Abstract & Commentary
Improving Risk Prediction in Atrial Fibrillation Patients on Anticoagulants
By Michael H. Crawford, MD, Editor
Source: Hijazi Z, et al. High-sensitivity troponin T and risk stratification in patients with atrial fibrillation during treatment with apixaban or warfarin. J Am Coll Cardiol 2014;63:52-61.
Accurate risk prediction in atrial fibrillation (AF) patients is important for the decision to prescribe oral anticoagulants. High-sensitivity troponin T levels (hs-TnT) have augmented prognostic predictions in coronary heart disease and heart failure. These investigators from the Apixaban for the Prevention of Stroke in Subjects with Atrial Fibrillation (ARISTOTLE) trial assessed the potential additive value of hs-TnT in predicting outcomes in patients with AF in this study of apixaban compared to warfarin. ARISTOTLE randomized more than 18,000 patients with AF and at least one CHADS2 risk factor for stroke to apixaban or warfarin. Study endpoints included stroke systemic embolism, mortality, myocardial infarction (MI), and major bleeding; average follow-up was about 2 years. Outcomes were evaluated in relationship to hs-TnT levels, and the CHADS2 and CHADS2-VASc (CH2V) scores. Measurable hs-TnT levels were found in 94% of the patients. With the exception of hypertension, all the CH2V risk factors were associated with higher hs-TnT levels. Overall, the rate of stroke or embolism was 1.4%/year, all-cause mortality 3.7%/year, cardiac deaths 1.9%/year, MI 0.5%/year, and major bleeding 2.6%/year. There was no difference in these events in relation to hs-TnT levels in the two treatment groups, so they were combined for this analysis. Higher hs-TnT levels were significantly related to outcome events, even when adjusted for baseline characteristics. A similar relationship was observed for the CH2V score. The predictive ability of the CH2V score was increased by 23% when the hs-TnT was added. However, stroke was predicted better by CH2V, and cardiac death and major bleeding by hs-TnT. The authors concluded that hs-TnT is almost always elevated in patients with AF, is independently associated with adverse events, and improves the risk prediction for cardiovascular (CV) events as compared to the CH2V score.
This study derived from the ARISTOTLE trial of apixaban vs warfarin in non-valvular AF patients makes several interesting observations. In these patients, all of whom were treated with an effective oral anticoagulant, mortality and major bleeding were more common than stroke and MI. When CH2V and troponin levels were assessed for their predictive ability, hs-TnT was better at predicting cardiac mortality, MI, and major bleeding, and CH2V was better at predicting stroke. This is not surprising since CH2V was designed to assess the risk of stroke in AF patients to guide preventive therapy. However, CV events are frequent in this population and hs-TnT does a better job of predicting these outcomes. In fact, in patients with low CH2V scores (0-2), hs-TnT culls out patients at higher risk of CV events and major bleeding on anticoagulants. Since troponin testing is widely available, this is an attractive option to further risk stratify AF patients on oral anticoagulants (OACs).
Unfortunately since all the patients in this study were on OACs, the results cannot be used to make decisions regarding starting anticoagulation therapy. Whether troponin testing would improve this decision needs to be tested. However, the ability of hs-TnT to improve the prediction of who is at risk for major bleeding could alert one to those patients who need to be followed more carefully. Whether troponin is better than the HASBLED scoring system for bleeding risk was not tested in this study, but other studies have shown that higher CH2V scores also predict bleeding risk about as well as HASBLED does, and troponin was additive to CH2V for predicting major bleeding. Also, the results of this study are consistent with those of a similar analysis of the RELY trial and if you combine both studies, the population tested is more than 21,000 patients.
Mortality was the most common endpoint observed in these patients on OACs for AF and it was best predicted by troponin levels. This implies that most AF patients on OACs die of cardiac causes rather than systemic embolic events. This makes sense since many patients with non-valvular AF have risk factors for atherosclerotic CV disease, such as hypertension. It also raises the question as to whether other markers of CV risk, such as brain natriuretic peptide or hs-CRP, would be useful predictors in this population. The difficulty in using hs-TnT to risk stratify patients in this trial is that almost all of them had levels above normal. So, you would need to compare your patients' level with the data in the paper to make a risk prediction. In addition, the cutoff values may vary with different assays for hs-TnT.