Which patients should have prolonged monitoring?

Risk factors and biomarkers of AF could in theory help to select patients for prolonged monitoring. However, many patient-related risk factors for AF, such as age and hypertension, are poorly discriminating, also being risk factors for atherosclerosis.32 Heart failure and mitral stenosis might be more specific,33 and the recently developed HAVOC risk score, which weights congestive heart failure heavily, has shown potential for triaging patients with cryptogenic stroke into low-risk, medium-risk and high-risk categories for AF detection.34 ECG biomarkers including PR interval prolongation and atrial premature beat count are also being developed.35 36

Clinical and radiological features might also help to identify patients likely to have AF. Large cardiac fibrin-rich thrombi tend to travel in medium-calibre high-flow vessels, leading to proximal arterial occlusion and a large infarct, or fragmentation and distal embolisation into terminal cortical branches. Consequently, the features of cardioembolic stroke include higher clinical severity, more cortical symptoms and signs, and certain radiological patterns (figure 1), including whole-territory infarction, wedge-shaped infarcts extending into the cortex, striatocapsular infarction (from transient M1 occlusion), isolated posterior cerebral artery infarction causing hemianopia, scattered distal infarcts within an arterial territory or external borderzone, and multiterritory bihemispheric or anterior and posterior circulation infarcts.37 Haemorrhagic transformation is probably also characteristic of cardiac embolism since cardiac, ‘red’ (fibrin-rich) thrombi are more liable to re-perfuse and consequently to have associated haemorrhage within the infarct. However, none of these findings is wholly reliable: even small subcortical infarction may occur in the presence of a definite embolic mechanism,38 and a recent meta-analysis showed that the AF detection rate at 7 days after ischaemic stroke is similar whether the stroke is classified clinicoradiologically as due to small or large vessel occlusion.39 Importantly, trials of prolonged monitoring gave high detection rates in relatively unselected patients with cryptogenic stroke. Therefore, until we have a validated biomarker-based approach, we advocate prolonged monitoring for all patients with cryptogenic stroke.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

MRI showing patterns of acute infarction associated with cardioembolism. Top row, from left: diffusion-weighted imaging showing isolated cortical infarct, small wedge-shaped predominantly cortical infarct and medium-sized cortico-subcortical infarct; T2-weighted imaging showing large wedge-shaped infarct with T2 hypointensity indicating haemorrhagic transformation. Bottom row, left to right: posterior cerebral artery territory infarction, scattered cortical and subcortical infarcts within right middle cerebral artery territory, striatocapsular infarction with additional small right middle cerebral artery territory infarcts (suggesting transient M1 occlusion then fragmentation and distal embolisation of thrombus), scattered bihemispheric lesions.

Monitoring techniques

The optimal techniques for ECG monitoring are uncertain, with a lack of head-to-head trials and recent rapid advances in device design. In most departments, the first-line method for non-invasive AF detection remains via Holter monitoring, developed in 1962 and limited by relatively short recording duration, bulky design, an inability to use while showering and delays in reporting. Newer hand-held patient-activated devices such as Kardia and Zenicor offer portability, cost-effectiveness and speed,40 but not the continuous monitoring required to detect asymptomatic AF. Our preference is for newer patch-monitoring devices, such as ZIO, Bardy or e-Patch, or lead monitoring via R-test or Apoplex (table 5, figure 2). These devices allow continuous monitoring for longer durations, a lightweight water-resistant design, easy application in clinic by a nurse or doctor, and email reporting within 72 hours.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Patch and lead monitoring devices (left to right): e-Patch, Zio, CaM, R-test, Apoplex.

After non-invasive monitoring, our practice is to proceed to prolonged monitoring via an implanted cardiac monitor in patients whose ischaemic stroke remains unexplained, with a high suspicion of cardiac embolism (eg, non-lacunar, especially if due to cortical branch occlusion, or in multiple vascular territories, or with other radiological characteristics shown in figure 1), and in whom we would recommend anticoagulation if AF is found. Traditional catheter-inserted monitors have now been replaced by subcutaneous devices, inserted via a <1 cm incision, which offer up to 3 years of continuous monitoring, MRI compatibility, automated arrhythmia detection, as well as a patient-activated event alarm, and wireless data download. The Reveal LINQ monitor (figure 3) has recently been the subject of a NICE MedTech Innovation briefing.41 Such systems are generally well tolerated, being small and lightweight, with a low rate of complications (in CRYSTAL-AF, the removal rate due to infection or erosion was 2.4%). The LOOP trial, reporting in 2020, is addressing whether their use translates into a reduced risk of stroke.42

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Reveal-LINQ is a small implanted cardiac monitor, inserted subcutaneously in the precordium via a <1 cm incision.

Low-burden and implanted device–detected AF

With the greater use of prolonged monitoring, there will be an increasing number of patients found who have short and infrequent paroxysms of AF—that is, a low ‘AF burden’. The patients in the key clinical trials of anticoagulation and risk score development cohorts will largely have been diagnosed by short-duration monitoring and are therefore more likely to have had a high AF burden. Therefore, whether patients with low-burden AF diagnosed through prolonged monitoring benefit similarly from anticoagulation is an important uncertainty. However, current national and international guidelines do not distinguish by AF burden, and our practice is to consider anticoagulation in any patient with proven AF of any duration. It is likely that AF burden increases over time, and the haemodynamic effects of irregular atrial contraction may not be the only mechanism of intracardiac thrombosis in AF.43 Most studies correlating AF burden with stroke risk have been performed in patients with pacemakers or implanted cardioverter-defibrillators able to detect paroxysmal atrial tachycardias termed ‘atrial high rate episodes’. The results of these studies are inconsistent, finding ischaemic stroke risk to be increased variously by the presence of an atrial high rate episode of >5 min, 6 min or 24 hours, or by a daily burden of 5.5 hours or more on at least 1 day during the monitoring period.44–48 Atrial high rate episodes correlate with AF detection on dedicated rhythm monitoring, but imperfectly; therefore, these results should not be extrapolated directly to patients with proven AF. Neurologists and stroke physicians should however be aware that interrogation of an implanted cardiac device in a patient with stroke may provide evidence of an increased risk of AF, and that two clinical trials (NOAH AFNET-6 and ARTESIA) are testing the benefit of anticoagulation in patients with atrial high rate episodes lasting less than 24 hours.49 50 In the meantime, a proposed approach for the management of atrial high rate episodes, based on European Society of Cardiology guidance, is to consider anticoagulation for patients with a previous ischaemic stroke, or with an atrial high rate episode >24 hours and two or more non-gender risk factors in CHA2DS2VASc.8 51 For other patients, an atrial high rate episode should prompt ECG monitoring, and consideration of anticoagulation if AF is then documented.

Neurogenic AF

A further complexity concerning AF diagnosed after stroke is that some of this may be caused by abnormal autonomic drive, possibly related to insular brain injury,52 and inflammation, especially if detected in the acute phase. This ‘neurogenic’ AF could be a transient phenomenon and not require long-term anticoagulation, unlike ‘cardiogenic’ AF due to structural heart disease.53 Although there is insufficient supporting evidence to change clinical practice, a recent report suggested that recurrent ischaemic stroke risk in patients diagnosed with AF after stroke is low compared with that of patients known to have AF before the stroke, and similar to that in patients in sinus rhythm.54 Repeat rhythm monitoring, delayed by several months, might help to stratify these patients’ longer-term risk.

Anticoagulation in the elderly

Beyond bleeding risk scores, increasing age is sometimes seen as a contraindication to anticoagulation, particularly in those over 75 years, with antiplatelets sometimes used instead. Older patients are at the highest risk of stroke, so have the most to gain from treatment, and the most to lose if anticoagulation is needlessly withheld. The BAFTA study67 showed warfarin to be superior to aspirin in stroke prevention in patients aged 75 or older (mean age 81.5±4.2 years), with a relative risk of all-cause stroke of 0.48 in the warfarin arm. The risk of haemorrhagic complications did not vary significantly between the two groups, and was low overall, with an annual risk of major haemorrhage of 1.9% with warfarin and 2.0% with aspirin (and 2.9% and 3.7%, respectively, in those >85 years). The relative efficacy of warfarin in ischaemic stroke prevention is maintained despite increasing age, so a greater absolute benefit is derived in older, and so higher risk, patients, outweighing a slower increase in intracranial haemorrhage risk. Overall, older patients seem to gain a greater net clinical benefit from anticoagulation.68 Secondary analyses of three pivotal DOAC trials (ARISTOTOLE, ROCKET-AF and RE-LY) show that older patients benefit similarly to younger patients from DOACs,69 70 although the use of dabigatran at the higher dose of 150 mg twice daily is associated with high rates of extracranial bleeding in those older than 80 years.71 We therefore do not recommend any upper age cut-off for considering anticoagulation.

Cognitive impairment, frailty and falls

These are common reasons for withholding anticoagulation, but this might be unjustified. Using a DOAC potentially avoids much of the difficulty of warfarin monitoring in cognitively impaired patients, but it might be challenging to measure adherence unless there is an available relative or carer to supervise its use. Frailty, usually defined as a decline in physiological reserve across multiple systems leading to increased vulnerability to stressors, is independently associated with reduced prescription of anticoagulants,72 but the limited evidence available does not support this: in the ORBIT-AF registry of nearly 10 000 patients, of whom over 500 met American Geriatrics Society criteria for frailty, survival was improved by anticoagulation even in frail patients, and frailty did not independently predict an increased risk of major bleeding or stroke.73 Falls are cited as a reason to withhold anticoagulation in over a quarter of patients with untreated AF.74 This is understandable: anticoagulation with warfarin triples the risk of subdural haemorrhage,75 thought often due to falls, and frequent falls quadruple the risk of traumatic intracranial haemorrhage.76 However, intracranial haemorrhage after a single fall is rare, even in anticoagulated patients,77 and the risk of subdural haemorrhage in patients with anticoagulated AF is very low compared with the risk of ischaemic stroke.75 As a result, it has been estimated that a patient with AF with a 6% annual ischaemic stroke risk would need to fall 295 times per year not to benefit from warfarin.78 Although this figure will be lower for patients with a lower ischaemic stroke risk, the average elderly patient who falls does so only 1.8 times each year, and most patients who fall will have an elevated stroke risk because of their age and comorbidities. Pragmatically, it is also not clear that being screened as ‘high falls risk’ or having fallen in the last year is associated with an increased risk of haemorrhage in anticoagulated patients.79 80 Therefore, we do not recommend withholding anticoagulation in patients who fall but do suggest a thorough assessment and modification of falls risk where possible.

Choosing between DOACs

Phase III trials showed some differences in DOACs’ performance against warfarin. Dabigatran 150 mg twice daily, edoxaban 60 mg and apixaban were superior for preventing stroke and systemic embolism, mainly driven by a reduced risk of intracranial haemorrhage, whereas rivaroxaban was non-inferior. Only dabigatran 150 mg twice daily was superior for preventing ischaemic stroke. Apixaban, edoxaban and dabigatran 110 mg twice daily were associated with a lower risk of major bleeding, and edoxaban 30 mg with a lower risk of gastrointestinal bleeding. Dabigatran 150 mg twice daily and rivaroxaban had higher risks of gastrointestinal bleeding. Given the absence of head-to-head randomised controlled trials, there have been attempts to infer their relative efficacy through network meta-analysis; that is, a systematic assessment of their performance against warfarin as a common comparator. In the largest analysis of this type, the authors concluded that apixaban offered the optimal balance of safety and efficacy, with dabigatran showing the greatest efficacy but a less favourable safety profile.88 We must interpret these results with caution. Although the authors showed a lack of effect modification by age, gender balance or CHADS2, they could not control for all possible sources of confounding. For instance, the ROCKET-AF trial of rivaroxaban included a substantially higher-risk population with more comorbidities than RE-LY or ARISTOTLE, not all of which are accounted for in the CHADS2 score. There is therefore still a need for high-quality randomised-controlled-trial evidence, though large numbers would be needed to compare treatments likely to have similar effect sizes. Until such evidence is available, we do not think there is sufficient evidence to make a general recommendation for a particular DOAC; rather, we suggest balancing the available evidence with patient characteristics and preferences in choosing an anticoagulant. For some patients, apixaban might be a reasonable choice, especially those with impaired renal function. We would consider dabigatran for patients with high ischaemic stroke risk and low bleeding risk but avoid it in patients with significant dyspepsia or previous major gastrointestinal bleeding. For patients who prefer once-daily dosing, edoxaban and rivaroxaban may aid adherence. Interactions with other medication should be considered. We recommend the recent American College of Chest Physicians guidelines for further advice on anticoagulant selection.63

Timing of anticoagulation

The optimal timing to start anticoagulation after a cardioembolic stroke is unclear. Based on studies of heparin, very early (<48 hours) anticoagulation increases the risk of symptomatic intracranial haemorrhage without reducing the risk of early recurrence (7–14 days), morbidity or mortality.89 Current practice is therefore to delay anticoagulation by up to 14 days. As the risk of recurrence in this time is around 5%, many clinicians anticoagulate earlier, according to the size of the infarct and presence of haemorrhagic transformation. The ‘1-3-6-12’ rule-of-thumb, based only on expert opinion, suggests anticoagulation on days 1, 3, 6 and 12, respectively, after transient ischaemic attack, minor, moderate and large infarcts.90 The lower risk of intracranial haemorrhage with DOACs may facilitate earlier anticoagulation, and observational studies do suggest that the risk of symptomatic intracranial haemorrhage in patients treated with DOACs within the first 5 days of ischaemic stroke is low, as is the risk of new asymptomatic haemorrhagic transformation,91 92 at least in patients with small infarcts or less severe strokes. Four large randomised controlled trials, OPTIMAS, TIMING, START and ELAN, will assess the benefit of early anticoagulation (<4 days) in patients with AF-related stroke, with OPTIMAS (due to recruit ~3500 participants throughout the UK from early 2019 (ClinicalTrials.gov identifier: NCT03759938).93

Management of recurrent ischaemic stroke

In this event, the first priority is to determine the patient’s eligibility for hyperacute treatment. In patients taking warfarin, point-of-care INR testing provides an immediate answer: an INR of 1.7 or less does not contraindicate thrombolysis.94 This is less straightforward with DOACs. Although they may influence some standard laboratory clotting indices (particularly the activated partial thromboplastin time), these are not reliable markers of the degree of anticoagulation. More valid assays, such as factor Xa concentrations (for apixaban, edoxaban and rivaroxaban) and dilute thrombin time (dabigatran) are not usually available quickly. In the absence of an assay result excluding a significant effect of a DOAC, US guidelines advise against thrombolysis unless it can be clearly established that the patient last took the medication more than 48 hours ago and has normal renal function.27 The European Society of Cardiology recommends also considering thrombolysis for patients who last took a DOAC between 24 and 48 hours ago, if renal function is normal and they are otherwise a good candidate.95 In patients not meeting these criteria, proceeding directly to mechanical thrombectomy should be considered in patients with thrombus within the basilar artery or proximal middle cerebral artery. Successful thrombolysis has been reported after dabigatran reversal. For patients taking factor Xa inhibitors, a pathway using a rapid anti-Xa activity assay (RivLev) to guide thrombolysis in patients taking rivaroxaban has been tested successfully in a single centre.96 Importantly, over half the patients included who would not have been eligible for thrombolysis based on last intake of DOAC had a RivLev result compatible with thrombolysis.97 We are introducing a similar pathway in our centre, and our practice in the absence of a factor Xa concentration reflects European guidelines.

Poor adherence is the most obvious cause for treatment failure. In this case, the underlying reasons should be addressed, which may involve continuing the existing treatment with measures to improve adherence or choosing a new oral anticoagulant more acceptable to the patient (for instance, a factor Xa inhibitor rather than dabigatran in patients with gastrointestinal side effects, a DOAC rather than warfarin in patients reluctant to undergo blood tests, or a once-daily rather than twice daily DOAC). If medication adherence is good, drug and dietary interactions should be considered. A few patients suffer a recurrent ischaemic stroke despite apparently therapeutic anticoagulation, perhaps due to an alternative mechanism such as atherosclerosis or small vessel occlusion. There is no accepted evidence-based strategy for such patients. Seeking any treatable alternative cause of stroke (eg, carotid stenosis) or stroke ‘mimic’ is essential. In those taking warfarin, a switch to a DOAC might be recommended. Increasing the target INR instead is associated with a high risk of bleeding. In those taking DOACs, an alternative preparation (with a different mechanism of action or more frequent dosing) could be used, or a switch to warfarin could be made if closer monitoring is thought desirable. Adding an antiplatelet is not generally recommended (at least not long term), as the concurrent use of aspirin with an anticoagulant was associated with increased bleeding but not a reduction in the risk of all-cause stroke and systemic embolism in the RE-LY98 and SPORTIF99 trials.

Management of oral anticoagulant–associated intracranial haemorrhage

The management of oral anticoagulant–associated intracranial haemorrhage entails supportive care, rapid reduction of systolic blood pressure to <140 mm Hg100 and reversal of anticoagulation. For patients taking warfarin, this is achieved with 5–10 mg vitamin K intravenously and replacing the vitamin K–dependent clotting factors (II, VII, IX and X) with four-factor prothrombin complex concentrate (eg, Beriplex, Octaplex) in preference to fresh-frozen plasma.101 For patients taking DOACs, there is no role for vitamin K supplementation, as DOACs act downstream of the vitamin K–dependent portion of the clotting cascade. Four-factor prothrombin complex concentrate has been used off-label, and is recommended by European Heart Rhythm Association guidelines,95 but there is only limited evidence for efficacy. For dabigatran, there is an available specific monoclonal antibody antidote, idarucizumab, which rapidly and safely normalised laboratory clotting indices and achieved normal intraoperative haemostasis in patients needing emergency surgery in an uncontrolled phase III study.102 We recommend the immediate use of this agent where available in dabigatran-related intracranial haemorrhage. A decoy protein reversal agent for factor Xa inhibitors, andexanet alfa, has been tested in a small clinical study103 and has recently been licensed in the USA. It remains expensive and clinically unproven for factor Xa inhibitor–related intracranial haemorrhage, with a potential procoagulant effect from binding tissue factor, which could account for an increased rate of thrombosis reported.102 Thus, if available, adexanet alfa can be considered for factor Xa–associated intracranial haemorrhage, ideally as part of a randomised controlled trial. Nevertheless, even without the use of a reversal agent, the outcome in patients with intracranial haemorrhage related to DOACs is no worse than that of patients with intracranial haemorrhage related to vitamin K antagonists in whom anticoagulation is reversed.104

Because of the early risk of haematoma expansion, anticoagulation is commonly withheld for at least 2 weeks after anticoagulant-related intracranial haemorrhage even in patients with a strong indication (most often AF). After this, the decision whether to restart anticoagulation is challenging. A meta-analysis of observational studies found that resuming anticoagulation reduces the risk of ischaemic stroke without increasing the risk of haemorrhage,105 with a median time to resumption of 10–39 days. Nationwide observational data from Sweden also suggest benefit from resuming anticoagulation, estimating the optimal timing to be 7–8 weeks after intracranial haemorrhage.106 Although it is likely that these results are subject to bias, with lower-risk patients more likely to be selected for resumption of anticoagulation, this does suggest that at least some patients can safely resume anticoagulation. Several randomised controlled trials—including SoSTART, APACHE-AF and PRESTIGE-AF—will provide more definitive evidence, and an individual patient data meta-analysis is planned through the COCROACH collaboration. In the interim, we suggest a careful re-evaluation of the bleeding and stroke risks, including location of the intracerebral haemorrhage (as the recurrence risk of lobar haemorrhage is about four times greater than non-lobar haemorrhage107) and MRI markers of cerebral amyloid angiopathy,19 careful control of modifiable risk factors and use of a direct oral anticoagulant in preference to warfarin. We strongly encourage randomisation of eligible patients into ongoing trials, but left atrial appendage occlusion may reasonably be considered if the risk of recurrent intracerebral haemorrhage is judged to be unacceptably high.

When anticoagulation is contraindicated

In reality, there is no agreement on what constitutes an absolute long-term ‘contraindication’ to oral anticoagulation but rather a spectrum of risk that can change over time. As clinical data indicate that 90% of all cardiac thrombi in non-rheumatic AF originate from the left atrial appendage,108 exclusion of the left atrial appendage from the circulation offers an alternative strategy for stroke prevention in patients in whom oral anticoagulant is thought to be unacceptably high risk. Surgical excision of the left atrial appendage is undertaken in patients undergoing cardiac surgery (eg, mitral valve or MAZE procedures) and is highly unlikely to be offered as an independent intervention purely to reduce stroke risk, whereas catheter-based left atrial appendage closure devices can achieve minimally invasive left atrial appendage occlusion.

The Boston Scientific Watchman is FDA approved. In two prospective randomised controlled trials comparing the Watchman to warfarin (PROTECT-AF109 and PREVAIL110), left atrial appendage occlusion was not overall inferior for the endpoint of ischaemic stroke and systemic embolism, although there was a higher rate of early, intervention-related complications including ischaemic stroke. Ischaemic events after the first 7 days were not significantly different, but rates of major bleeding and particularly haemorrhagic stroke were significantly lower in the groups undergoing left atrial appendage occlusion. The overall findings persisted over 5-year follow-up, indicating that the devices are comparable with warfarin in reducing the rate of ischaemic stroke but with a lower risk of haemorrhage (particularly intracerebral haemorrhage).111 The European Society of Cardiology AF Management guidelines8 recommend left atrial appendage occlusion for high-risk patients in whom warfarin is contraindicated (IIb indication, level of evidence B). The recent American College of Chest Physicians guidelines also suggest left atrial appendage occlusion in patients with a strong contraindication to oral anticoagulation.63

However, these trials of left atrial appendage occlusion were undertaken in predominantly intermediate-risk patients (CHA2DS2VASc 1–2) and excluded patients in whom warfarin was contraindicated. The Watchman also requires short-term oral anticoagulation and, based on currently available data, long-term antiplatelet therapy. This appears anomalous to the clinical guideline recommendations: it seems unlikely that many patients with a true contraindication to long-term anticoagulation would tolerate short-term anticoagulation and long-term antiplatelet therapy.112 Observational evidence suggests that dual antiplatelet therapy for 3–6 months in place of anticoagulation may be a safe alternative, even in patients with previous intracranial haemorrhage.113 114 However, given that long-term single antiplatelet therapy is still generally used, which arguably has a bleeding risk similar to DOAC therapy,115 it remains unclear whether left atrial appendage occlusion offers a clear benefit over contemporary medical therapy. It should also be noted that incomplete left atrial appendage occlusion may increase thromboembolic risk and so mandate anticoagulation,112 and that the insertion procedure is not without risk (notably, a roughly 3% risk across PREVAIL and PROTECT-AF of serious pericardial effusion requiring treatment). Therefore, though current international guidelines support left atrial appendage occlusion and it is an option that should be discussed with the cardiologists, the key clinical trials have yet to be performed to demonstrate its efficacy and safety in moderate-risk and high-risk patients.

There is also interest in whether electrophysiological ablation can reduce the risk of ischaemic stroke in people with AF. We believe this should be viewed as a potential adjunct to anticoagulation, not an alternative, not least due to the risk of AF recurrence.116 Although there is observational evidence to suggest ablation may reduce stroke risk,117 the recent CABANA trial of ablation against medical therapy (both with anticoagulation) did not meet its primary end-point of a composite of death, disabling stroke, major bleeding and cardiac arrest in an intention-to-treat analysis.118 While there was a high cross-over rate from ablation to medical treatment, and a positive result was obtained in a per-protocol analysis, publication of the detailed results is awaited, and per-protocol findings should be considered hypothesis generating.

Embolic stroke of undetermined source

Twenty to twenty-five per cent of ischaemic strokes remain cryptogenic despite routine investigations, including non-invasive monitoring. Many of these may be embolic, for instance from low-risk cardiac sources, aortic arch or non-occlusive carotid plaque, or covert paroxysmal AF.119 It has been argued that anticoagulation in these patients might be more effective in reducing stroke risk than an antiplatelet. The concept of ‘embolic stroke of undetermined source’ as a target for anticoagulation has been tested in three clinical trials. NAVIGATE ESUS was recently stopped early on the basis of a planned interim analysis showing an excess of bleeding with rivaroxaban 15 mg once daily compared with aspirin, and a low probability of a reduction in stroke risk.120 RE-SPECT ESUS, with a similar design using dabigatran, is yet to finally report121; however, data presented in abstract form at the World Stroke Conference in October 2018 reported the rate of recurrent stroke (primary outcome) was 4.1% per year with dabigatran and 4.8% per year with aspirin (HR 0.85; p=0.1). The rate of major bleeding was similar in both arms: 1.7% per year with dabigatran and 1.4% per year with aspirin. ATTICUS will test apixaban in a population enriched for probable cardioembolism, with radiological re-infarction at 12 months as the primary end point.122

As well as the echocardiographic markers (left atrial dilation, spontaneous left atrial appendage echo contrast, reduced left atrial appendage flow velocity) used for patient selection in ATTICUS, ECG markers of left atrial function may have a role in identifying patients with ‘embolic stroke of undetermined source’ more likely to benefit from anticoagulation. For example, increased p-wave terminal force, defined as the amplitude of the terminal negative deflection of the p-wave in lead V1 multiplied by its duration, is a risk factor for cryptogenic or cardioembolic stroke, even in patients without AF.123 This lends support to the concept of ‘left atrial cardiopathy’ as an independent cause of cardioembolic stroke and a potential target for future anticoagulation trials.43 However, in the absence of positive trial results, we recommend usual antiplatelet treatment combined with intensive investigation for established high-risk cardiac sources in patients with suspected but unproven cardioembolic stroke.

Patients with cardiovascular comorbidities

Whereas cardioembolism is most effectively prevented by anticoagulation, atherosclerotic disease affecting the coronary, cervical or peripheral arteries is usually treated with an antiplatelet, based on the pathophysiological concept of the ‘red’, fibrin-rich thrombus forming under slow-flow conditions and the ‘white’, platelet-rich thrombus forming on ulcerated plaque. People with AF commonly have arterial disease, sometimes prompting dual therapy. Large registry studies indicate that combining warfarin and an antiplatelet clearly increases bleeding risk124 without improving cardiovascular outcomes in patients with coronary or peripheral arterial disease.125 126 Post hoc analyses of ROCKET-AF suggest a similar pattern with rivaroxaban.127 The distinction between anticoagulation-responsive thrombi and antiplatelet-responsive thrombi is unlikely to be absolute—for instance, both warfarin and rivaroxaban have shown efficacy in place of an antiplatelet in treating coronary artery disease.128 129 Reflecting European Heart Rhythm Association guidelines, in patients with anticoagulated AF, we do not add an antiplatelet for secondary prevention in atherosclerotic disease. In patients with AF undergoing percutaneous coronary intervention, triple therapy with dual antiplatelets and a DOAC is generally recommended for 1 to 6 months (based on an assessment of atherothrombotic and bleeding risk), then dual therapy to complete 12 months, then anticoagulation monotherapy. In patients with severe or symptomatic carotid stenosis, endarterectomy should be performed in preference to stenting, with the addition of aspirin immediately before surgery, and for 10 days afterwards.95

The findings of the COMPASS trial129 raise a possible alternative approach to stroke prevention in patients with AF with atherosclerotic comorbidities. The combination of rivaroxaban 2.5 mg twice daily and aspirin improved cardiovascular outcomes, including ischaemic stroke risk, in a population with stable atherosclerotic cardiovascular disease compared with aspirin alone, whereas rivaroxaban 5 mg twice daily alone did not. These results may also have implications for patients with AF who suffer a recurrent ischaemic stroke despite therapeutic anticoagulation. However, both rivaroxaban doses used were low compared with that used for stroke prevention in AF, and very few patients with AF were included (392/27395). Randomised-controlled-trial evaluation of such an approach in patients with AF with atherosclerotic comorbidities, or stroke despite oral anticoagulation, might be of interest.