Atrial fibrillation (AF) is associated with an increased risk of ischemic stroke, and oral anticoagulants (OACs) with vitamin K antagonist (VKA), e.g. warfarin, or non-VKA OACs (NOACs), could effectively reduce the risk of AF-associated stroke.1)2) Compared to warfarin, NOACs were much safer regarding the risk of intra-cranial hemorrhage (ICH) in the pooled meta-analysis of 4 pivotal radomized trials.2) In the real-world daily practice, the annual risk of ICH of NOAC-treated patients was lower than 1% even for those aged >90 years.3) Actually, the introduction of NOACs has transformed the landscape for stroke prevention and led to a better clinical outcome in AF patients.4) For stroke prevention in AF, it is crucial to assess the stroke and bleeding risks of individual AF patients to make appropriate decisions for OAC use and achieve optimal patient care. This review article aimed to provide an overview on the correct use of CHA₂DS₂-VASc and HAS-BLED scores, and to update important revisions about the definitions of some of the CHA₂DS₂-VASc components. Also, data about the biomarkers and potential roles of AF duration and left atrial functions in the prediction of stroke in AF were also discussed.

Four important international guidelines, including 2017 Asia Pacific Heart Rhythm Society, 2018 American College of Chest Physicians (ACCP), 2019 American College of Cardiology/American Heart Association (ACC/AHA)/Heart Rhythm Society and 2020 European Society of Cardiology (ESC), all recommend the use of CHA₂DS₂-VASc score for stroke risk assessment (Table 1).5)6)7)8)9) There were some differences regarding the definitions of “C: congestive HF” and “V: vascular diseases” components between the original version and revised version of CHA₂DS₂-VASc score in 2020 ESC guidelines.

The suggestions of the bleeding scheme used for bleeding risk assessment are listed in Table 1.5)6)7)8) Most guidelines recommend the use of the HAS-BLED score for bleeding risk assessment, while the ACC/AHA did not propose any specific bleeding risk scheme.9) HAS-BLED score has been well-validated in Asian AF patients which could predict the occurrences of major bleeding or ICH more accurately than other bleeding prediction schemes.22) A major concern about the use of bleeding schemes is that physicians may inaccurately regard a high bleeding risk score as a reason to withhold OACs. The 2020 ESC guidelines clearly stated that the HAS-BLED score should be considered to help address modifiable bleeding risk factors, and to identify patients at high risk of bleeding (HAS-BLED score >3) for early and more frequent clinical review and follow-up (class IIa recommendation)”, rather than being used to guide treatment decisions to use OAC or not (class III recommendation).8) Actually, in the mobile AF application (mAFA-II) randomized trial, HAS-BLED score was monitored prospectively using mAFA, and calculated as 30 days, days 31–60, days 61–180, and days 181–365.23) Among 1077 (60%) patients who had 4 bleeding risk assessments, incident bleeding events decreased significantly from days 1–30 to days 181–365 (1.2% to 0.2%, respectively, p<0.001), while the total OAC usage increased from 63.4% to 70.2% (p trend<0.001).23) A recent study from Taiwan also showed that for AF patients with a low bleeding risk (HAS-BLED score 0–2) who received OACs, the continuations of OACs once their HAS-BLED scores increased to >3 were associated with a lower risk of ischemic stroke (HR, 0.60) and all-cause mortality (HR, 0.88).24) All these data support the concept that a high HAS-BLED score itselft should not be the sole reason to withhold OACs for stroke prevention.

When assessing the stroke risk of AF patients, it is important to recognize the dynamic nature of stroke risk of AF patients.25)26)27)28) Among newly-diagnosed AF patients initially with a CHA₂DS₂-VASc score of 0 (males) or 1 (females) who were not indicated for OACs, around 16% of them would have a CHA₂DS₂-VASc score >1 (males) or >2(females) at 1 year,29) and the use of OACs once their CHA₂DS₂-VASc scores increased was associated with a better net clinical outcome.28) The follow-up CHA₂DS₂-VASc scores or its changes (delta CHA₂DS₂-VASc scores) were more accurate than the baseline scores in the prediction of ischemic stroke.26) Until now, there were no high-quality prospective studies investigating the optimal frequency of stroke risk re-assessment and subsequent changes of stroke prevention strategy. The only available data were reported from Taiwan AF cohort showing that for AF patients who acquired new stroke risk factors, 80% of them would acquire these co-morbidities after 4.2 months of AF diagnosis. The duration from the acquirement of incident co-morbidities to the occurrence of ischemic stroke was longer than 4.4 months for 90% of the patients.28) Therefore, 3–4 months may be a reasonable timing interval at which stroke risk should be reassessed so that OACs could be prescribed in a timely manner for stroke prevention.28) However, the optimal and reasonable frequency of stroke risk re-assessement should also incorporate compliance of patients and cost-benefit ratio which would vary widely in diverse healthy care systems. Several key points about the use of CHA₂DS₂-VASc and HAS-BLED scores for stroke and bleeding risk assessments are listed in Figure 1.

In most important international guidelines, OACs were not recommended for patients with a CHA₂DS₂-VASc score of 0 (males) or 1 (females), and were recommended/indicated for those with a CHA₂DS₂-VASc score >2 (males) or >3 (females).9) However, the recommendations about use of OACs for AF patients with single one stroke risk factor beyond gender (CHA₂DS₂-VASc score 1 for males and 2 for females) are different between 2020 ESC (class IIa recommendation) and 2019 ACC/AHA (class IIb recommendation) guidelines. For Asian AF patients, a more personalized approach may be helpful since these AF patients were not the homogenous population regarding their stroke risks.30) There are 2 important points should be considered before determining the use of NOACs for Asian AF patients with a CHA₂DS₂-VASc score 1 (males) or 2 (females). First, which kind of the non-sex stroke risk factor contributes to the CHA₂DS₂-VASc score and second, how old is the patient. It is well recognized that different risk factors were associated with different stroke risks,31)32) with age being as a very important determinant.17)33) Therefore, different age thresholds should be considered for the use of NOACs for patients with different comorbidities. Based on the recent report, the proposed age thresholds were as follows: age 35 years for HF, 50 years for hypertension or diabetes, and 55 years for vascular disease,30)33) which have been also mentioned in 2020 ESC guideines.8) The considerations about the use of NOACs for Asian AF patients with single one stroke risk factor beyond gender are summarized in Figure 2. These proposed age thresholds should be validated and tested in further studies from various dataset.

Other than the components of the CHA₂DS₂-VASc scheme, some biomarkers and several parameters derived from imaging tools which were reported to be associated with stroke events in AF patients are summarized in Table 2.34)35)36)37)38)39)40)41)42)43)44)45)46)47)48)49)50)51)52) A biomarker-based stroke risk score, the ABC (age, biomarkers [high-sensitivity troponin and N-terminal pro B-type natriuretic peptide; NT-proBNP], and clinical history of prior stroke/transient ischemic attack) score, has been developed to predict stroke or SE in 14,701 anticoagulated AF patients enrolled in ARISTOTLE trial, and was externally validated in 1,400 participants with AF or atrial flutter (689 on OACs) in the STABILITY trial.40) The ABC-stroke score yielded higher c-indices than the CHA₂DS₂-VASc score in both the derivation cohort (0.68 vs. 0.62, p<0.001) and the external validation cohort (0.66 vs. 0.58, p<0.001).40) The ABC-stroke score were subsequently validated in patients enrolled in RE-LY and ENGAGE AF-TIMI 48 trials.41)42) However, the follow-up duration was often short in the trial cohort; whether the baseline ABC-stroke score could predict long-term outcomes of patients was not clear. In a study from Spain analyzing 1,125 AF patients who were stable on VKAs and followed-up for a median of 6.5 years, the ABC-stroke score does not offer significantly better predictive performance compared with the CHA₂DS₂-VASc score.53) Furthermore, levels of biomarkers, such as NT-proBNP, could change significantly over time. In the nested prospective biomarker study from ENGAGE AF-TIMI 48 trial which included 6308 patients, levels of high-sensitivity troponin T (hsTnT) and NT-proBNP were measured at baseline and 12 months.54) The results demonstrated that hsTnT was dynamic in 46.9% (≥2 ng/L change) and NT-proBNP in 51.9% (≥200 pg/mL change) during 12 months, and re-assessment of ABC-stroke risk score at 12 months accurately reclassified a significant proportion of patients compared with their baseline risk (net reclassification improvement 0.50; 95% confidence interval, 0.36–0.65).54) These findings demonstrated that how biomarkers were dynamic, just like the changes of CHA₂DS₂-VASc scores, and should also be re-assessed. Given the costs of the measurements of biomarkers and the uncertainty of the intervals at which biomarkers should be rechecked, the CHA₂DS₂-VASc score should still be the clinically preferred tool for stroke risk assessment unless further studies could clearly prove that ABC-stroke score-guided OAC treatments are better than CHA₂DS₂-VASc score-based approach.

Artificial intelligence (AI) is developing rapidly and may improve stroke risk stratification in AF. However, in an analysis of the ORBIT-AF and GARFIELD-AF registries including 74,792 AF patients, machine learning techniques did not improve prediction modelling of death, major bleeding, or stroke.52) More efforts are necessary to see whether more advanced AI models could improve risk prediction of AF patients and change clinical practice.

The personalized and precision evaluation of stroke risk of each AF individuals is the future of stroke prevention in AF, and the AF duration and left atrial function may play an important role. Although the current stroke risk stratification scheme did not take the duration of AF into consideration, more and more data suggesting that the stroke risk may be a function of AF duration and clinical comorbidities. In a recent study which included 21,768 non-anticoagulated patients with cardiovascular implantable electronic devices, both increasing AF duration (p<0.001) and increasing CHA₂DS₂-VASc score (p<0.001) were signifcantly associated with annualized risk of ischemic stroke and SE (SSE).55) When patients were categorized into 3 groups according to the maximum daily AF duration (no AF, 6 minutes to 23.5 hours, and >23.5 hours), a interaction between AF duration and CHA₂DS₂-VASc scores were observed, as shown in Figure 3.55) For patients with a low CHA₂DS₂-VASc score (0 or 1), the risk of SSE was lower than 0.9%/year even for those with a maximum AF duration longer than 23.5 hours, and therefore, NOACs were generally not necessary. On the contrary, for patients with a high CHA₂DS₂-VASc score (>5), NOACs should be considered even their AF burdens were low. Most importantly, information about AF duration may be able to further stratify patients with a middle-range CHA₂DS₂-VASc score (2 or 3–4) into different groups with a SSE rate higher or lower than 0.9%/year. Further studies should be performed to understand whether the incorporations of AF duration into the CHA₂DS₂-VASc score could really improve stroke risk assessment.

Atrial cardiomyopathy was defined as “any complex of structural, architectural, contractile or electrophysiological changes affecting the atria with the potential to produce clinically-relevant manifestations”,56)57) and left atrium (LA) could be regarded as a “target organ” which would suffer from damages due to systemic diseases (like hypertension, HF, and diabetes). Peak atrial longitudinal strain derived from 2-dimensional speckle tracking echo (LA strain) is a more sensitive marker to detect early functional remodelling than overt structural change, such as LA dilatation. Several studies regarding the potential role of LA strain in stroke risk stratifications have been published and summarized in Figure 4.50)51)58) A lower LA strain is commonly observed in AF patients irrespective of different age strata compared to normal subjects,58) and independently associated with an increasec risk of ischemic stroke (7.8% lower risk of stroke per 1% increase in LA strain).50) The degree of decrease of each AF individual's LA strain from reference values can be applied in a stratified manner as standard score for organized risk stratification.51) In the previous study performed by Liao et al.,51) AF patients with a mean LA strain lower than reference values (mean values of LA strain of healthy subjects) for more than 3 standard deviations were associated with a more than 2-fold higher risk of ischemic stroke compared to those whose LA strain was lower for only 2–3 standard deviations. Besides, the consecutive 3-beat averaged value of LA strain is a feasible and reproducible method to overcome the irregular cardiac cycle of AF and correlated well with the index-beat method and 10-second average method.50) Further integration of LA strain with current risk stratification schemes might be a potential manner for prognostication in the era of individualized precision medicine.

Stroke and bleeding risks of AF patients should be assessed using CHA₂DS₂-VASc and HAS-BLED scores, respectively, and should be regularly re-assessed. The AF duration and left atrial function may play an important role for personalized evaluation of individual stroke risk while more studies are necessary.