Female gender as a risk factor for stroke associated with atrial fibrillation

This editorial refers to ‘Revisiting sex differences in outcomes in non-valvular atrial fibrillation: a population-based cohort study’†, by C. Renoux et al., on page 1473. Several epidemiological studies have identified gender-related differences in the incidence of stroke attributable to atrial fibrillation (AF), with a relative increase ranging from 20% to 70% across the studies, even after correction for age, risk factors, co-morbidities, and the use of oral anticoagulation (OAC; Table 1). Furthermore, the effect of AF per se on risk of stroke was 2- to 4.5-fold greater in women compared with men, after adjustment for other major risk factors,1 , 2 and AF-related strokes in women were twice as likely to result in disabling or fatal consequences.3 Meta-analysis of 17 reports which included 5 randomized controlled trials and 12 prospective observational studies has shown a 1.3-fold greater risk of AF-related stroke in women, which has not been modified by anticoagulation therapy.4 This has led to the inclusion of female gender as a component of the risk stratification schemes, such as CHA2DS2VASc (‘Sc’ stands for the ‘sex category’ and is assigned 1 point) which has been adopted by the majority of guidelines on stroke prevention in AF.5 Female gender as risk factor for stroke AF, atrial fibrillation; HR, hazard ratio; NVAF, non-valvular atrial fibrillation; OAC, oral anticoagulation; OR, odds ratio; RCT, randomized controlled trial; RR, relative risk; TIA, transient ischaemic attack. Female gender as risk factor for stroke AF, atrial fibrillation; HR, hazard ratio; NVAF, non-valvular atrial fibrillation; OAC, oral anticoagulation; OR, odds ratio; RCT, randomized controlled trial; RR, relative risk; TIA, transient ischaemic attack. The exact mechanism by which gender may affect predisposition to stroke has not been clearly understood, albeit that several possible hypotheses, such as genetics including factors that regulate the response to anticoagulation therapy, vascular physiology, and endothelial function, hormonal influences, modulating the expression of tissue factors and biomarkers involved in the prothrombotic processes, factors promoting hypercoagulability, and possibly less optimal treatment of underlying cardiovascular disease, have been suggested. Hormone replacement therapy may increase risk of stroke seen in women with AF. Gender-specific interactions between AF and other risk factors, such as hypertension and renal dysfunction, are possible, but no direct evidence supporting this hypothesis exists. However, higher rates of thrombo-embolic events in women have been chiefly reported in earlier cohorts such as the Atrial Fibrillation Investigators analysis of five studies, Framingham Heart Study, and ATRIA (AnTicoagulation and Risk factors In Atrial fibrillation), which were conducted in the era of a relatively low awareness of AF and, consequently, less aggressive management of the arrhythmia and associated risk factors, including OAC, which was probably likely in women.6–8 The recognition of AF as a clinically relevant entity and improvements in treatment of underlying heart disease may have contributed to the elimination of gender differences in outcome, specifically stroke. The historic surveys generally included smaller overall numbers of patients, and of women in particular (less than one-third in some surveys), and were possibly ‘biased’ to select either more symptomatic or ‘sicker’ women who were likely to have a high baseline risk. The contemporary analyses, particularly those based on the nationwide data registers involving hundreds of thousands of patients, are largely devoid of these limitations and are better positioned to overcome the random effect of gender on stroke rates, but even these have reported different results with respect to the gender effect on stroke risk9 – 12 (Table 1). Some reports highlighted an as yet unexplained paradox of lower all-cause and cardiovascular death rates in women, despite a higher likelihood of stroke.2 While younger women (under 65 years of age) in fact had a lower risk of stroke compared with male counterparts, the sharp increase in the rates of stroke in women at older age (an arbitrary threshold is 75 years), being at 10–20% higher than in men of similar age, which were not affected by the use of warfarin, suggesting that the role of female gender as a risk factor for stroke is closely linked to age and possibly other major risk factors, but in younger women it does not have a sufficient power to trigger anticoagulation.11 , 12 There has been no evidence that women may be less compliant with treatment since the rates of filling the prescription for anticoagulants appear to be similar between genders. A meta-analysis has also found that risk of stroke was the highest in women aged ≥75 years.4 The reasons for the differential effect of gender on stroke rates in older patients are multifactorial, and one of the recent developments was the finding of increased amyloid production which is linked to oestradiol metabolism and the greater prevalence of amyloid atrial cardiomyopathy in women, particularly of older age, reflecting the gender-related arrhythmodenic substrate which may also promote the thrombogenic milieu. Isolated atrial amyloidosis is a common finding in older individuals, found in >90% of left atrial tissue samples in octogenarians, whereas amyloid deposits are facilitated by AF duration and female gender.15 The pathophysiological mechanism involves increased expression of atrial natriuretic peptide (ANP) by stimulation of oestrogen receptors in the presence of 17β-oestradiol, with persistently elevated ANP levels promoting amyloid formation and deposition, particularly if further ANP secretion is induced by atrial dilation and stretch associated with AF.16 In this issue of the journal, Renoux et al.,17 using the healthcare database involving the residents in Québec Province of Canada, identified 147 622 individuals with non-valvular AF between 2000 and 2009 in order to investigate the gender-related difference in the rates of stroke as well as mortality. The database was the linkage of three components: the demographic database, the medical services database, and the prescription database, and covers the entire population of Québec, thus offering the potential advantage of minimizing the selection basis. Unlike recent similar nationwide reports, the study population included both in- and outpatients. This analysis follows the earlier report on the incidence of AF and associated mortality which demonstrated a higher age-related case fatality in men with AF.18 In the time-fixed model with adjustment for confounders, female gender appears to be associated with a 16% higher risk of stroke, which is somewhat expected and is in line with many previous reports. The strength of this study, however, is that in addition to the standard time-fixed analysis, time-dependent adjustment for confounders has also been performed in which female gender no longer poses any additional risk of stroke compared with male counterparts (relative risk, 1.01; 95% confidence interval, 0.97–1.05). The time-dependent adjustment for covariates offers the opportunity to unravel the specific effect of a variable on the endpoint of interest. The study took the previous observations of the association between gender and stroke to different levels by assessing the continuous effect of gender in the changing settings. Albeit that the mean follow-up may seem relatively short, the large sample size is a significant asset towards the credibility of these findings. The results of recent large surveys which did not support female gender as an independent risk factor for stroke have paved the way to a more ‘relaxed’ approach to treating the female gender per se as an independent component of stroke risk stratification, leading to the increase in the threshold for the absolute indication for OAC (class I recommendation) in women in the 2016 edition of the European Society of Cardiology (ESC) Guidelines.19 In addition, given controversial reports and uncertainty about risk of stroke in patients with a low CHA2DS2VASc score, the ESC Task Force commissioned analysis of stroke risk in men and women with additional risk factors. This analysis was based on the linked electronic health record database of 70 206 individuals with AF, and it found significant positive net clinical benefit of vitamin A antagonists (VKAs) from CHA2DS2VASc ≥2 in men and from CHA2DS2VASc ≥3 in women.20 In contrast to the previous set of the guidelines, OAC must be implemented in women with two additional major risk factors for stroke, whereas the strength of the recommendation has been downgraded to IIA, meaning that albeit that OAC is still advocated in women with an additional risk factor from the CHA2DS2VSc scale bar gender itself, the decision should be based on individual characteristics and patient preferences. However, there are several consideration to take into account. First, about half the patients in the Quebec database were anticoagulated (with a VKA), with the majority (80%) starting the treatment within 3 months following AF diagnosis. The endpoints of stroke clustered in the relatively short space of time—during the first months, possibly prior to anticoagulation. Hence, the mild association between gender and stroke (a 16% increase in women) found in time-fixed analysis may be real, whereas anticoagulation may have negated the gender effect. Secondly, the endpoint of ischaemic stroke has also included transient ischaemic attack, which is notoriously difficult to diagnose and, given the inherent limitations of administrative databases where information is derived form the International Classification of Diseases (ICD) codes, some events may have been misclassified, albeit that the effect of such misclassification was likely to be small, given the large number of events (>11 000 of endpoints were reported). Due a generally better safety profile of non-vitamin K oral anticoagulants (NOACs), the threshold for the initiation of anticoagulation in order to achieve the positive net effect is lower for this type of drugs compared with VKAs (<1% vs. <2%). The net clinical benefit analysis that has influenced the strength of the recommendation in low-risk patients has demonstrated a positive but non-significant treatment benefit of VKA in individuals with a CHA2DS2VASc score of 1, but did not include patients treated with NOACs, which could have led to different results.20 The Québec study provides a valuable addition to understanding the gender-related outcomes associated with AF. It is generally accepted that, in the modern era, the gender factor per se should not affect major outcomes of cardiovascular morbidities including stroke, but gender-specific interactions related to age and responses to therapies have not been well understood. The more pronounced effect of AF on mortality, the effectiveness of NOACs, and the differences in AF management in women of different age and with different co-morbidities. remain the subject of further research. Conflict of interest: none declared.