Benefit of Warfarin in Older Persons with Atrial Fibrillation

Atrial fibrillation (AF) is a potent risk factor for ischemic stroke, and anticoagulation therapy significantly mitigates this risk. Conspicuously missing from an understanding of this risk reduction in older adults is the effect of the “competing risk” of death not related to stroke. Because AF is a condition whose prevalence rises with increasing age, deaths from conditions unrelated to stroke are, unfortunately, not uncommon. Individuals with AF who are not treated with anticoagulation are older and more often frail than those who are treated.1 Consequently, estimates of the long-term effect of anticoagulation on risk of stroke may be substantially biased if the competing risk of nonstroke-related death is not considered. To address this gap in knowledge, and using data from the Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA) study,2 Ashburner and colleagues conducted an analysis that provides important insight into the management of AF in older adults.3 Ashburner and colleagues provide a clear demonstration of how warfarin's abatement of the risk of stroke is strongly diminished as people grow older and increasingly die of nonthromboembolic causes.3 In multivariable analysis with a traditional cause-specific approach that censors (ignores nonstroke death), the use of warfarin is associated with a 43% lower cause-specific hazard of stroke over a 7-year period. In contrast, the same multivariable model using an approach that incorporates the competing risk of nonstroke death4 shows only a 13% lower subdistribution hazard. This suggests that, for this study population and other populations with similar rates of nonstroke death, covariates, and stroke, increasing levels of nonstroke death appear to greatly reduce, but not eliminate, warfarin's protective association.5 A supplemental analysis of the same ATRIA data that exhibits a similar reduction in warfarin's protective association coinciding with rising age further supports these results. An early meta-analysis showed a risk reduction of 68% for use of warfarin,6 and the original ATRIA results reported a 64% reduction in the odds of stroke over a 3-year period.7 Those results, which did not adjust for the competing risk of death, are consistent with the 43% reduction in the hazard of stroke over a 7-year period reported in the cause-specific model of Ashburner and colleagues.3 To the knowledge of the authors of this editorial, this is the first documented case in which an evaluation of warfarin's effectiveness in preventing stroke has included a rigorous examination of the competing risk of nonstroke death. For this reason, the shrinkage of warfarin's protective association that accompanies older age and greater likelihood of death is novel and informative. There is general consensus that use of warfarin is effective in preventing stroke among older adults with AF.8 At the same time, there is research suggesting that, even in individuals at high risk of stroke, oral anticoagulants are widely underused.9 A plausible explanation for such underuse appears to be that long-term practitioners prioritize concern about the risk of bleeding over stroke prevention in some older adults with AF.10 There is also agreement that use of anticoagulants such as warfarin is most effective in individuals with high risk of stroke, whereas treatment with aspirin appears to suffice for persons at lower risk.11 The marked attenuation of the protective effect of warfarin reported here provides a new perspective on the decision-making process that providers and older adults must navigate. In addition to the older adult's individual likelihood of stroke, the decision to prescribe anticoagulants must also consider chronic conditions and the overall risk of death from nonstroke causes.12 Ashburner and colleagues are to be congratulated for presenting this timely and meaningful examination of the competing risk of nonstroke death. Their study elegantly combines the use of the highly regarded subdistribution approach with a clever subsetting of the longitudinal data. These techniques clearly demonstrate how older age and mortality concur with a discounting of the protective effect of warfarin in older adults with AF. Because the use of warfarin has been associated with nearly double the hazard of intracranial hemorrhage,7 concern about bleeding that prevents many long-time practitioners from prescribing it to older adults with AF is not unfounded.10 These results reinforce the growing understanding that, when treating older adults, providers can no longer consider any disease or condition in isolation from the individual's complete profile of health concerns.13 Conflict of Interest: The authors declare no competing interests. Author's Contributions: Both authors contributed to the drafting and writing of the editorial. Sponsor's Role: None.