Optimal long-term antithrombotic management of atrial fibrillation: life cycle management

Given the significant antithrombotic therapeutic consequences and their potential sequelae, it all starts with unequivocal electrocardiographic documentation of AF. Next, identifying clinical AF duration and type, adequate treatment of symptoms, underlying cardiovascular and pulmonary disease and/or other triggers. However, it is worth noting that the indication for OAC is independent of the type of AF. A clinically challenging scenario is the patient with secondary AF: a paroxysmal episode only documented during a temporary, reversible disease state (e.g. thyrotoxicosis, post-operative inflammation, and infections such as pneumonia). These patients have both a higher risk of ‘future’ AF and worse long-term cardiovascular outcomes [4]. However, optimal long-term antithrombotic management of secondary AF remains unclear in the absence of randomised controlled trials (RCTs). Despite the association of OAC use with reduced long-term mortality in post-operative AF patients [4], appropriate risk assessment, timing and patient preference need to be taken into account. Currently, the same accounts for atrial sensing of asymptomatic high atrial rate episodes by implantable devices. There is supporting evidence that the majority of these patients with ‘subclinical AF’ also have an increased stroke risk [5, 6], it is still unclear whether and when they benefit from OAC, also considering the bleeding risk. OAC has to be considered in patients with two additional CHA₂DS₂-VASc risk factors (i.e. ≥2 in males, ≥3 in females) and an AF burden >24 h (if there are no contraindications) [7]. Lower duration may merit OAC if multiple risk factors are present.

The vascular nature of AF is underlined by the high prevalence of cardiovascular co-morbidities, which represent the intrinsic long-term stroke risk. The most important independent stroke risk factors are incorporated into the CHA₂DS₂VASc score and its use is recommended for stroke risk assessment in AF patients [2]. Considering the increasing aging of our society it is important to emphasise that age is an extremely strong, continuous, risk factor for stroke [11] and aging potentiates other risk factors such as heart failure. Therefore, the CHA₂DS₂VASc score should be reassessed over time [12]. Untreated, one out of every four octogenarians with AF suffers a stroke during their remaining lifetime [1]. Still, across various health care settings studies demonstrated age, frailty and dementia to be some of the strongest independent predictors of withholding anticoagulation in the eldest elderly [13‐15]. However, in these patients the key question is not whether, but how we should organise optimal anticoagulation.

The downside of any antithrombotic drug is an increase of patients’ intrinsic risk of bleeding. It therefore makes sense to identify bleeding risk factors. Practical bleeding risk scores, such as HASBLED [16], assist in particular with identifying modifiable risk factors. Thereby minimising the risk of major bleeding during OAC treatment. However, we should keep in mind that the available risk scores fail to capture important frailty aspects in the eldest elderly with AF [17] and, similarly in elderly with venous thromboembolism [13]. Here, previous reports demonstrated that age itself proved to be a strong determinant of risk. Especially the eldest elderly (85+) appear to have a substantial increased risk of major bleeding [18, 19]. This enhanced bleeding risk also applies for patients with a poor quality of anticoagulation (i.e. time in therapeutic range [TTR] <65%, patients spent a median 14%, 71%, and 15% of time below, within, and above the intended therapeutic range, respectively) [20].

According to the current European Society of Cardiology/European Heart Rhythm Association (ESC/EHRA) guidelines men and women with a CHA₂DS₂VASc score of 2 and 3, respectively, clearly benefit from OAC. In case of a CHA₂DS₂VASc score of 1 in men and 2 in women, OAC is likely to provide net clinical benefit and individual characteristics and patients’ preferences should be considered [2]. However, a primary care study among 260 AF patients aged 70–85 years that used decision analysis regarding antithrombotic treatment preference showed that approximately half of the patients with a preference for OAC did not receive it [21]. On the other hand, 47% of patients were not being prescribed warfarin although the results of their decision analysis suggested they wanted to be. A researcher-administered questionnaire study using a fictional AF patient and pictograms to display the 10-year risk of stroke showed that 100% of the elderly chose OAC over no treatment. However, when the necessity of many tablets with multiple blood tests, alcohol restrictions and an increasing risk of intracranial haemorrhage was introduced, the choice for anticoagulation dropped. A daily tablet corresponds to 94%, an intracranial haemorrhage risk of 0.1%/year corresponds to 99% and 0 units of alcohol/day corresponds to 89% who would still choose OAC [22]. Although most physicians will acknowledge its importance, practical and evidence-based methods of shared decision making in AF patients are an unmet need (Table 1).

The downside of any antithrombotic drug is the increase of the patients’ intrinsic bleeding risk. Undeniably, fear for bleeding has caused physicians to choose aspirin over VKA for perceived lower bleeding risk. However, it is important to realise that there is no relationship between a drug’s ability to prevent stroke and its potential to harm through bleeding (safety paradox). In fact, irrespective of age, aspirin use in AF patients is associated with a similar rate of major bleeding events compared with VKA but with significantly inferior stroke protection [23, 24]. This also holds true for the eldest elderly (n = 366 ≥ 85 years): aspirin vs. apixaban, 4.9% vs. 4.7% major bleeding events and 7.5% vs. 1.0% strokes, respectively, per year in one study [19]. Furthermore, the combination of aspirin with clopidogrel also provides inferior stroke protection in AF patients compared with VKA [25]. Finally, OAC is the only antithrombotic regimen that reduces all-cause death [26].

Thus, even in AF patients with a low stroke risk (i.e. CHA₂DS₂VASc score <2 in men and <3 in women), aspirin is not a suitable alternative, it warrants either no antithrombotic therapy or OAC [2]. However, we should keep in mind that these patients may require aspirin for the prevention of other atherosclerotic complications.

The collaborative evidence from multiple, large clinical trials randomising elderly AF patients to a NOAC or VKA consistently demonstrated at least similar efficacy and safety, but significantly less intracranial haemorrhage with all NOACs compared with VKA [2]. In addition, compared with VKAs, NOACs are much easier to use with a rapid onset and offset (comparable with low-molecular-weight heparins), standard daily dosing (no need for frequent routine blood tests) and limited drug interactions. Therefore, NOACs are the current first-line oral anticoagulant in the absence of a contra-indication [2].

NOACs are only contra-indicated in the infrequent case of, usually rheumatic, moderate/severe mitral stenosis, end-stage renal disease, or mechanical heart valves. These subgroups were excluded from the NOAC trials. Furthermore, the thromboembolic pathophysiology in both rheumatic valve disease [1] and mechanical heart valves is significantly different. For now, in the case of the above-mentioned groups VKAs are indicated.

In the absence of head-to-head randomised controlled trials there is no superior NOAC. The value of these drugs lies in the significant benefits over VKAs and aspirin as discussed above. Still, nuances between the respective landmark trials (e.g. baseline stroke risk factors, age, the occurrence of side effects), practical aspects (e.g. once vs. twice daily, tablet/capsule size, blister packaging, antidote) and pharmacokinetics (e.g. reduced renal function, drug interactions) allow for choice (Fig. 2). Furthermore, the prescribing physician’s experience with and preference for specific NOACs should play a role in this choice.

Of much greater importance is choosing the appropriate NOAC dose in order to replicate RCT results in daily clinical practice and pursue optimal stroke protection (Fig. 2). This seems logical, but ‘real-world’ data consistently show label-discordant dosing, in particular an inappropriately lower dose [27‐32]. Although the reasons remain speculative, it could be because of a perceived increased risk of bleeding. However, physicians should be aware of the safety paradox: inappropriate low dosing results in potentially preventable strokes [33, 34], not significantly fewer major bleeding events [33, 35]. In order to appropriately dose NOACs the relevant criteria must be checked. Renal function is important for all NOACs, but information about recent renal function was lacking in a third of the AF patients in a large international prospective NOAC registry [32]. This calls for nationwide prospective registries, such as DUTCH-AF (Project number 848050006, ZonMw-programma Goed Gebruik Geneesmiddelen).

As pointed out by Maikranz et al. [37] it appears that very few patients with long-term OAC know how to act in case of unexpected major bleeding events. However, it is important to realise (and communicate) that these major events occur despite optimal preventive measures and to discuss with the patient how to act.

A predominantly upper gastrointestinal bleeding is the most frequent major bleeding complication in patients using long-term antithrombotic treatment, both aspirin and OAC [38]. Although gastrointestinal bleeding in itself is often non-fatal, it is a prognostic marker as reflected by a death rate of 30–35% in two years following the index event [39, 40]. Restarting OAC following a gastrointestinal bleeding is the only regimen associated with both a very significant reduction of all-cause mortality [39‐43] and stroke. Although the risk of major bleeding increases with reinitiating OAC, the risk of a recurrent gastrointestinal bleed does not [39, 40], except in patients with end-stage renal disease [43]. In case of repetitive gastrointestinal bleeding, quality of life also plays an important role in the shared decision to whether or not restart OAC therapy. Furthermore, it is important to take the potential mechanism into account. For instance, esophagogastroduodenoscopy confirmed ulcer healing before initiation of OAC significantly reduces the rate of major bleeding [44]. Also, good (>65%) opposed to poor (<65%) TTR did not increase the risk of a recurrent ulcer bleeding but provided better stroke prevention [42]. The optimal time window of reinitiating OAC following gastrointestinal bleeding remains unclear. Qureshi et al. found that restarting OAC seven compared with thirty days after a gastrointestinal bleeding did not increase the risk of major bleeding but was associated with lower mortality and stroke rates [39].

Regarding the most feared and fatal bleeding complication, intracranial haemorrhage, a Danish nationwide study revealed that AF patients who survived the initial six weeks following an intracranial haemorrhage had a substantial reduction of all-cause mortality and stroke compared with no OAC during one-year follow-up (hazard ratio 0.55, confidence interval 0.39–0.78) [45]. A Swedish nationwide study showed that of the surviving 1,454 patients of a first intracranial haemorrhage under OAC, 10.4% resumed OAC within three and 21.2% within twelve months [46]. Oral anticoagulants should probably not be resumed in patients with a lobar intracerebral haemorrhage caused by cerebral amyloid angiopathy, given the high risk (3–14% within one year) of recurrent intracranial haemorrhage [47]. Although RCTs answering the question whether and when to resume antithrombotic medication in patients who suffered an intracranial haemorrhage are lacking, one is currently recruiting in the Netherlands (APACHE AF trial) (Table 2). Hence, resuming antithrombotic medication should be a multidisciplinary decision, involving a cardiologist, neurologist and a geriatric specialist or general practitioner [48].

Early stroke recurrence of cardioembolic origin is as high as 14% within two weeks. OAC substantially reduces this risk. Even, or perhaps especially, the eldest elderly benefit as Appelros et al. demonstrated a reduced all-cause mortality and stroke rate in nonagenarians (between 90 and 100 years of age) [49]. Initiation of OAC monotherapy, which outperforms low-molecular-weight heparin alone or followed by OAC, 4–14 days following the index stroke opposed to outside this window provides the best protection against recurrent stroke [50]. Perhaps beside the use of the National Institutes of Health stroke severity scale [2], the novel ALESSA score might assist in deciding upon early initiation or re-initiation [51].

Side effects could trigger incorrect or non-use of OAC leading to a vicious circle. It therefore deserves special attention during follow-up. However, it seems as if a magnifying glass is being held over the side effects of NOACs as reflected by the registration in the Netherlands (Lareb). Interestingly, the industry filed most (35%) of the reports followed by patients (17–20%) and pharmacists (16%), whereas physicians (8%) provided the least of the reports. In the Netherlands, rivaroxaban and dabigatran became available in 2011, followed by apixaban (2013) and edoxaban (2015) (Fig. 3a). Since then, the Netherlands pharmacovigilance centre Lareb received a total of 1,487 (609 were severe) side effect notifications among 84,047 NOAC users. A NOAC specific breakdown is displayed in Fig. 4a. During the same time period (2011-present) among the >400,000 VKA users (Fig. 3b) Lareb received a total of 1,237 (366 were severe) side effect reports (Fig. 4b). Perhaps, this provides food for thought on how we can optimise the monitoring and reporting of long-term side effects in the near future. Furthermore, results of real-life studies and registries are instrumental in understanding how new therapies find their way in daily clinical practice and for the identification of pitfalls.

Several ongoing and future trials will also provide answers on current knowledge gaps (Table 2).