Atrial fibrillation (AF) is the most common arrhythmia and carries a significant risk of stroke and thromboembolism; hence, oral anticoagulation (OAC) is recommended in guidelines as a fundamental pillar for its appropriate management.1,2

Nonetheless, OAC use requires stroke risk stratification, which should also be balanced with bleeding risk assessment, using clinical risk scores such as the CHA2DS2-VASc and HAS-BLED scores,3,4 as recommended by most clinical practice guidelines. Unfortunately, such risk assessment is usually performed at baseline, as a one-off evaluation that views risk as static process, and these values are usually applied to predict clinical outcomes that occur many years later.

However, risk factors for stroke and bleeding are not static over time, but are rather dynamic in nature, and most AF patients will develop at least 1 new risk factor before presenting with a thromboembolic event.5 These dynamic changes may increase the scores of the CHA2DS2-VASc and HAS-BLED assessed initially, thus modifying the absolute risk (and rate) of stroke and bleeding. Hence, the baseline estimated risk of these outcomes may worsen, due to aging and other incident comorbidities.6,7

The aim of the present study was to investigate whether dynamic evaluation of CHA2DS2-VASc and HAS-BLED scores over time would improve risk prediction in a cohort of AF patients taking OAC therapy and who were prospectively enrolled in the Murcia AF Project.

From May 1, 2007, to December 1, 2007, we consecutively included adult outpatients with permanent or paroxysmal AF who were stable on OAC therapy with vitamin K antagonists in the preceding 6 months (ie, INR from 2-3, so the time in therapeutic range at enrolment was 100%) attending our anticoagulation clinic. Patients with prosthetic valves or rheumatic AF were excluded, as well as those who had had an acute coronary syndrome, stroke, surgical interventions, hospitalizations, or any hemodynamic instability in the previous 6 months.

The study protocol was approved by the Ethics Committee of Hospital General Universitario Morales Meseguer and was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki. All patients gave informed consent for participation.

This study included 1361 patients, of which 693 (50.9%) were women, with a median age of 76 [IQR 71-81] years. A summary of other baseline characteristics is shown in table 1.

The mean baseline CHA2DS2-VASc was 4.0±1.7, whereas the mean CHA2DS2-VASc scores at 2 years and at 4 years were 4.3±1.6 and 4.5±1.6, respectively. Compared with baseline CHA2DS2-VASc, CHA2DS2-VASc at 2 years and at 4 years were significantly higher (both P <.001). CHA2DS2-VASc at 4 years was also significantly higher than CHA2DS2-VASc at 2 years (P <.001). Accordingly, the proportions of participants categorized as low, intermediate or high risk by CHA2DS2-VASc score at baseline were 2.1% (28), 4.8% (65) and 93.2% (1268), respectively. Of the 1062 patients who were alive at 4 years, 8 (0.8%) were categorized as low risk, 24 (2.3%) as intermediate risk and 1030 (97.0%) as high risk by recalculated CHA2DS2-VASc score (figure 1).

Figure 1.

Alluvial plots showing baseline stroke and bleeding risk stratification (baseline CHA2DS2-VASc and HAS-BLED), and reclassification into different risk categories during follow-up. Green, low risk of stroke or bleeding; orange, moderate risk of stroke or bleeding; red, high risk of stroke or bleeding.

(0.26MB).

The mean HAS-BLED score was 2.7±1.2 at baseline, 2.9±1.3 at 2 years and 3.1±1.2 at 4 years. Compared with baseline HAS-BLED, HAS-BLED at 2 years and at 4 years were significantly higher (both P <.001). HAS-BLED at 4 years was also significantly higher than HAS-BLED at 2 years (P <.001). At baseline, the proportions of participants categorized as low, moderate or high risk by HAS-BLED score were 10.4% (n=141), 28.4% (n=387) and 61.2% (n=833), respectively. Of the 1062 patients who were alive at 4 years, 68 (6.4%) were categorized as low risk, 318 (29.9%) as moderate risk, and 676 (63.7%) as high risk by the recalculated HAS-BLED score (figure 1).

Primary outcomes and predictive abilities

During 6 years of follow-up, 156 (11.5%) patients had an ischemic stroke/TIA and 269 (19.8%) had a major bleeding event. In addition, 472 (34.68%) patients died. The period with the highest incidence rates for ischemic stroke/TIA and major bleeding was from 2 to 4 years, which showed a significantly higher incidence rate ratio compared with the others for major bleeding outcomes. Complementary information on clinical outcomes and incidences among the different periods is shown in appendix C and table 1 of the supplementary data.

The probability of experiencing an ischemic stroke/TIA increased with the use of the recalculated CHA2DS2-VASc, with the 4-year assessment showing the highest higher hazard ratio even after adjustment (table 2 of the supplementary data). Regarding the predictive performance for ischemic stroke/TIA, the CHA2DS2-VASc recalculated at 2 years had significantly higher predictive ability during the period from 2 to 4 years compared with the baseline CHA2DS2-VASc, whereas IDI and NRI showed improvements in sensitivity and reclassification (table 2). Similarly, the CHA2DS2-VASc recalculated at 4 years yielded significantly better predictive performance for ischemic stroke/TIA during the period from 4 to 6 years compared with the baseline CHA2DS2-VASc (figure 2). Again, IDI reported improvement in sensitivity and there was an enhancement of the reclassification ability based on NRI (table 2).

Table 2.

C-indexes, c-indexes comparison, IDI and NRI of the dynamic CHA2DS2-VASc compared with the original score

Risk score assessment	C-index	95%CI	Z score*	P*	IDI	95%CI	P	NRI	95%CI	P
Ischemic stroke/TIA at 4 y
CHA2DS2-VASc at baseline	0.604	0.576-0.631	-	-	-	-	-	-	-	-
CHA2DS2-VASc at 2 y	0.701	0.675-0.727	3.628	<.001	0.014	0.007/0.020	<.001	0.677	0.427/0.926	<.001
Ischemic stroke/TIA at 6 y
CHA2DS2-VASc at baseline	0.682	0.653-0.710	-	-	-	-	-	-	-	-
CHA2DS2-VASc at 2 y	0.670	0.640-0.697	0.889	.374	0.002	-0.001/0.004	.211	0.209	-0.092/0.511	.173
CHA2DS2-VASc at 4 y	0.761	0.734-0.786	2.234	.026	0.030	0.016/0.044	<.001	0.757	0.496/1.018	<.001

95% CI, 95% confidence interval; IDI, integrated discrimination improvement; NRI, net reclassification improvement.

*

For c-index comparison.

Figure 2.

Receiver operating characteristic curves of the baseline and dynamic (at 2 and 4 years) CHA2DS2-VASc/HAS-BLED scores for the prediction of ischemic stroke/TIA or major bleeding. Blue line, baseline scores; green line, scores recalculated at 2-years; red line, scores recalculated at 4-years.

(0.13MB).

As for stroke, the probability of major bleeding was higher with the use of the recalculated HAS-BLED score (table 3 of the supplementary data). At 2 years, the recalculated HAS-BLED score also showed a significantly higher predictive ability than the baseline HAS-BLED score for major bleeding events during the period from 2 to 4 years. IDI and NRI demonstrated significant improvements compared with the baseline HAS-BLED score (table 3). For major bleeds from 4 to 6 years, the c-index of the HAS-BLED score recalculated at 4 years was numerically higher but not statistically significant compared with baseline. There was a slight enhancement in sensitivity as assessed by the IDI but was not significantly better when reclassified by NRI (table 3).

Table 3.

C-indexes, c-indexes comparison, IDI and NRI of the dynamic HAS-BLED compared with the original score

Risk score assessment	C-index	95%CI	Z score*	P*	IDI	95%CI	P	NRI	95%CI	P
Major bleeding at 4 y
HAS-BLED at baseline	0.663	0.632-0.693	-	-	-	-	-	-	-	-
vs HAS-BLED at 2 y	0.709	0.680-0.738	2.987	.003	0.016	0.006/0.026	.001	0.444	0.287/0.600	<.001
Major bleeding at 6 y
HAS-BLED at baseline	0.623	0.593-0.652	-	-	-	-	-	-	-	-
vs HAS-BLED at 2 y	0.613	0.582-0.642	0.467	.640	-0.001	-0.003/0.001	.419	-0.063	-0.246/0.119	.498
vs HAS-BLED at 4-years	0.631	0.601-0.660	0.318	.751	0.009	0.001/0.016	.018	-0.002	-0.174/0.170	.977

95% CI, 95% confidence interval; IDI, integrated discrimination improvement; NRI, net reclassification improvement.

*

For c-index comparison.

DCAs would help in the estimation of patients who will experience any of the primary endpoints, based on the predictions of the baseline risk scores in comparison with those recalculated at 2 and 4 years of follow-up. In figure 3, the prediction models that are the farthest away from the dashed black line (ie, assume all events) and the horizontal black line (ie, assume none event) had the highest net benefit. DCAs demonstrated that using the dynamic CHA2DS2-VASc and HAS-BLED scores was clinically useful and provided an overall improvement in the net benefit for the prediction of ischemic stroke/TIA and major bleeding, respectively. A detailed estimate of the net benefit at each threshold of probabilities from 0% to 20% is shown in table 4. A summary of the study findings is shown in figure 4.

Figure 3.

Decision curve analysis of the baseline and dynamic (at 2 and 4 years) CHA2DS2-VASc/HAS-BLED scores for ischemic stroke/TIA or major bleeding. Solid black line, assumes all patients will suffer an adverse event; dashed black line, assumes no patient will suffer an adverse event; blue line, baseline scores; green line, scores recalculated at 2-years; red line, scores recalculated at 4-years.

(0.18MB).

Table 4.

Net benefits for baseline, 2-year, and 4-year CHA2DS2-VASc/HAS-BLED at different threshold probabilities

Threshold probability	Ischemic stroke/TIA				Major bleeding
	Net Benefit for CHA2DS2-VASc				Net Benefit for HAS-BLED
	Atbaseline	At 2 y	At 4 y	Difference in net benefit(baseline vs recalculated at 4 y)	Atbaseline	At 2 y	At 4 y	Difference in net benefit(baseline vs recalculated at 4 y)
5%	0.13%	0.12%	1.32%	1.20%	2.00%	1.88%	2.20%	0.20%
10%	0.00%	0.07%	0.23%	0.23%	0.36%	0.04%	0.15%	−0.20%
15%	0.00%	0.00%	−0.07%	−0.07%	−0.01%	0.00%	0.01%	0.02%
20%	0.00%	0.00%	0.05%	0.05%	0.00%	0.00%	0.24%	0.24%
Overall improvement				1.41%				0.24%

Figure 4.

Central illustration. TIA, transient ischemic attack.

(0.37MB).

In this real-world cohort study, our principal findings are as follows: a) consecutive reassessment of stroke and bleeding risks through CHA2DS2-VASc and HAS-BLED scores demonstrated a significantly higher predictive ability and net benefit compared with the baseline scores; and b) the risks of stroke and bleeding in AF patients are dynamic and change during follow-up.

The incidence of new risk factors and the temporal trends in the CHA2DS2-VASc score have been previously investigated. Chao et al.5 included 14 606 AF patients who did not receive antiplatelet agents or OACs, with a baseline CHA2DS2-VASc score of 0 (men) or 1 (women) and with incident risk factors, and observed a dynamic increase in the CHA2DS2-VASc score.5 Indeed, this often translated into a dynamic change in the risk category.6,11,12 A recent study demonstrated that changes in CHA2DS2-VASc score over time were associated with the incidence of stroke, thus indicating that stroke risk is not static.13 In fact, the CHA2DS2-VASc score of AF patients would increase as patients become older and they accumulate more comorbidities, and we observed a significant progressive increase in the CHA2DS2-VASc score in accordance with previously published literature,5,6,11 again emphasizing the dynamic nature of thromboembolic risk.

Although the concept or risk reassessment is well accepted.5,14,15 prior clinical guidelines did not provide clear and concise recommendations on this issue. The landscape has changed with the publication of the latest international guidelines, which emphasize the need for dynamic assessment, at least annually, of thromboembolic risk.1,2,16–19 However, the time interval in which thromboembolic risk should be reassessed is controversial. In a study published by Chao et al.,5 the CHA2DS2-VASc score in patients initially classified as low risk increased by approximately 12.1%/y. The authors therefore suggest that, among low-risk patients, stroke risk should be reassessed every 4 months, with the aim of prescribing OAC therapy in those with an increased CHA2DS2-VASc score.

Regarding bleeding risk, guidelines recommend the use of HAS-BLED for the evaluation of bleeding risk, suggesting its frequent reassessment with particular attention to modifiable bleeding risk factors in patients with a high bleeding risk (HAS-BLED ≥3).1,2,16,18–21 Although the dynamic nature of the variables associated with bleeding is well accepted, few studies have specifically investigated this. As with stroke risk, bleeding risk assessment is often done at baseline only, at the beginning of OAC therapy, while bleeding events can be observed many years later. This may reflect that bleeding risk assessment has been subject to misuse and misinterpretation, and modifiable bleeding risk factors should be addressed as part of a holistic approach to AF patient assessment and management.22 Bleeding risk stratification should be used to flag patients with a high risk of severe bleeding for a more careful and closer follow-up to manage modifiable factors and reduce the potential risk of a major bleeding event. This approach has been prospectively tested in the mAFA-II trial, where the intervention arm using HAS-BLED had a lower risk of major bleeding at 1 year and an increase in OAC use compared with usual care, which showed higher bleeding and a decline in OAC use.23

One study that focused on the dynamic nature of bleeding risk found that dynamic assessments of HAS-BLED had a better risk predictive ability than baseline assessment alone for the prediction of major bleeding.24 These results are similar to that observed in our study, where the predictive ability of the HAS-BLED score calculated at 2 years was significantly better than the baseline assessment. Although there were no significant differences between the dynamic HAS-BLED at 4 years and the baseline score, the dynamic HAS-BLED showed a slight improvement in sensitivity, and the overall clinical usefulness and net benefit of the dynamic HAS-BLED scores were still higher.

Overall, our results reinforce those of previous studies since the use of the dynamic CHA2DS2-VASc and HAS-BLED scores was associated with a higher net benefit and therefore with increased clinical usefulness than the baseline scores. Even though dynamic evaluation and reassessment of stroke and bleeding risks are currently widely accepted, many international clinical practice guidelines for AF patients do not as yet include clear and concise recommendations on how to perform and address this dynamic risk monitoring. The 2021 Asia Pacific Heart Rhythm Society AF guidelines provide recommendations on the dynamic nature of risk in patients with AF, whereby frequent reassessment of patients with AF is suggested, using the CHA2DS2-VASc and HAS-BLED scores.2,25

In our study, we show that changes in the overall score of CHA2DS2-VASc also corresponded to variations in the risk category of AF patients, showing an evolutionary increase in the proportion of patients classified within the high-risk group, to the detriment of a decrease in the proportion of low- and moderate-risk patients. For this reason, periodic reassessment of stroke risk is of particular interest in patients classified as low risk by a baseline CHA2DS2-VASc score (ie, 0 in men; 1 in women) given that OAC is not required in these patients but the progression of aging and new risk factors/comorbidities would change the overall CHA2DS2-VASc score, and therefore OAC may become indicated.15,26,27 Although there is special clinical interest in reassessment of the CHA2DS2-VASc score to reconsider the decision to initiate OAC in these patients initially classified as low-risk, it is also important to reevaluate the risk of stroke in high-risk AF patients already receiving OAC. Indeed, stroke rates can vary significantly between patients with a CHA2DS2-VASc from 3 to 9, even though all of them are in the same high-risk category, and not all stroke risk factors have the same impact.28 Additionally, several risk factors for stroke are also risk factors for bleeding, hence modifiable risk factors could be identified to be managed and reduced appropriately in high-risk patients.

In AF patients, stroke and bleeding risks are dynamic and change over time with aging and incident comorbidities. The CHA2DS2-VASc and HAS-BLED scores (and clinical risk profiles) should be regularly reassessed, which is particularly necessary for appropriate stroke and bleeding risk prediction.

This work was supported by the Spanish Ministry of Economy, Industry, and Competitiveness, through the Instituto de Salud Carlos III after independent peer review (research grant: PI17/01375 cofinanced by the European Regional Development Fund) and group CB16/11/00385 from CIBERCV.

The study protocol was approved by the Ethics Committee of Hospital General Universitario Morales Meseguer and was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki. All patients gave informed consent for participation. The possible variables of sex and gender have been taken into account in this work in accordance with the SAGER guidelines. In fact, the study is balanced between male and female participants (48.7% men, 51.3% women).

No artificial intelligence was used in the preparation of this work.

M.J. Serna, J.M. Rivera-Caravaca, and V. Roldán contributed to data collection, performed statistical analyses, and drafted the manuscript. E. Soler-Espejo and R. López-Gálvez critically revised the manuscript. G.Y.H. Lip and F. Marín conceived and supervised the study and critically revised the manuscript. All authors read and approved the final version of the manuscript.

G.Y.H. Lip is a consultant and speaker for BMS/Pfizer, Boehringer Ingelheim and Daiichi-Sankyo. No fees are received personally.

J.M. Rivera-Caravaca is a consultant for Idorsia Pharmaceuticals LTD.

The remaining authors have nothing to disclose.