke stroke and risk of ischemic stroke (C-index 0.677 vs 0.642 and 0.624; P < 0.001 for both comparisons).

Results were consistent across various subgroups.

Moreover, the ABC-AF scores “provided added benefit concerning individually tailored decisions on stroke-prevention treatment in patients with AF treated with oral anticoagulation with a residual risk of ischemic stroke in the clinically relevant ranges between 0.5% and 3% per year for the ABC-AF-istroke score and 0.5% and 5% per year for the ABC-AF-stroke score,” the researchers write.

Where Does ABC-AF Fit In?

In an accompanying editorial, William McIntyre, MD, PhD, and Aristeidis Katsanos, MD, PhD (Population Health Research Institute, Hamilton, Canada), ponder how the ABC-AF scores could be most useful when managing patients with AF, stating that they would not be helpful at the low end of the stroke risk spectrum when there is a question about the need for oral anticoagulation.

“In the middle of the risk spectrum, the ABC-AF score is again likely to lose out to CHA₂DS₂-VASc because risk-based decisions with the incumbent score are not in doubt and the new model requires biomarkers and a nomogram or an online calculator to execute,” they write.

Where there is potential utility, however, is among patients with the highest risk of stroke, who are already candidates for oral anticoagulation, McIntyre and Katsanos indicate, noting that ongoing trials are examining the impact of additional interventions like LAAO and placement of permanent bilateral carotid filters.

“If these therapies show efficacy for further reducing stroke,” they write, “the ABC-AF model could serve as a practical guide for identifying oral anticoagulation-treated AF patients who continue to have a high risk of stroke and who could be candidates for combined pharmacological/mechanical stroke prevention.”

Still, they say, ABC-AF is not likely to replace CHA₂DS₂-VASc for risk stratification in everyday practice.

Wallentin agreed that “for the routine question—anticoagulate or not—I don’t think you really need the score. You can use it, but you don’t necessarily need it for that simple question today because of the safety and the very positive benefit-risk profile of drugs like apixaban and edoxaban.”

In addition to perhaps guiding use of additional interventions on top of oral anticoagulation, the ABC-AF score could be useful in the future for tailoring use of additional medical therapies like the factor XIa inhibitors, which will be more expensive, Wallentin suggested.