Abstract 9885: Overestimation of Anticoagulant Benefit in Patients with Atrial Fibrillation and Low Life Expectancy: Evidence from 12 Randomized Trials

Abstract

Introduction: Formative anticoagulant RCTs for stroke prevention in atrial fibrillation (AF) were completed before the advent of competing risk methods. Neglecting the competing risk of death can overestimate treatment benefit. This is noteworthy because many patients with AF are frail with limited life expectancy. We determined the overestimation of anticoagulant benefit as a function of life expectancy. Methods: We used patient-level data from 12 anticoagulant RCTs that compared warfarin vs. aspirin or placebo. We estimated each patient’s life expectancy using CDC life tables modified for age, sex, comorbidities, and trial enrollment year. We predicted cumulative stroke incidence with and without anticoagulants at 3 years using a Cox model treating death as a censoring event (“standard model”) and again using a Fine-Gray model treating death as a competing event (“competing risk model”). For each patient, we plot the absolute and relative standard model overestimation compared to the competing risk model. We describe the relationship between life expectancy and log-transformed benefit overestimation (Figure A, B). Results: For the 9548 patients (35% women, mean age 74yrs), the median life expectancy was 11.4yrs (IQR 8.4, 15.3). The overall estimated reduction in stroke incidence at 3 years was 7.7% using a standard model and 6.3% using a competing risk model. The absolute and relative overestimation of the standard model increased exponentially as life expectancy decreased. For example, with a life expectancy of 3 years, a standard model would overestimate absolute benefit by 2.8% (95%CI 2.7-2.9%) and the relative benefit by 29.4% (95%CI 29.3-29.4%). Conclusions: For patients with low life expectancy, the anticoagulant benefit is substantially overestimated using standard survival methods. While anticoagulants reduce the risk of stroke, decision tools should account for life expectancy when estimating benefits for patients with AF.