Moving the tipping point: the decision to anticoagulate patients with atrial fibrillation

Abstract

Background: The rate of ischemic stroke associated with traditional risk factors for patients with atrial fibrillation has declined over the past 2 decades. Furthermore, new and potentially safer anticoagulants are on the horizon. Thus, the balance between risk factors for stroke and benefit of anticoagulation may be shifting.

Methods and results: The Markov state transition decision model was used to analyze the CHADS(2) score, above which anticoagulation is preferred, first using the stroke rate predicted for the CHADS(2) derivation cohort, and then using the stroke rate from the more contemporary AnTicoagulation and Risk Factors In Atrial Fibrillation cohort for any CHADS(2) score. The base case was a 69-year-old man with atrial fibrillation. Interventions included oral anticoagulant therapy with warfarin or a hypothetical "new and safer" anticoagulant (based on dabigatran), no antithrombotic therapy, or aspirin. Warfarin is preferred above a stroke rate of 1.7% per year, whereas aspirin is preferred at lower rates of stroke. Anticoagulation with warfarin is preferred even for a score of 0 using the higher rates of the older CHADS(2) derivation cohort. Using more contemporary and lower estimates of stroke risk raises the threshold for use of warfarin to a CHADS(2) score ≥2. However, anticoagulation with a "new, safer" agent, modeled on the results of the Randomized Evaluation of Long-Term Anticoagulation Therapy trial of dabigatran, leads to a lowering of the threshold for anticoagulation to a stroke rate of 0.9% per year.

Conclusions: Use of a more contemporary estimate of stroke risk shifts the "tipping point," such that anticoagulation is preferred at a higher CHADS(2) score, reducing the number of patients for whom anticoagulation is recommended. The introduction of "new, safer" agents, however, would shift the tipping point in the opposite direction.

Figure 1. One-way sensitivity analysis: Annual rate of ischemic stroke

Quality-Adjusted Life Expectancy for each of the 3 strategies (Warfarin, Aspirin, and No Antithrombotic Therapy) is shown as a function of the annual rate of ischemic stroke ranging from 0 to 0.15/year. There are two secondary horizontal axes showing the corresponding CHADS₂ scores. The upper secondary axis uses the CHADS₂ derivation cohort (see appendix table 2), while the lower axis maps the CHADS₂ predictors to the annual stroke rate found in the more contemporary ATRIA cohort. The threshold lines divides the decision space into three regions. To the far left, at low rates of ischemic stroke (< 0.2%/year), no antithrombotic therapy is best, while to the far right at stroke rates greater than 1.7%/year, anticoagulation with warfarin is best. There is a small region between these two thresholds in which aspirin use is preferred. Using more contemporary data fr stroke risk (bottommost horizontal axis), anticoagulation is only preferred at a higher CHADS₂ score (≥ 2) compared with stroke risk predicted by the CHADS₂ derivation model (top secondary horizontal axis), for which warfarin is preferred even with a CHADS₂ score less than 0.

Figure 2. One-way sensitivity analysis: Annual rate of ischemic stroke with addition of anticoagulation with a new, “safer” anticoagulant

The axes are the same as in Figure 1. With the addition of a new, “safer” agent as another option for anticoagulation, the “tipping point” above which the risk and outcomes of ischemic stroke outweigh the risk and outcomes of major hemorrhage shifts to the left. Anticoagulation with the new drug is preferred at annual stroke rates above 0.9%/year (CHADS₂ score < 0 in the derivation model; and CHADS2 score of 1 or greater using the ATRIA data).

Figure 3. Three-way sensitivity analysis: Relative hazard of intracerebral hemorrhage and relative hazard of ischemic stroke (new, “safer” anticoagulant vs warfarin), and quality of life

The decision space is divided into two regions. At the lower left, where the relative hazards of intracerebral hemorrhage (ICH) and ischemic stroke while receiving the new, “safer” anticoagulant vs warfarin are low, the new agent is preferred. At the upper right, where the relative hazards of these events is high, warfarin is preferred. Three threshold lines are shown for varying quality of life while taking the new anticoagulant (0.98, 0.99, and 1.0). If the new anticoagulant has no detrimental impact on quality of life (eg quality of life of 1.0), then the region in which it is favored is largest. As the quality of life while taking the new anticoagulant decreases, the size of this region becomes smaller. The ellipse demonstrating the base case values for the relative hazards of ICH and ischemic stroke, along with their 95% confidence intervals, falls within the region in which the new anticoagulant is preferred if the quality of life while taking this new agent is 0.99 or greater. Hypothetical new and safer agents characterized by higher efficacy and lower bleeding risk (lower left corner) would be preferred over warfarin.