Can We Predict Stroke in Atrial Fibrillation?
Stroke prevention with appropriate thromboprophylaxis still remains central to the management of atrial fibrillation (AF). Nonetheless, stroke risk in AF is not homogeneous, but despite stroke risk in AF being a continuum, prior stroke risk stratification schema have been used to ‘artificially’ categorise patients into low, moderate and high risk stroke strata, so that the patients at highest risk can be identified for warfarin therapy. Data from recent large cohort studies show that by being more inclusive, rather than exclusive, of common stroke risk factors in the assessment of the risk for stroke and thromboembolism in AF patients, we can be so much better in assessing stroke risk, and in optimising thromboprophylaxis with the resultant reduction in stroke and mortality. Thus, there has been a recent paradigm shift towards getting better at identifying the ‘truly low risk’ patients with AF who do not even need antithrombotic therapy, whilst those with one or more stroke risk factors can be treated with oral anticoagulation, whether as well-controlled warfarin or one or the new oral anticoagulant drugs.
The new European guidelines on AF have evolved to deemphasise the artificial low/moderate/high risk strata (as they were not very predictive of thromboembolism, anyway) and stressed a risk factor based approach (within the CHA2DS2-VASc score) given that stroke risk is a continuum. Those categorised as ‘low risk’ using the CHA2DS2-VASc score are ‘truly low risk’ for thromboembolism, and the CHA2DS2-VASc score performs as good as—and possibly better—- than the CHADS2 score in predicting those at ‘high risk’. Indeed, those patients with a CHA2DS2-VASc score = 0 are ‘truly low risk’ so that no antithrombotic therapy is preferred, whilst in those with a CHA2DS2-VASc score of 1 or more, oral anticoagulation is recommended or preferred.
Given that guidelines should be applicable for >80% of the time, for >80% of the patients, this stroke risk assessment approach covers the majority of the patients we commonly seen in everyday clinical practice, and considers the common stroke risk factors seen in these patients. The European guidelines also do stress that antithrombotic therapy is necessary in all patients with AF unless they are age <65 years and truly low risk. Indeed, some patients with ‘female gender’ only as a single risk factor (but still CHA2DS2-VASc score of 1, due to gender) do not need anticoagulation, especially if they fulfil the criterion of “age <65 and lone AF, and very low risk”. In the European and Canadian guidelines, bleeding risk assessment is also emphasised, and the simple validated HAS-BLED score is recommended. A HAS-BLED score of ≥3 represents a sufficiently high risk such that caution and/or regular review of a patient is needed. It also makes the clinician think of correctable common bleeding risk factors, and the availability of such a score allows an informed assessment of bleeding risk in AF patients, when antithrombotic therapy is being initiated. Clin. Cardiol. 2012 DOI: 10.1002/clc.20969
Dr. Lip received an honorarium through an educational grant from Sanofi Aventis for his time and expertise spent writing this article. He has served as a consultant for Bayer, Astellas, Merck, AstraZeneca, Sanofi, BMS/Pfizer, Biotronik, Portola and Boehringer Ingelheim and has been on the speakers bureau for Bayer, BMS/Pfizer, Boehringer Ingelheim, and Sanofi-Aventis.
Atrial Fibrillation and Stroke
Atrial fibrillation (AF) is the most common sustained cardiac rhythm disorder. AF commonly coexists with cardiovascular risk factors, and the presence of these adds to the development of new onset AF, especially if inadequately managed,1 and to complications associated with AF, such as stroke.2
AF predisposes to stroke by its fulfillment of Virchow's triad of thrombogenesis,3 whereby thrombus formation occurs in the presence of “vessel wall abnormalities” (now recognized as representing structural heart disease or vascular disease), “abnormal blood flow” (as evidenced by stasis in association with this arrhythmia), and “abnormal blood constituents” (given the abnormalities of clotting and platelets that are present in AF).
Strokes in AF are associated with greater mortality and morbidity, with more disability, longer hospital stays, and lower rates of discharge to the patient's own home. Clearly, AF represents a major public health problem, with a significant impact on healthcare costs.
How Do We Reduce this Risk?
Compared with placebo, adjusted-dose oral anticoagulation (OAC) with warfarin reduces stroke risk by 64% (95% confidence interval [CI], 49%–74%) and all-cause mortality by 26% (95% CI, 3%–43%).4
In this meta-analysis, antiplatelet therapy reduced strokes by 22% (95% CI, 6%–35%) compared to control, which is a figure consistent with the reduction in stroke seen by use of antiplatelet therapy in non-AF patients with vascular disease.5,6 When the analysis by Hart et al4 is confined to the aspirin-only trials, aspirin showed a nonsignificant 19% (95% CI, −1% to 35%) reduction in the incidence of stroke, with no significant effect on mortality (relative risk reduction, 14% (−7 to 31).
Although there was no statistical heterogeneity among the trials, the stroke risk reduction with aspirin compared to placebo (−19% stroke risk reduction) was driven by the 1 single positive trial, the SPAF-1 trial,7 which reported a 42% stroke risk reduction with 325 mg aspirin daily, compared to placebo. This single trial had internal heterogeneity between the anticoagulation-eligible and anticoagulation-ineligible arms of the trial (94% vs 8% stroke risk reduction, respectively); also, aspirin was ineffective in those aged >75 years and did not prevent severe strokes. The SPAF-1 trial was stopped at an interim stage and its result may be exaggerated.
In contrast, OAC was associated with a 39% (95% CI, 0.22–0.50) risk reduction compared with antiplatelet therapy.4 More contemporary randomized trials very clearly confirm that aspirin is inferior to OAC for stroke prevention, and the risks of bleeding (including intracranial hemorrhage) with aspirin are not significantly different to OAC, especially in the elderly.8–10
In low-risk AF patients, 1 prospective randomized trial showed no difference between aspirin (150–200 mg daily; N = 426) and control for the primary endpoint of thromboembolism-related complications (cardiovascular death, symptomatic brain infarction, or transient ischemic attack), with a trend toward more major bleeding (and intracranial hemorrhage) in aspirin-treated patients.11 In this trial, the Japanese AF trial, there were 27 primary endpoint events (3.1% per year; 95% CI, 2.1–4.6) in the aspirin group versus 23 (2.4% per year; 95% CI, 1.5–3.5) in the control group, and treatment with aspirin caused a nonsignificant trend for an increased risk of major bleeding (1.6%) (vs controls, 0.4%; P = 0.101).
Aspirin-clopidogrel combination therapy was compared to warfarin for stroke prevention in 1 trial, but this was stopped early due to a clear superiority of warfarin over aspirin-clopidogrel for stroke prevention.12 When aspirin-clopidogrel was compared to aspirin alone in a trial of patients who had refused or were deemed inappropriate for warfarin,13 there was a 28% reduction in ischemic stroke with aspirin-clopidogrel compared to aspirin alone. However, the risk of major bleeding with aspirin-clopidogrel was approximately 2% per year, which was >50% higher compared to aspirin alone, and broadly comparable to major bleeding rates seen with OAC.
As the patient gets older, the absolute beneficial effect of OAC increases, while the effect of aspirin declines markedly.14 Of note, the risk of stroke (and vascular events) rises with increasing age, from age 65 years upward, but serious bleeding only showed a small rise with increasing age, with no marked difference between OAC and aspirin.
What Are the Risk Factors for Thromboembolism?
Whereas it is recognized that AF increases the risk of stroke and thromboembolism by 5-fold, we do increasingly appreciate that that this risk is not homogeneous and is altered by the presence of various stroke risk factors. Also, the risk profile changes over time. For example, in a 12-year follow-up study of patients with newly-diagnosed carefully-defined lone AF patients, the annual rate of thromboembolism and heart failure during follow-up was low (0.4% each), and 5 patients (1.4%) died; however, multivariate predictors of adverse outcomes, including thromboembolism, were arrhythmia progression (ie, from paroxysmal to permanent AF), development of cardiac diseases, and older age (all P<0.05).15
Two comprehensive systematic reviews from the Stroke in AF Working Group and the UK National Institute for Health and Clinical Evidence (NICE) guidelines have summarized the published evidence for various risk factors on stroke, largely based on nonwarfarin arms of clinical trials and a few epidemiological cohorts.16,17 These risk factors have been used to formulate stroke risk stratification schemes, such as the CHADS2 (Congestive heart failure, Hypertension, Age ≥75, Diabetes mellitus, and prior Stroke or transient ischemic attack [doubled]) score,18 as discussed further below.
These risk factors have been use to formulate and (perhaps artificially) categorize AF patients into low, moderate/intermediate, and high risk strata, so that the “high risk” subjects could at least be targeted for anticoagulation therapy with warfarin, which is an “inconvenient” (and sometimes dangerous) drug.
What are the common stroke risk factors in AF populations? The Stroke in AF Working Group analysis16 identified previous stroke/transient ischemic attack (TIA) (adjusted relative risk [RR] 2.5), age (RR 1.5/decade), hypertension (RR 2.0), diabetes (RR 1.8), and female gender (RR 1.6) as risk factors. Interestingly, history of heart failure was not a significant risk factor, but moderate-systolic left ventricular dysfunction is still an independent predictor of thromboembolism. The NICE guideline systematic review17 identified history of stroke or TIA, increasing age (from age 65 years upward), hypertension, and structural heart disease (left-ventricular dysfunction or hypertrophy) to be good predictors of stroke risk in AF patients. In this NICE review, the evidence regarding diabetes mellitus, gender, and other patient characteristics as stroke risk factors in AF was less consistent.
It is worth remembering that in many of the older trial datasets (and many of these trials only randomised <10% of AF patients who were screened), some common stroke risk factors have not been systematically looked for and/or recorded, and unsurprisingly, there are many inconsistencies in definition in the various different published studies. In addition, large epidemiological and cohort studies have provided much additional information on common stroke risk factors in everyday clinical practice, among AF populations, although some inconsistencies in definitions of stroke risk factors may still be evident.
In the Loire Valley AF Project, among nonanticoagulated AF patients with no CHADS2 risk factors (N = 1035), the impact of age was clearly seen, because the stroke/thromboembolic event rate per 100 person-years was 0.23 (95% CI, 0.08–0.72), 2.05 (1.07–3.93), and 3.99 (2.63–6.06), in those aged <65, 65 to 74, and ≥75 years, respectively. In both univariate and multivariate analyses, heart failure, previous stroke, and vascular disease were significantly associated with increased risk of stroke/thromboembolism.19
In 1 recent nationwide cohort study from Taiwan, risk factors for ischemic stroke in AF patients on multivariate analysis were age (where stroke risk increased from age ≥65 years; odds ratio [OR] = 1.338 for age 65–74 years vs age 20–64 years, P = 0.014; OR = 1.652 for age ≥75 years vs age <65 years, P<0.001), hypertension (OR = 2.656, P<0.001), diabetes mellitus (OR = 1.341, P = 0.005), heart failure (OR = 1.611, P<0.001), previous ischemic stroke or TIA (OR = 2.752, P<0.001), and peripheral arterial disease (PAD) (OR = 1.814, P = 0.006).20
In another Danish nationwide cohort study, the presence of vascular disease also increased the risk of thromboembolism significantly at 5 and 10 years of follow-up, with hazard ratios (HRs) of 2.04 and 2.22, respectively.21 In an epidemiological cohort, the Danish Diet, Cancer and Health (DCH) study, vascular disease (prior myocardial infarction [MI] and PAD) was an independent risk factor for the primary endpoint of “stroke or death” in patients with AF, even after adjustment for the commonly used CHADS2 risk score.22 There is no doubt that vascular disease is an important consideration in the risk assessment of stroke and cardiovascular events in AF patients.23
What Are the Different Scoring Systems for Stroke Prediction, How Do They Compare, and What Should We Use Now?
The CHADS2 score17 is the most commonly used and validated risk score for stroke in AF; this simple score was derived by amalgamation of the AF Investigators and SPAF-1 risk schema (both trial-based risk stratification schema, with the limitations associated with historical trial datasets mentioned above). The pros and cons of the CHADS2 score have been recently discussed, particularly its noninclusion of many common stroke risk factors.24 A recent systematic review and metaanalysis on validation of the CHADS2 clinical prediction rule to predict ischaemic stroke concluded that the pooled c statistic and calibration analysis suggested minimal clinical utility of the CHADS2 in predicting ischaemic stroke across all risk strata in AF.25 The authors even suggested that further validation of CHADS2 should perhaps be undertaken.
For example, stroke risk in AF rises from age 65 years upward, and does increase stroke by 1.5-fold per decade,15 as illustrated by other analyses described above. Thus, the age criterion in the CHADS2 score needed to be re-examined, and given that age is a powerful driver for stroke (and mortality) in AF, extra weight was needed for the older subject (eg, age ≥75 years), who was likely to benefit more from OAC, given that aspirin is no safer than OAC and inferior to OAC for stroke prevention.7,26
Also, vascular disease is a risk factor for stroke, with AF patients with PAD being at very high risk for stroke and death.20–22 As discussed above, OAC has a significant impact on reducing stroke and all-cause mortality in AF patients, compared to placebo or control.4 In the Loire Valley AF Project, vascular disease significantly improved the predictive ability of the CHADS2 score (net reclassification improvement [NRI] 0.40, integrated discrimination improvement [IDI] 0.031).19 Female gender has also been associated with a higher risk of stroke in some studies,27 but many of these studies have included much older female subjects, and clearly, young female subjects (eg, age <65 years) with lone AF would be considered at low risk.
To complement the CHADS2 score, the new European Society of Cardiology guidelines28 have deemphasized the low, moderate, and high risk stratification, given the poor predictive value of such artificial strata, and more emphasis was put on a risk factor–based approach. For example, in the 12-year follow-up analysis of lone AF (discussed above), the baseline CHADS2 score was not predictive for thromboembolism (c-statistic 0.50; 95% CI, 0.31–0.69).14
After all, any stroke risk factor when present with AF causes a stroke, and with the improved knowledge on how to “handle” warfarin (ie, much lower risk of bleeding and improved stroke reduction with well-controlled warfarin, as reflected by high time in therapeutic range) and the imminent arrival (and regulatory approval) of new oral anticoagulant drugs, such as the oral direct thrombin inhibitors and oral Factor Xa inhibitors, the emphasis should focus instead on identifying ‘truly low risk’ patients with AF who do not even need any antithrombotic therapy, whilst those with one or more stroke risk factors can be considered for oral anticoagulation therapy.24
Thus, the European Society of Cardiology (ESC) guidelines recommended use of the CHA2DS2-VASc (Congestive heart failure, Hypertension, Age ≥75 years [double points], Diabetes mellitus, previous thromboembolism [double points], Vascular disease, Age 65–74 years, and female gender) score (see Table 1) to complement the initial assessment with the CHADS2 score.28 The CHA2DS2-VASc score includes more of the common stroke risk factors seen in everyday clinical practice.
Table 1. Assessment of Stroke (CHA2DS2-VASc) and Bleeding Risk (HAS-BLED) in Atrial Fibrillation Patients28
|Congestive heart failure||1||Hypertension (systolic blood pressure >160 mm Hg)||1|
|Hypertension||1||Abnormal renal and liver function (1 point each)||1 or 2|
|Age ≥75 years||2||Stroke||1|
|Diabetes mellitus||1||Bleeding tendency/ predisposition||1|
|Stroke/TIA/TE||2||Labile INRs (if on warfarin)||1|
|Vascular disease (prior MI, PAD, or aortic plaque)||1||Elderly (eg, age >65 years)||1|
|Age 65–74 years||1||Drugs or alcohol (1 point each)||1 or 2|
|Sex category (ie, female gender)||1|| || |
|Maximum score||9||Maximum score||9|
Given that guidelines should be applicable for >80% of the time, for >80% of the patients, the stroke risk assessment approach in the ESC guideline is applicable to most of the patients we commonly seen in everyday clinical practice, and considers the common stroke risk factors in such patients. The ESC guidelines do stress that antithrombotic therapy is necessary in all patients with AF unless they are age <65 and truly low risk. Indeed, some AF patients with ‘female gender’ only as a single risk factor (thus, still CHA2DS2-VASc score of 1) do not need anticoagulation, especially if they fulfil the criteria of “age <65 and lone AF, so very low risk of stroke”.
The first derivation and validation of the CHA2DS2-VASc score was performed in an European cohort from the EuroHeart survey on AF,29 to provide European data for the European guideline. This initial analysis had various limitations (including a proportion of patients lost to follow-up), but other validations of the CHA2DS2-VASc score have since been published.
In a nationwide Danish cohort study of 73,538 hospitalized nonanticoagulated patients with AF in Denmark reported that in “low risk” subjects (CHA2DS2-VASc score = 0), the 1-year rate of thromboembolism per 100 person-years was 0.78 (0.58–1.04), in contrast to an event rate with the CHADS2 score of 1.67 (95% CI, 1.47–1.89).21 The c-statistics (a statistical measure of the predictive value of a risk score) with the CHA2DS2-VASc score clearly outperform the CHADS2 at 1, 5, and 10 years of follow-up (see Table 2).
Table 2. Event Rates and c-Statistics Comparing the CHADS2 and CHA2DS2-VASc Scores at 1-, 5-, and 10-Year Follow-upa
|At 1 year|
| CHADS2||1.7||4.8||12||0.72 (0.69–0.75)|
| CHA2DS2-VASc||0.78||2.0||8.8||0.85 (0.83–0.87)|
|At 5 years|
| CHADS2||1.3||3.7||8.3||0.80 (0.78–0.81)|
| CHA2DS2-VASc||0.69||1.5||6.0||0.88 (0.87–0.89)|
|At 10 years|
| CHADS2||1.2||3.6||8.0||0.81 (0.80–0.83)|
| CHA2DS2-VASc||0.66||1.5||5.7||0.89 (0.88–0.90)|
These c-statistic data are consistent with another recent study investigating how AF burden improved clinical stroke risk assessment, which reported that the c-statistics of the CHADS2 and CHA2DS2-VASc for predicting thromboembolism were 0.653 (95% CI, 0.50–0.81) and 0.898 (95% CI, 0.84–0.96), respectively, and that the c-statistic for the CHADS2 score could be improved by the addition of AF burden data.30
A further large validation study was performed in a United Kingdom cohort of 79,884 AF patients aged ≥18 years in the UK General Practice Research Database, who were followed for an average of 4 years.31 Again, low-risk subjects (CHA2DS2-VASc score = 0) were truly low risk (with annual stroke events <0.5%) with the CHA2DS2-VASc score.
In an analysis of the Stroke Prevention Using Oral Thrombin Inhibitor in Atrial Fibrillation (SPORTIF) trial population, an anticoagulated AF cohort (N = 7329 subjects),32 the CHA2DS2-VASc scheme correctly identified the greatest proportion of AF patients at high risk, and the negative predictive value (ie, the percent categorized as “not high risk” actually being free from thromboembolism) for CHA2DS2-VASc was 99.5%. Finally, Poli et al33 also reported a “real world” of 662 consecutive elderly anticoagulated AF patients, in which the CHADS2 and CHA2DS2-VASc schemes had the best c-statistics (0.717 and 0.724, respectively).
Notwithstanding the above comments, all the published stroke risk scores (including CHADS2 and CHA2DS2-VASc) are not very useful in the elderly, with only limited ability to predict the risk of stroke. Thus, the Birmingham Atrial Fibrillation in the Aged (BAFTA) Investigators recently concluded that given the systematic undertreatment of older people with anticoagulation, and the relative safety of warfarin versus aspirin in the elderly, there could be a pragmatic rationale for classifying all patients over 75 years of age as “high risk” until better tools are available.34
In summary, the CHA2DS2-VASc score clearly outperforms the CHADS2 in identifying the “truly low risk” subjects with AF, and is at least as good as—and possibly better than—the CHADS2 score in predicting the “high risk” subjects.
Do We Need to Take the Risk of Bleeding into Account, and How Is This Quantified?
Risk factors for bleeding on OAC have been identified from general anticoagulated cohorts, but few bleeding risk assessment tools have been derived and validated in AF populations.35
In the NICE guidelines systematic review, the following patient characteristics were identified as having supporting evidence for being risk factors for anticoagulation-related bleeding complications: advanced age; uncontrolled hypertension; history of myocardial infarction or ischemic heart disease; cerebrovascular disease, anemia, or a history of bleeding; and the concomitant use of other drugs such as antiplatelet agents.36 The presence of diabetes mellitus, controlled hypertension, and gender were not identified as significant risk factors for bleeding.
Many risk factors for anticoagulation-related bleeding are also risk factors for stroke in AF patients. Older bleeding risk tools were also complicated and difficult to use in everyday clinical practice.37 Thus, many prior guidelines have avoided formal recommendation of bleeding risk assessment using a specific bleeding risk score in AF patients.
More recently, the ESC guidelines28 and Canadian guidelines38 have recommended that bleeding risk assessment be done, and the simple HAS-BLED (Hypertension, Abnormal renal/liver function, Stroke, Bleeding history or predisposition, Labile international normalized ratio [INR], Elderly, Drugs/alcohol concomitantly) score should be used.
The HAS-BLED score (see Table 1) was initially derived and validated in an AF population (N = 3978) from the EuroHeart survey, in which the predictive accuracy in the overall population (c-statistic 0.72) was consistent when applied in several subgroups.39 Also, application of HAS-BLED had c-statistics of 0.91 and 0.85, in patients receiving antiplatelet agents alone or no antithrombotic therapy, respectively.
The HAS-BLED score has also been validated in a clinical trial population, as well as in a nationwide cohort study. In the analysis from the SPORTIF trial dataset,40 multivariate analyses demonstrated that significant predictors of bleeding were concurrent aspirin use (HR 2.10; 95% CI, 1.59–2.77; P<0.001); renal impairment (HR 1.98; 95% CI, 1.42–2.76; P<0.001); age ≥75 years (HR 1.63; 95% CI, 1.23–2.17; P = 0.0008); diabetes (HR 1.47; 95% CI, 1.10–1.97; P = 0.009), and heart failure or left ventricular dysfunction (HR 1.32; 95% CI, 1.01–1.73; P = 0.041). Of the tested schemas, the new HAS-BLED score performed best, with a stepwise increase in rates of major bleeding with increasing HAS-BLED score (P-trend < 0.0001).
In the nationwide Danish cohort,41 the c-statistics for the HAS-BLED and HEMORR2HAGES (Hepatic or renal failure, Ethanol abuse, Malignancy, Older [age >75 years], Reduced platelet count or function, 2 points for Rebleeding risk Hypertension [uncontrolled], Anemia, Genetic factors, Excessive fall risk [including neurodegenerative and psychiatric disorders] and history of Stroke) schemes were 0.795 (0.759–0.829) and 0.771 (0.733–0.806), respectively, with comparable results found in AF patients not receiving OAC. Thus, in this unselected nationwide cohort of hospitalized patients with AF, the HAS-BLED score performed similar to HEMORR2HAGES in predicting bleeding risk, but the HAS-BLED score is much simpler and easier to use in everyday clinical practice.
How should we use this bleeding risk score? In the ESC guidelines, a HAS-BLED score of ≥3 represents a sufficiently high risk such that caution and/or regular review of a patient is needed. It also makes the clinician think of correctable common bleeding risk factors; eg, improving blood pressure control and labile INRs would reduce points on the HAS-BLED score. The availability of such a score allows informed decisions when deciding whether to use a low-dose or high-dose regime of new OACs such as dabigatran.28
Bleeding risk is clearly multifactorial and apart from those with risk factors for bleeding, there are high-risk periods when OAC is initiated.42 Given the importance of bleeding risk assessment and management in AF patients, the European Heart Rhythm Association recently published a position document, endorsed by the ESC Working Group on Thrombosis, with the objective of summarizing “best practice” when approaching antithrombotic therapy in AF patients.44 This document addresses the epidemiology and size of the problem of bleeding risk in AF and reviews established bleeding risk factors. Patient values and preferences in balancing the risk of bleeding against thromboembolism are considered, and the prognostic implications of bleeding are discussed. The document also reviews bleeding risk stratification and currently published bleeding risk schema.
How should we approach stroke risk assessment in AF? By being more inclusive, rather than exclusive, of common stroke risk factors in AF, we can clearly do much better in reducing stroke and mortality in AF, which are the 2 main endpoints that are significantly reduced by oral anticoagulation, compared to placebo/control. Indeed, those categorized as “low risk” using CHA2DS2-VASc as recommended in the ESC guidelines were “truly low risk” for thromboembolism, and the CHA2DS2-VASc scheme performs as good as—and possibly better than—the CHADS2 score in predicting those at “high risk.”
The availability of the new OAC offers exciting new possibilities43 so that once we identify the “truly low risk” patients (eg, CHA2DS2-VASc score = 0), OAC is recommended or preferred for AF patients with ≥1 stroke risk factors (CHA2DS2- VASc ≥1).45
Risk assessment must also consider bleeding risk when initiating OAC therapy. Thus, the HAS-BLED score is simple and is recommended as an informed assessment of bleeding risk in AF patients when antithrombotic therapy is being initiated.