Despite the impressive reduction of risk of stroke and death in patients with atrial fibrillation (AF) conferred by anticoagulation with warfarin or non–vitamin K antagonist oral anticoagulants,1,2 there is still an appreciable stroke risk during anticoagulant treatment, approximating 1.7% per year for warfarin and 1.4% per year for non–vitamin K antagonist oral anticoagulants at 2.2 years’ follow-up.2 This residual stroke rate is often regarded as treatment failure, but, to our knowledge, it has not been compared with a matched control population and could instead reflect the stroke rate in people of a comparable baseline risk without AF.
We reanalyzed data from a cohort study of 5555 patients with incidentally detected ambulatory AF3: 2492 (44.9%) received warfarin with (817 of 2492 [32.8%]) or without (1675 of 2492 [67.2%]) an antiplatelet and 1603 (28.8%) received an antiplatelet alone; 1460 (26.3%) received no antithrombotic. The study was approved by the Independent Scientific Advisory Committee for Clinical Practice Research Datalink research; no further ethics approval was required for the analysis of the data.
These cohorts were compared with 24 705 age- and sex-matched control patients without AF seen in general practice on the same day. The patient and control cohorts were derived between January 1, 2001, and March 31, 2009, from the UK Clinical Practice Research Datalink,3 a fully annotated general practice database including prescriptions linked to hospital discharges and deaths. Full methods and baseline characteristics have been published.3 In the group with incidentally detected ambulatory AF, 3422 (61.6%) were men; mean (SD) age was 70.9 (10.1) years. In the control group, 15 123 (61.2%) were men; mean age was also 70.9 (10.1) years.
Cumulative stroke incidence estimates were derived from competing risk regression analysis (death as a competing risk),4 using multivariable models adjusted for comorbidities, standardizing to the prevalence of all CHA2DS2VASc elements, smoking, previous major bleeding events, cancer, prior aspirin use, and Charlson Comorbidity Index score in the warfarin-treated cohort.5 Because the proportional hazards and subdistribution hazards assumptions had been fulfilled for only 1.5 years, regression analysis was limited to this timespan. Crude and adjusted subdistribution hazard ratios were calculated for stroke and death (Table). Data analysis was also conducted between January 1, 2001, and March 31, 2009.
At 1.5 years after AF diagnosis (Figure, A), the adjusted cumulative stroke incidence was 3.9% in untreated patients, 1.3% in those receiving warfarin, and 1.2% in the controls without AF (P = .75). There was no significant difference when only nonfatal strokes were considered (Figure, B and Table). The adjusted cumulative mortality at 1.5 years was 7.2% in patients with untreated, incidentally detected ambulatory AF; 4.2% in those receiving warfarin; and 2.5% in the matched controls (Figure, C). Although absolute mortality reduction was similar to that for stroke, residual risk in the treatment group was significantly greater than the risk in the control group, even after adjustment for baseline differences (P = .005). Antiplatelet therapy did not reduce the incidence of either stroke or death.3
The results of this study are consistent with the hypothesis that residual stroke risk in anticoagulant-treated patients with AF is unlikely predominantly due to treatment failure for cardioembolic stroke. Rather, the findings suggest that the risk may be related to noncardioembolic stroke that may be expected in individuals of similar age, sex, and other stroke risk factors but without AF. Because unknown AF has been detected in 3% of the population aged 75 to 76 years after 2 weeks of intermittent handheld electrocardiographic recordings,6 it is possible that unknown AF was present in 3% of the controls. When we recalculated the control stroke risk, assuming 3% undiagnosed AF, findings were unchanged. Because this was an observational cohort study, residual confounding could have played a role in our results showing no difference in stroke between patients receiving anticoagulant treatment for AF and controls.
The findings for mortality differed from those for stroke. Anticoagulation only partially reduced the adjusted excess mortality compared with age- and sex-matched controls. Other conditions that may result from AF include heart failure and myocardial infarction, but only myocardial infarction would likely be reduced by warfarin therapy.3 Atrial fibrillation may also be a marker for other conditions that predispose to death but are not influenced by anticoagulant therapy.
We conclude that residual stroke risk in anticoagulant-treated patients with AF is more likely to reflect the population-associated risk of noncardioembolic strokes in patients of similar age, sex, and comorbidity without AF rather than anticoagulant treatment failure. However, this finding is not true for death: AF incurs a direct or associated mortality risk that is only partially reduced by anticoagulant therapy.
Corresponding Author: Ben Freedman, MB, PhD, Heart Research Institute, Level 3E D17, Charles Perkins Centre, University of Sydney, Sydney NSW 2006, Australia (email@example.com).
Published Online: May 4, 2016. doi:10.1001/jamacardio.2016.0393.
Author Contributions: Dr Martinez had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Freedman, Martinez.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Freedman, Martinez.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Martinez, Katholing, Rietbrock.
Administrative, technical, or material support: Martinez.
Study supervision: Martinez.
Conflict of Interest Disclosures: The authors have completed and submitted the ICJME Form for Disclosure of Potential Conflicts of Interest. Dr Freedman reported receiving research grants to conduct investigator-initiated studies from Bayer Pharma, BMS/Pfizer, and Boehringer-Ingelheim, manufacturers of 3 different non–vitamin K antagonist oral anticoagulant drugs used in atrial fibrillation; personal fees from Bayer Pharma and BMS/Pfizer as well as personal fees from AstraZeneca, Gilead, and Servier related to drugs that are not used in atrial fibrillation; and nonfinancial support from Bayer Pharma and Boehringer-Ingelheim. Dr Martinez reported receiving grants from Bayer Pharma AG and BMS/Pfizer and personal fees from Boehringer Ingelheim outside the submitted work. Ms Katholing and Dr Rietbrock reported receiving grants from Bayer Pharma AG and grants from BMS/Pfizer outside the submitted work. No other disclosures were reported.
MI. Meta-analysis: antithrombotic therapy to prevent stroke in patients who have nonvalvular atrial fibrillation. Ann Intern Med
. 2007;146(12):857-867.PubMedGoogle ScholarCrossref
et al. Comparison of the efficacy and safety of new oral anticoagulants with warfarin in patients with atrial fibrillation: a meta-analysis of randomised trials. Lancet
. 2014;383(9921):955-962.PubMedGoogle ScholarCrossref
SB. Adverse prognosis of incidentally detected ambulatory atrial fibrillation: a cohort study. Thromb Haemost
. 2014;112(2):276-286.PubMedGoogle ScholarCrossref
RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc
. 1999;94(446):496-509.Google ScholarCrossref
PW. Adaptation and validation of the Charlson Index for Read/OXMIS coded databases. BMC Fam Pract
. 2010;11:1.PubMedGoogle ScholarCrossref
M. Mass screening for untreated atrial fibrillation: the STROKESTOP Study. Circulation
. 2015;131(25):2176-2184.PubMedGoogle ScholarCrossref