Major progress in anticoagulant uptake for atrial fibrillation at last: does it translate into stroke prevention?

This editorial refers to ‘A 10 year study of hospitalized atrial fibrillation-related stroke in England and its association with uptake of oral anticoagulation’†, by J.C. Cowan et al., on page 2975. A number of randomized controlled trials in atrial fibrillation (AF) conducted in the last century showed that vitamin K antagonists, (VKAs), principally warfarin, were remarkably effective in stroke prevention, reducing stroke by 64% and all-cause mortality by 26%,1 figures rarely matched in treatments of other diseases. With the introduction of the non-VKA anticoagulants (NOACs) in the last 8 years, stroke reduction in the randomized trials was similar to VKAs, but with less intracranial haemorrhage and a further 1% absolute risk reduction in total mortality;2 results confirmed in many ‘real-life’ studies using administrative or registry data.3 Many guidelines, including the first joint AHA/ACC/ESC guideline in 2001,4 recommended oral anticoagulant (OAC) therapy for those with AF at elevated risk of stroke, as have subsequent joint and ESC guidelines and updates in 2006, 2010, 2012, and 2016. But OAC uptake has been less than ideal for some time,5 , 6 hovering in the UK at ∼50% of those for whom OAC was guideline recommended, in the 10 years from 2001 to 2011. The data from the Swedish RIKS registry7 of >94 000 strokes help identify the two major gaps3 in prevention of AF-related strokes, which represent 33% of all ischaemic strokes (Take home figure). In 9% of all strokes, AF was not known prior to stroke: prevention would require screening to detect and treat unknown AF.8 In another 20%, AF was known but patients were either on no antithrombotic therapy (8%) or on antiplatelet therapy alone (12%), which is largely ineffective for AF thrombo-prophylaxis.9 Thus, up to 20% of ischaemic strokes might be prevented if the evidence–practice gap of OAC prescription was eliminated. It is therefore very pleasing to see in this issue of the European Heart Journal that in the UK, OAC use in patients with elevated stroke risk has increased from <50% to almost 80% in the 10 years from 2006 to 2016, mostly in the latter 5 years, while antiplatelet therapy has reduced from 43% to 16%.6 This is a great success story, observed to some degree in many countries, but by no means uniformly, even within the same country.10 , 11 Stroke prevention gaps in atrial fibrillation. Data from the RIKS stroke and Swedish patient register (Friberg et al. 7). Relationship to AF for 94 083 ischaemic strokes between 2005 and 2010 according to whether AF was previously known or unknown, and treatment at time of stroke for those with known AF. For further information on calculation of numbers, see Freedman et al. (online supplementary material).3 Trying to pinpoint the reason for this improvement is not easy. Success has many fathers, while failure is an orphan. Cowan and colleagues6 postulate that successive ESC and NICE guidelines, quality improvement incentivization programmes such as GRASP and QOF, or licensing and marketing of NOACs, accentuating their ease of use, could all have played a role, but it is difficult to determine which is the major one. ESC guidelines and their updates were recommending OAC in 2001, 2006, and 2010, and the GRASP quality initiative from 2009, with little change in OAC uptake, while QOF did not commence until 2012. It is instructive to look at our own UK data using a higher resolution temporal trend of OAC initiation within 90 days of incident AF12 (Figure 1). The graph plots total OAC initiations for patients with CHA2DS2-VASc ≥2. A rise in VKA OAC prescription was observed from early 2011, when limited marketing preceded the availability of NOAC drugs, followed by an accentuated increase due to NOAC prescription accompanied by a plateau in VKA prescription. It is quite likely that at this time point, the availability and marketing of NOACs played a dominant role, with this influence increasing as additional NOACs became available. The final increase from 2013 to 2106 is probably a combination of all three factors: an effective national QOF quality framework, gradual percolation of guidelines and updates, and continued marketing of NOACs. Time trend of the proportion of patients with incident AF and CHA2DS2-VASc score ≥2 having OAC treatment (total, VKAs, and NOACs) within 90 days of AF diagnosis, between 2011 and 2014. Data from UK Clinical Practice Research Datalink (CPRD). Reproduced with permission from Martinez et al. 12 Ensuring that a high proportion of patients with AF are taking effective doses of OAC rather than aspirin requires first that OAC is offered by the physician and accepted by the patient, secondly that patients are adherent to taking the drugs as prescribed, achieving >70% time in the therapeutic international normalized ratio (INR) range for VKAs, and finally, that patients persist with taking OAC therapy long-term. There is a high non-persistence rate for OAC in the first 2 years after initiation, significantly greater for VKAs than for NOACs,12 but not ideal for either, and greater emphasis needs to be placed by physicians on keeping therapy persistence high to prevent stroke. It is intuitive to presume that as anticoagulant usage increases for AF, pari passu with a decline in aspirin usage as a single thrombo-prophylactic, the AF-related stroke rate would fall. This is why the study of Cowan et al. 6 is so interesting, as the authors used multiple databases and registries between 2006 and 2016 to determine whether a reduction in stroke rate was occurring. They showed that the prevalence of AF increased gradually from 692 054 to 983 254 over 10 years, with no difference in the rate of increase between 2006–2011 and 2011–2016. AF incidence has also been noted to rise in the UK between 2001 and 2013, largely resulting from increased numbers of older people with a non-primary AF hospital discharge diagnosis.13 Atrial fibrillation-related stroke increased in the first half of the 10-year period, as the AF prevalence rose. It was very reassuring to note that in the second half of this period, even while AF prevalence continued to rise, the AF-related stroke rate plateaued and then fell. Plotting the AF-related stroke rate per 100 000 AF as in their summary figure, made the rise, plateau, and fall seem even more striking. Modelling then showed that the fall in stroke rate was related to the increasing use of OAC and decreasing use of antiplatelet drugs. Taken at face value, this is an amazing success story for stroke prevention in AF. Rarely is it shown that increased adherence to guideline-recommended therapy for a serious condition translates to a reduction in a feared and incapacitating outcome such as stroke. Of course, as the authors point out, association between increased OAC uptake and reduced AF-related stroke is not proof of causation. The path of truth is strewn with chance associations that are not causal. In this instance, there is a strong a priori case that stroke rate should reduce if enough people with AF are taking effective thrombo-prophylaxis against cardioembolic stroke, and therefore must remain the most likely explanation. Are there other plausible alternative explanations? It is certainly possible that better treatment of co-morbidities such as hypertension and hyperlipidaemia were occurring at the same time and could account for some of the reduction in the AF-related stroke rate. Levels of co-morbidities actually increased over time as judged by the Charlson index. It is also possible that paroxysmal AF, which has a slightly lower stroke risk than persistent AF, was more commonly diagnosed in the 10-year study time frame, so the overall stroke risk would fall, although there are no data to support this possibility. Finally, the data on AF prevalence and stroke rates are aggregated data from different data sources, rather than tracking of individual patients in a large cohort, and, as the authors concede, could lead to a ‘potential “ecological fallacy” in which the findings are attributed incorrectly to the individual patient level’. There is also a downside to anticoagulant therapy, principally bleeding, with intracranial haemorrhage and haemorrhagic stroke the most feared bleeds. The reductions in stroke rate were accompanied by a much smaller increase in both intracranial haemorrhage and haemorrhagic stroke, mostly in the first 5 years, and no significant increase in the second 5 years, possibly because NOACs were increasingly used rather than VKAs in the latter period. Those with an eye to history will note the parallels of the observations of Cowan et al. in this issue with studies of temporal trends in AF-related stroke in the last two decades of the 20th century.14 , 15 These studies noted a steady decrease in stroke rates associated with an increase in OAC prescription, most marked between 1995 and 2002. In the Olmstead county study,14 OAC use increased to 30% in 1995–2000 from a low base of 9% in 1980–1985. The authors also noted a decrease in systolic blood pressure during the same period, and ascribed the reduced stroke rate to both improved control of hypertension and increased OAC uptake. In a cohort derived from a 5% US Medicare sample,15 OAC use increased from 24% to 56% between 1992 and 2002, but US OAC usage has remained rather static since then: in the large Pinnacle registry,11 OAC use rose only to 59% in 2013–2014 from 55% in 2007–2008 (OAC rates calculated from data in the publication). The major association with non-prescription of OAC in that study was antiplatelet use, even in those without vascular disease, highlighting the likely impact of misperceptions of antiplatelet efficacy and safety on appropriate prescription of oral anticoagulant for AF.9 We do seem to be witnessing a success story in the progress of stroke prevention in AF, starting with findings from randomized clinical trials leading to implementation in guidelines, quality improvement, new drugs, and finally changes in management at scale. This has resulted in an estimated annual prevention of >4000 strokes in 2015/16 in the UK. While progress has had a long gestation, hopefully it will be realized globally to reduce the AF-related stroke burden. Conflict of interest: B.F. reports grants to the institution for investigator-initiated studies, personal fees, and non-financial support from Bayer Pharma AG, grants to the institution for investigator-initiated studies, and personal fees and non-financial support for workshops from BMS/Pfizer, and personal fees for lectures in official meeting satellite symposia from Daiichi-Sankyo.