Anticoagulation variability: is it the physician, patient or hospital?

Oral anticoagulation (OAC) with vitamin K antagonists (VKA), such as warfarin, has demonstrated overwhelming evidence for the reduction of stroke and thromboembolism in patients with nonvalvular atrial fibrillation (NVAF) [1, 2]. Despite an impressive evidence-base from numerous clinical trials in recognizing stroke risk factors [1, 2] as well as the publication of various thromboprophylaxis guidelines for atrial fibrillation (AF) [3–5], the rate of utilization of VKAs for stroke prevention in atrial fibrillation (AF) is still inadequate [1]. Why is OAC utilization still suboptimal given its demonstrable benefits and what contributes to the large variation in anticoagulation rates? There are numerous reasons why patients with AF do not receive OAC [6–10]. Indeed, three crucial elements act as barriers to OAC prescription in patients with AF, so-called ‘patient factors’, ‘physician factors’ and ‘health-care system factors’. ‘Patient factors’, which may effect acceptance of, and compliance with, OAC therapy include patient’s treatment preferences and beliefs about their health [11, 12], their understanding of the necessity for OAC [13, 14], satisfaction with their medical care and physician [15] and psychosocial factors [6]. Further, problems inherent to OAC therapy, namely, the inconvenience of regular International Normalised Ratio (INR) monitoring and dietary and alcohol restrictions, all impact on the uptake and adherence of OAC therapy. Amongst physicians, bleeding risk is most commonly cited as the reason for not initiating OAC therapy [6–10]. Physicians’ perceptions of risk and benefit are not always reliable with bleeding risks often overestimated [16, 17] and thromboembolic risk underestimated, particularly, among older patients. Physicians are prepared to maintain anticoagulation at a lower target INR range (e.g. 1.6 to 2.5) in older patients, which only offers modest (and probably suboptimal) stroke protection, especially when INRs fall <2.0. Increasing age in itself should not be a contraindication to anticoagulation, although the former is a risk factor for bleeding, especially in those aged over 85 years [18]. Qualitative research has demonstrated that physicians’ personal experiences with OAC and the number of years in medical practice influences whether or not OAC therapy is initiated [10]. Treatment priorities also differ, physicians would rather avoid major bleeding than prevent stroke, whilst patients place more emphasis on stroke prevention and would be prepared to accept the risk of bleeding with OAC to avoid a stroke [13, 19]. Clinical guidelines, as useful and relevant as they are, could still seem irrelevant to some physicians who would question the applicability to their own patients [6]. Physicians reluctance to prescribe OAC therapy is reflected in the Euro Heart Survey [20], which demonstrated that only 67% of patients eligible for OAC therapy are actually prescribed it. Does the evidence of the benefit of OAC seen in clinical trials translate into ‘real life’ practice? It is often argued that patients enrolled into clinical trials are not representative of the types of AF patients we routinely see in clinical practice. This may be true with <10% of patients screened in the initial clinical trials comparing VKA with control being subsequently randomized. Patients in clinical trials often are less fragile, more mobile and compliant with medication, with less comorbidities and have undergone stringent trial inclusion/exclusion criteria checks before being recruited. However, two studies have demonstrated the clinical effectiveness of warfarin in ‘real life’ AF patients with a 51% risk reduction in thromboembolic events in a cohort of 12 000 AF subjects in Northern California [16] and a 40% stroke risk reduction in a cohort of 5100 male Danish patients [17]. The reported stroke risk reductions are lower than the established 64% risk reduction evident from randomized controlled trials [1], but such studies should reassure physicians that warfarin can still be utilized effectively outside the clinical trial setting. Even in the primary care setting with an elderly population (>75 years), AF patients benefitted greatly in terms of thromboprophylaxis with warfarin, with no difference in major bleeding rates when compared to aspirin [21]. Further, the majority of physicians believe that INR monitoring in clinical trials and real-life practice is not comparable because of paucity of resources in the latter and several barriers related to the health care system. Unsurprisingly, the percentage of time spent within the therapeutic target INR range (2.0 to 3.0 for AF), is unlikely to be on par with the level achieved in clinical trials. Indeed, in one study conducted in a university teaching hospital involving 2223 NVAF patients, almost a third of the treatment time and close to half of the INR readings were outside target therapeutic range [22]. Self-monitoring of the INR with dose adjustment has been shown to improve the percentage of time spent in the therapeutic INR range and to reduce the frequency of major bleeding in older patients commencing long-term OAC therapy [23]. Further, a recent systematic review and meta-analysis of 14 RCTs of self-monitoring [24] revealed that self-monitoring alone significantly reduced major haemorrhage (OR 0.65; 95% CI 0.42–0.99), thromboembolic events (OR 0.45, 95% CI 0.30–0.68) and all-cause mortality (OR 0.61, 95% CI 0.38–0.98), and that a combination of self-monitoring with self-adjusted OAC therapy also demonstrated a significant decrease in death and thromboembolic events, but not haemorrhage. However, self-monitoring requires the infrastructure within the healthcare service to support such care, but these results suggest that it may be important to develop alternatives to clinic monitoring of INR. In this current issue of the Journal of Internal Medicine, Hansen et al. [25] performed an observational study via nationwide Danish administrative registers (Patient Registry and Registry of Medicinal Product Statistics) between 1995 and 2004, evaluating hospital variation in usage of OAC therapy (with, warfarin or phenprocoumon) applicable to first-time hospitalized AF patients and their clinical consequences. Three types of hospitals were included: tertiary university hospitals, main regional (district) and local community hospitals. Only 37% of the patients received OAC, with substantial variation between hospitals, ranging from 17% in the hospitals with the lowest rates, to 50% in those with the highest OAC usage. Understandably, cardiology departments had the highest usage of OAC with 40.1% of patients receiving OAC compared to other departments; interestingly, local community hospitals, which lacked cardiology input, had higher OAC use (39.3%) in comparison to university (35.5%) and main regional hospitals (36.5%). These differences translated to a significant increase in thromboembolic risk in hospitals with a low OAC use (hazard ratio 1.16, 95% CI 1.10–1.22), with minimal augmented risk of intracranial haemorrhage from hospitals with higher OAC usage. Despite increased OAC usage over time, there were still persistent hospital variations. Two salient points are worth highlighting: (i) we cannot be certain that the higher risk of thromboembolism is directly related to not being on OAC therapy in hospitals with low OAC use or because of some other reason (for e.g. poorly controlled hypertension) and (ii) that the overall uptake of OAC (37% patients) was fairly low in comparison to the Euro Heart Survey (67%) [20]. In addition, we do not know how the AF diagnosis was derived in the present study [25], which is based on administrative registers with limited detail on clinical data. Further, were the physicians from hospitals with high OAC usage much better in risk-stratifying AF patients and hence, this contributed to the high OAC usage? The results from Hansen et al. [25] are presumably based on mainly white-Europeans, with minimal patients from ethnic minorities. Thus, the data cannot be extrapolated to nonwhite populations because ethnic origin has some influence on the epidemiology and some management aspects of AF. For example, there is a lower risk and prevalence of AF in both Indo-Asians and Afro-Caribbeans despite a greater prevalence of diabetes and hypertension, when compared to whites [26]. Moreover, different ethnic groups have different degrees of knowledge and perceptions about AF coupled with its management; in the West Birmingham Atrial Fibrillation Project, Indo-Asians had the least awareness about warfarin and thromboembolism associated with AF [27]. Indeed, it may not be easy to overcome patient-centred obstacles, especially when it comes to knowledge and perceptions coupled with language barrier and time constraint in the out-patient setting. However, a pilot study demonstrated that a brief educational intervention with information booklet in an AF cohort can help to somewhat improve their knowledge about their condition and OAC use [14]. Perhaps a more rigorous educational intervention could appreciably improve patients’ wisdom about AF, perceptions of disease and uptake of OAC therapy; this strategy would be accompanied by a significant cost and whether it could be successfully applied internationally is still unknown. Of the various patient educational priorities associated with AF management, the recognition of bleeding risks is evidently imperative to avoid complications of OAC use. Until now, the published models in predicting bleeding risks in patients on anticoagulation therapy have not been fully validated in large prospective cohorts of AF patients [28]. Conceivably, if there is a well-validated bleeding risk predictor tool or even a composite stroke and bleeding risk predictor, both physicians and patients could make an informed decision, weighing up the perceived benefits and risks [as a ‘net clinical benefit’ assessment] prior to initiating anticoagulation therapy. No conflict of interest was declared.