Can Simple Biomarkers Improve Stroke Prediction in AF?

John Mandrola, MD


October 14, 2016

This is 2016. It is past time. We need a more precise way to predict stroke in patients with AF.

Seemingly easy questions like, "Do I need a blood thinner?" or "Can I stop my blood thinner?" are as hard to answer now as they were a decade ago.

That's why I am pleased to tell you about a nifty study published recently in JAMA Cardiology.[1]

Dr Christian Ruff (Harvard University, Boston) and colleagues set out to develop and test a biomarker score for patients with AF. In a prespecified analysis of the edoxaban-vs-warfarin ENGAGE-AF-TIMI 48 trial, they studied a group of nearly 5000 patients who had three biomarkers—troponin I, N-terminal pro-B-type natriuretic peptide (NT-proBNP), and d-dimer—measured at the time of randomization.

The background for this study centers on the three big problems with ceding anticoagulation decisions to the CHA2DS2-VASc score.

The first problem is the statistics of stroke. Namely, the most likely outcome for patients with AF whether they take an anticoagulant or not is nothing—neither stroke nor bleeding. (Without treatment, a patient with a CHA2DS2-VASc score of 3 has a greater than 96% chance of not having a stroke in the next year.)

The second problem with the CHA2DS2-VASc is its poor predictive ability. The C statistic, a measure of predictive accuracy, is approximately 0.6[2]—with 0.5 being a pure play of chance.

The third problem is the magnitude of committing someone to take a potent anticoagulant drug for the next 2 or 3 decades. A decision like this demands more precision. Does a healthy 65-year-old woman with a few nonbothersome short-lived episodes of AF really need to take an anticoagulant for the rest of her life?

For the present study, Ruff and colleagues stratified the three everyday biomarkers into quartiles (troponin into a quintile) and assigned each a value based on the nearest whole number of the associated hazard ratio (adjusted for the CHA2DS2-VASc). The three values were totaled to form an aggregate biomarker score from 0 to 11. To test the hypothesis that a biomarker score would outperform CHA2DS2-VASc, they measured the predictive performance of each biomarker and the composite biomarker score using the C statistic.


Within each biomarker group, they found a statistically significant graded association between increasing biomarker levels and the rates of stroke, systemic embolism (SEE), or death.

Using the biomarker score, the rate of stroke, SEE, or death ranged from 1.2% per year for patients with a 0 score to 21% per year in patients with a score of 10 to 11. Using the CHA2DS2-VASc score, the range was much narrower, at 2.2% to 9.9%. After adjustment for CHA2DS 2-VASc, the biomarker score still identified more than a 15-fold gradient of risk.

Within each CHA2DS2-VASc category, investigators noted a broad gradient of risk for events. For instance, in patients with a CHA2DS2-VASc of 3, the risk of events varied from 2% with low (0–4) biomarker scores to more than 12% in patients with high (8–11) biomarker scores.

The biomarker risk score alone had significantly greater prognostic accuracy than the CHA2DS2-VASc score, with a C statistic of 0.70 (95% CI 0.68–0.72) compared with 0.59 (95% CI, 0.57-0.61) for the CHA2DS2-VASc score. (P<0.001).


These are exciting data. The business of predicting the future will always be tough, but these results suggest three simple blood tests might help doctors and patients make tough decisions.

Everyday clinicians sense that there are, within CHA2DS2-VASc category, degrees of unhealthiness. For instance, smoking, obesity, history of venous thrombosis, left ventricular hypertrophy, poor exercise tolerance, and stress may be important stroke risk factors not considered in the CHA2DS2-VASc.

The notion of adding biomarkers to clinical scores in patients with AF is not new. In 2006, Lip et al[3] noted that the addition of plasma von Willebrand factor refined stroke prediction. More recently, a group of investigators (Hijazi et al[4]) developed and validated a novel biomarker-based risk score for predicting stroke among patients with AF enrolled in the ARISTOTLE and STABILITY trials. Their ABC (age, biomarkers, clinical history) index using NT-proBNP and high-sensitivity cardiac troponin assays performed significantly better than the CHA2DS2-VASc for predicting embolic events.

Of course, adding more and more biomarkers will improve stroke prediction. But simplicity and pragmatism at the bedside remain critical. The CHA2DS2-VASc score works because it's easy to use. And medicine has decided it gets us close enough, especially with very low-risk patients. Fine, that was then, but in this era of big data and applied math, we can do better.

Biomarkers like BNP, troponin, and d-dimer are easy to measure. A biomarker score could be done in outpatients with AF and reported like a glomerular filtration rate. Wouldn't it be valuable to know whether one's stroke risk is 1.2% per year or 21%?

A couple of caveats about the paper: first, ENGAGE-AF-TIMI-48 did not enroll patients with CHA2DS2-VASc scores of 0 or 1. We need to apply the biomarker idea to this group. I've recently seen two AF patients (CHA2DS2-VASc 0–1) who suffered massive embolic strokes. There must be something about these patients that could have warned us of their upcoming stroke.

I asked Dr Ruff about biomarker variability. In an email, he explained that most of the patients in the substudy were stable outpatients and most were in AF at the time of measurement. This brings up the main challenge of biomarker use in clinical practice: one has to view biomarkers in clinical context. Patients admitted for flares of diastolic heart failure, pneumonia, or postoperative AF will have transient alterations of biomarkers. It's akin to measuring mitral regurgitation—it always looks worse during a flare of heart failure. Use of biomarkers will still require a human brain.

Finally, to the credit of the authors, they make it clear that these data are proof of concept. They say the biomarker score needs "validation and refinement in additional studies that should include cohorts not receiving anticoagulants outside of the highly selected clinical-trial population."

If the patients I recently saw with AF-related embolic stroke and low CHA2DS2-VASc scores had had elevated biomarkers, I might have been able to tell them to take anticoagulants before the event. Likewise, a person with intermittent AF and one risk factor but all normal biomarkers may be able to avoid lifelong anticoagulants.

Offering our patients more precision on future stroke risk would be a huge step forward for modern cardiology. There's lots of work to do, but these are encouraging signals that are surely worth more research effort.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: