Validation of the CHA2DS2-VA Score (Excluding Female Sex) in Nonvalvular Atrial Fibrillation Patients: A Nationwide Population-Based Study.

Sex (i.e., female sex) confers one point for the CHA2DS2-VASc score. For this reason, females with atrial fibrillation (AF) always have a CHA2DS2-VASc score of at least 1. To compare the CHA2DS2-VA (excluding female sex) and CHA2DS2-VASc scores in Korean AF patients using the Korean National Health Insurance Service database, we analyzed the risk of ischemic stroke in nonvalvular AF patients between 2013 and 2017. The predictive values of the CHA2DS2-VA and CHA2DS2-VASc scores for ischemic stroke were evaluated using the C-statistic and net reclassification improvement (NRI). The primary outcome was the occurrence of ischemic stroke. A total of 185,637 patients with AF (49.7% women) were included in this study. The mean ages were 66.5 years for females and 64.9 years for males. The incidence of ischemic stroke in male patients was similar to females (3.63%/year vs. 3.72%/year, p = 0.273, respectively). In addition, no sex difference was apparent for stroke risk in AF patients stratified by risk factor component and age group. In the C-statistic analysis, the predictive ability of the CHA2DS2-VA score for ischemic stroke was similar to the CHA2DS2-VASc score. Additionally, CHA2DS2-VA performed better for predicting ischemic stroke in AF patients with risk scores of ≥2 (AUC 0.701 vs. 0.689, z = 4.596, p < 0.001) or those aged ≥75 years (AUC 0.715 vs. 0.701, z = 4.957, p < 0.001). In Korean AF patients, female sex is not a specific risk factor that contributes to the development of ischemic stroke. The CHA2DS2-VA score, which excludes female sex, may be a more suitable risk score for guiding anticoagulation decisions in Korean AF patients.

1. Introduction

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia disorder and confers a five-fold increased risk of stroke [1]. Stroke risk stratification is recommended for AF patients when assessing the risk of stroke and guiding oral anticoagulation (OAC) therapy decisions. Most major international guidelines recommend the CHA2DS2-VASc score for stroke risk stratification [2,3,4,5], which includes female sex as a risk component [6].

There are gender differences in the incidence and prevalence of AF. While the incidence of AF is lower in females, the prevalence of AF in males and females ≥75 years of age is greater for females due to their increased longevity, so the absolute numbers of males and females with AF are roughly equal based on population [7,8]. A higher risk of stroke in females than in males with AF has been evident in several studies [9,10,11,12,13]. However, the inclusion of female sex as an independent risk factor has been the subject of recent examination and remains controversial [9,10,11,12,13,14,15,16,17,18,19,20,21]. The latest Canadian, Australian, and Japanese guidelines do not use female sex as a risk factor [22,23,24]. There is an exception to the use of the CHA2DS2-VASc score in the European and American guidelines, whereby female patients aged <65 years without other risk factors (CHA2DS2-VASc score of 1) do not require anticoagulation therapy [2,5]. In addition, the European guidelines recommend (or at least consider) OAC therapy in males with a CHA2DS2-VASc score of ≥1 and females with a CHA2DS2-VASc score of ≥2. Although a CHA2DS2-VASc score of 0 indicates an extremely low risk of stroke, a female would be allocated 1 point in this scoring system from birth. It is therefore apparent that the CHA2DS2-VASc scoring system has limitations. The aim of this study was to compare the CHA2DS2-VASc and CHA2DS2-VA scores (sex-independent risk score) in Korean AF patients.

2. Materials and Methods

This study used the national health claims database established by the National Health Insurance Service (NHIS) of Korea. The NHIS is a mandatory universal health insurance service that provides comprehensive medical care coverage for up to 97% of the Korean population. The database contains each patient’s sociodemographic information, diagnoses, procedures, prescription records in inpatient and outpatient services, and mortality data. Diagnoses are recorded using the International Classification of Disease-10th Revision-Clinical Modification (ICD-10-CM) codes. The database is open to researchers whose study protocols are approved by the official review committee. Approval for this study was obtained from the Ethical Review Board Ethical Review Board of Dong-A University Hospital (15–130), and the requirement for informed consent was waived.

2.1. Study Population

From January 2013 to December 2017, 185,637 nonvalvular AF patients aged ≥20 years out of a total of 363,188 patients were selected from the National Health Insurance Service (NHIS) database as the study population. Prevalent nonvalvular AF was identified using International Classification of Disease, Tenth Revision, Clinical Modification (ICD-10-CM) codes (I48) and the baseline absence of mitral stenosis or mechanical heart valves (ICD-10 codes I05 or Z952–Z954). We excluded patients who had a history of thromboembolic events or intracranial hemorrhage (ICH) and only included those with newly diagnosed AF. Patients receiving oral anticoagulants (OAC) (warfarin or non-vitamin K antagonist oral anticoagulant (NOAC)), aspirin, or other antiplatelet agents at the baseline were also excluded.

2.2. Risk Stratification Schemes for Ischemic Stroke

The CHA2DS2-VASc score was calculated by assigning 1 point each for ages between 65 and 74 years; a history of hypertension, diabetes mellitus, congestive heart failure, or vascular disease (myocardial infarction or peripheral artery disease); and female sex and 2 points each for a history of stroke or transient ischemic stroke or ages ≥75 years [6]. The risk score excluding female sex from the CHA2DS2-VASc score was defined as the CHA2DS2-VA score (Table 1).

Table 1

Risk factors included in the CHA2DS2-VASc score and CHA2DS-VA score.

Risk Factor	CHA2DS2-VASc Score	CHA2DS2-VA Score
Congestive heart failure/LV dysfunction	1	1
Hypertension	1	1
Age ≥75	2	2
Diabetes mellitus	1	1
Stroke/TIA/thromboembolism	2	2
Vascular disease	1	1
Age 65–74	1	1
Sex category (i.e., female sex)	1	0

2.3. Clinical Endpoint

The primary study endpoint was the occurrence of ischemic stroke (ICD-10 code I63 or I64). Patients were followed until an endpoint occurred and censored at the outcome event or at the end of the study period.

2.4. Statistical Analysis

Continuous variables were expressed as mean values with standard deviations, and categorical variables were presented as frequencies (percentages). For the clinical endpoints, we calculated incidence rates per 100 person/years and estimated confidence intervals (CI) for the incidence rates assuming that the number of cases followed a Poisson distribution. The prognostic utility of the risk models for ischemic stroke was assessed by an analysis of their C-statistic, while calibration of the models was achieved using the Hosmer–Lemeshow goodness-of-fit statistical analysis. Net reclassification improvement (NRI) represents the average weighted improvement in discrimination. The impact of the reclassification procedure using the superior score was assessed using the NRI approach. Positive values for NRI indicated a predominance for correct reclassification, while negative values indicated a predominance for incorrect reclassification. All reported p-values were two sided, and p-values < 0.05 were considered statistically significant. Data manipulation and statistical analyses were conducted using SAS® Version 9.3 (SAS Institute Inc., Cary, NC, USA).

3. Results

3.1. Baseline Characteristics

Table 2 summarizes the characteristics of the 185,637 AF patients included in the analysis: 93,395 (50.3%) were male. The mean male age was 64.9 ± 11.5 years, and the mean female age was 66.5 ± 12.1 years. At the time of the AF diagnoses, hypertension was the most prevalent comorbidity in 65.5% of the total patients (male, 64.4% and female, 66.6%). With the exception of dyslipidemia, the remaining comorbidities (hypertension, diabetes mellitus, heart failure, and vascular disease) were higher in females than in males. Nineteen point two percent of the study patients were classified as having a CHA2DS2-VASc score of 0 or 1, and 26.5% of the patients were classified as having a CHA2DS2-VA score of 0 or 1. The median follow-up duration was 2.0 years (interquartile range (IQR): 1.1–2.7) for males and 2.2 years (1.2–2.6) for females.

Table 2

Baseline characteristics.

Variables	Total(Total = 185,637)	Male Patients(n = 93,395)	Female Patients(n = 92,242)
Age, years	65.7 ± 12.3	64.9 ± 11.5	66.5 ± 12.1
<55 years, n (%)	38,639 (20.8)	21,574 (23.1)	17,065 (18.5)
55–64 years, n (%)	41,327 (22.3)	22,602 (24.2)	18,725 (20.3)
65–74 years, n (%)	52,320 (28.2)	23,909 (25.6)	28,411 (30.8)
≥75 years, n (%)	53,352 (28.7)	25,310 (27.1)	28,042 (30.4)
Clinical history, n (%)
Hypertension	121,580 (65.5)	60,146 (64.4)	61,433 (66.6)
Diabetes mellitus	41,200 (22.2)	19,800 (21.2)	21,400 (23.2)
Heart failure	60,232 (32.4)	29,700 (31.8)	30,532 (33.1)
Dyslipidemia	67,306 (36.3)	34,836 (37.3)	32,469 (35.2)
Vascular disease	44,441 (23.9)	20,827 (22.3)	23,614 (25.6)
Prior myocardial infarction	7981 (4.3)	3923 (4.2)	4059 (4.4)
Prior peripheral artery disease	36,738 (19.8)	16,998 (18.2)	19,740 (21.4)
CHA2DS2-VASc score
Score = 0	8996 (4.8)	8996 (9.6)	0 (0.0)
Score = 1	26,757 (14.4)	21,312 (22.8)	5445 (5.9)
Score = 2	43,344 (23.3)	29,955 (32.1)	13,389 (14.5)
Score ≥ 3	106,540 (57.4)	33,132 (35.5)	73,408 (79.6)
CHA2DS2-VA score
Score = 0	14,441 (7.8)	8996 (9.6)	5445 (5.9)
Score = 1	34,701 (18.7)	21,312 (22.8)	13,389 (14.5)
Score = 2	57,093 (30.8)	29,955 (32.1)	27,138 (29.4)
Score ≥ 3	79,402 (42.8)	33,132 (35.5)	46,270 (50.2)
Medication
No therapy	17,207 (9.3)	8395 (9.0)	8812 (9.6)
Aspirin	42,163 (22.7)	20,386 (21.8)	21,777 (23.6)
Warfarin	61,128 (32.9)	29,981 (32.1)	31,147 (33.8)
NOAC	65,139 (35.1)	34,633 (37.1)	30,506 (33.1)
Rivaroxaban	27,970 (15.1)	14,599 (15.6)	13,371 (14.5)
Dabigatran	16,837 (9.1)	9285 (9.9)	7552 (8.2)
Apixaban	15,608 (8.4)	8574 (9.2)	7034 (7.6)
Edoxaban	4724 (2.5)	2175 (2.3)	2549 (2.8)
Follow up, years (IQR)	2.1 (1.4–2.6)	2.0 (1.1–2.7)	2.2 (1.2–2.6)

Values are n (%), mean ± SD (standard deviation), or median IQR (interquartile range). NOAC, non-vitamin K antagonist oral anticoagulant.

3.2. Risk of Ischemic Stroke in Male and Female Patients Stratified by Risk Score Components and Age

During follow-up, the male subpopulation accrued 3111 cases of ischemic stroke, while 3140 ischemic events were reported in females. The incidence of ischemic stroke was 3.63 per 100 person/year and 3.72 per 100 person/year for all males and females, respectively. However, no sex difference was found (HR = 1.11 for females to males, 95% CI 0.93–1.29, p = 0.273) (Figure 1). Even when the male and female patients were divided into two groups according to the number of risk score components (0 or 1 and ≥2 risk factors) and into three subgroups for age (<65, 65–74, and ≥75 years), there were no sex differences in the incidence of ischemic stroke in the subgroups.

Figure 1

The risk of ischemic stroke for male and female patients stratified by risk factor components (A) and age groups (B). Hazard ratio of female to male patients in the incidence of ischemic stroke. CI, confidence intervals; HR, hazard ratio; IR, incidence rate (events divided by 100 person/year, %/year).

3.3. Risk of Ischemic Stroke in AF Patients Stratified by Risk Scores

The incidence of ischemic stroke for AF patients stratified by the CHA2DS2-VASc and CHA2DS2-VA scores is shown in Table 3. There was a significant increase in the risk of ischemic stroke with the increasing CHA2DS2-VASc and CHA2DS2-VA scores. For risk scores of 0 or 1, ischemic stroke rates using CHA2DS2-VASc and CHA2DS2-VA were not significantly different (0.65%/year vs. 0.72%/year, p = 0.498 and 1.01%/year vs. 1.09%/year, p = 0.414, respectively). For risk scores of ≥2, the incidence rate of ischemic stroke was significantly higher with the CHA2DS2-VA score.

Table 3

Risk of ischemic stroke in AF patients stratified by the CHA2DS2-VASc and CHA2DS2-VA scores.

	No. of Events	IR (95% CI)	p-Value	HR (95% CI)	p-Value
CHA2DS2-VASc of 0	157/8996	0.65 (0.60–0.70)	0.498	Reference	0.498
CHA2DS2-VA of 0	260/14,441	0.72 (0.68–0.76)		1.12 (0.95–1.29)
CHA2DS2-VASc of 1	517/26,757	1.01 (0.97–1.05)	0.483	Reference	0.414
CHA2DS2-VA of 1	690/34,701	1.09 (1.06–1.12)		1.16 (0.99–1.33)
CHA2DS2-VASc of 2	1501/43,344	2.12 (2.07–2.17)	0.004	Reference	< 0.001
CHA2DS2-VA of 2	1619/57,093	2.44 (2.39–2.49)		1.33 (1.17–1.49)
CHA2DS2-VASc of ≥3	4076/106,540	5.18 (5.10–5.26)	<0.001	Reference	< 0.001
CHA2DS2-VA of ≥3	3682/79,402	5.46 (5.41–5.51)		1.49 (1.34–1.64)

CI, confidence intervals; HR, hazard ratio; IR, incidence rate (events divided by 100 person/year, %/year).

3.4. Comparison of Risk Scores

The predictive value of the CHA2DS2-VA and CHA2DS2-VASc scores for ischemic stroke was compared by the C-statistic and NRI in AF patients stratified by risk scores and age groups (Table 4). Across the total number of patients, the C-statistic for the CHA2DS2-VA score was similar to the CHA2DS2-VASc score (0.671 and 0.668, respectively). For low-risk AF patients (risk scores of 0 or 1), the discriminant ability of the CHA2DS2-VA (C-statistic 0.628) and CHA2DS2-VASc scores (C-statistic 0.626) was not significantly different (C-statistic difference 0.002, Z = 1.099, p = 0.272; NRI 0.019, 95% CI 0.001–0.032, p = 0.321). However, among the AF patients with risk scores of ≥2, the C-statistic values for the CHA2DS2-VA score were higher than for the CHA2DS2-VASc score (0.701 vs. 0.689, C-statistic difference 0.012, Z = 4.596, p < 0.001; NRI 0.126, 95% CI 0.081–0.172, p = 0.005). The predictive abilities of the CHA2DS2-VA (C-statistic 0.642 and 0.668, respectively) and CHA2DS2-VASc scores (C-statistic 0.639 and 0.663, respectively) were very similar for AF patients aged <65 years (C-statistic difference 0.003, Z = 0.879, p = 0.379; NRI 0.021, 95% CI 0.009–0.042, p = 0.309) and 65–74 years (C-statistic difference 0.005, Z = 1.193, p = 0.233; NRI 0.034, 95% CI 0.016–0.059, p = 0.227). However, in AF patients aged ≥75 years, the CHA2DS2-VA score was significantly superior compared to the CHA2DS2-VASc score in the C-statistic (0.715 vs. 0.701, C-statistic difference 0.014, Z = 4.957, p < 0.001) and NRI (0.159, 95% CI 0.119–0.196, p = 0.001). The cumulative incidences of ischemic stroke in males and females are shown in Figure 2. The incidence rates of ischemic stroke were not significantly different between male and female patients, regardless of the number of risk score components and age groups.

Table 4

Comparison of the scores for ischemic stroke predictions in AF patients stratified by risk scores and age groups.

All Patients	C-Statistic		SE	95% CI		p-Value
CHA2DS2-VA score	0.671		0.005	0.663–0.679		<0.001
CHA2DS2-VASc score	0.668		0.004	0.661–0.675		<0.001
Score difference	C-Statistic Analysis				NRI Analysis
	C-Statistic Difference	Z	p -Value	NRI	95% CI	p -Value
	0.003 (0.000–0.003)	0.998	0.318	0.031	0.002–0.037	0.118
Risk score of 0 or 1		C-Statistic	SE	95% CI		p-Value
CHA2DS2-VA score		0.628	0.005	0.619–0.637		<0.001
CHA2DS2-VASc score		0.626	0.005	0.618–0.634		<0.001
Score difference	C-Statistic Analysis				NRI Analysis
	C-Statistic Difference	Z	p -Value	NRI	95% CI	p -Value
	0.002 (0.000–0.004)	1.099	0.272	0.019	0.001–0.032	0.321
Risk score of ≥2		C-Statistic	SE	95% CI		p-Value
CHA2DS2-VA score		0.701	0.003	0.692–0.713		<0.001
CHA2DS2-VASc score		0.689	0.003	0.675–0.703		<0.001
Score difference	C-Statistic Analysis				NRI Analysis
	C-Statistic Difference	Z	p -Value	NRI	95% CI	p -Value
	0.012 (0.009–0.021)	4.596	< 0.001	0.126	0.081–0.172	0.005
Aged < 65 years		C-Statistic	SE	95% CI		p-Value
CHA2DS2-VA score		0.642	0.006	0.630–0.654		<0.001
CHA2DS2-VASc score		0.639	0.005	0.629–0.649		<0.001
Score difference	C-Statistic Analysis				NRI Analysis
	C-Statistic Difference	Z	p -Value	NRI	95% CI	p -Value
	0.003 (0.002–0.006)	0.879	0.379	0.021	0.009–0.042	0.309
Aged 65–74 years		C-Statistic	SE	95% CI		p-Value
CHA2DS2-VA score		0.668	0.005	0.658–0.678		<0.001
CHA2DS2-VASc score		0.663	0.004	0.655–0.671		<0.001
Score difference	C-Statistic Analysis				NRI Analysis
	C-Statistic Difference	Z	p -Value	NRI	95% CI	p -Value
	0.005 (0.003–0.008)	1.193	0.233	0.034	0.016–0.059	0.227
Aged ≥ 75 years		C-Statistic	SE	95% CI		p-Value
CHA2DS2-VA score		0.715	0.004	0.707–0.723		<0.001
CHA2DS2-VASc score		0.701	0.005	0.691–0.711		<0.001
Score difference	C-Statistis Analysis				NRI Analysis
	C-Statistic Difference	Z	p -Value	NRI	95% CI	p -Value
	0.014 (0.011–0.022)	4.957	<0.001	0.159	0.119–0.196	0.001

CI, confidence intervals; HR, hazard ratio; IR, incidence rate (events divided by 100 person/year, %/year); NRI, net reclassification; SE, standard error.

Figure 2

Cumulative incidence curves for ischemic stroke in male and female patients stratified by risk factor components and age groups.

4. Discussion

The present study aimed to identify whether sex is a prognostic factor for ischemic stroke in AF patients. We sought to validate the CHA2DS2-VA score, a risk score excluding female sex from the CHA2DS2-VASc score, in a large cohort of Korean AF patients. We found that there was no sex difference for the risk of ischemic stroke in AF patients between risk groups (low- and high-risk) and age groups (aged < 65 years, 65–74 years, and ≥75 years). The CHA2DS2-VA and CHA2DS2-VASc scores demonstrated similar predictive abilities with respect to ischemic stroke in all AF patients, but the C-statistic values for the CHA2DS2-VA scores were higher than for the CHA2DS2-VASc scores in high-risk (risk scores of ≥2) or elderly (aged ≥ 75 years) AF patients. These findings indicate that female sex did not have a consistent impact on the CHA2DS2-VA and CHA2DS2-VASc scores, and the CHA2DS2-VA score may be more suitable for risk stratification for ischemic stroke in Korean AF patients, especially in high-risk or elderly patients.

Stroke prevention is the principal management priority in AF patients, so appropriate risk stratification is needed to balance the benefits of intervention against the risk of bleeding [25]. The CHA2DS2-VASc score, which includes female sex as a risk component [6], is formally recommended for stroke risk stratification in most major international guidelines [2,3,4,5]. However, recent scientific focus has shifted toward the sex category criterion as an independent risk factor for thromboprophylaxis, and the inclusion of female sex as a risk factor remains controversial [9,10,11,12,13,14,15,16,17,18,19,20,21]. The CHA2DS2-VASc risk score overestimates the risk conferred by female sex at the low end of the scale. These results support the European and American AF guidelines giving equal recommendations to both males and females without a sex risk factor [2,5]. In addition, the latest Canadian, Australian, and Japanese guidelines do not use the female sex criterion as a risk factor [22,23,24]. A number of studies have demonstrated that sex is not a significant ischemic event risk factor [14,15,16,17,18]. Additionally, recent large Japanese and Danish cohort studies have led to the implementation of the CHA2DS2-VA score (excluding the sex category criterion) as the preferred stroke risk score for guiding anticoagulation decisions in AF patients [19,20].

While females have a lower incidence of AF, the prevalence of AF in males and females ≥75 years of age is greater in women due to their increased longevity, so the absolute numbers of males and females with AF are roughly equal based on a population [7,8]. In addition, the risk of ischemic stroke is higher in females compared to males, but this association appears to be the result of confounding by age [10,11,12].

In this study, we evaluated the risk of ischemic stroke in male and female Korean AF patients, and no sex difference was found in patients when stratified by risk factor components and age groups. We also validated the CHA2DS2-VA score (excluding the sex category criterion) for ischemic stroke in Korean AF patients, and the predictive ability of CHA2DS2-VA score was similar to the CHA2DS2-VASc score. In addition, the CHA2DS2-VA score showed greater predictive accuracy for ischemic stroke than the CHA2DS2-VASc score, especially in identifying high-risk (risk score ≥ 2) or elderly (aged ≥ 75 years) Korean AF patients.

Study Limitations

This study was based on a nationwide cohort study using Korean NHIS data and might be limited by errors of coding, missing data, and laboratory measurements. Furthermore, because the AF diagnosis and the estimation of the clinical outcomes were based on diagnostic codes registered by physicians, the diagnosis of AF and events could be inaccurate. Second, the registry data also failed to provide any details regarding drug changes over time, as well as some unmeasurable confounding factors such as physicians’ decisions. For this reason, in the main analysis, we sought to adjust for several lists of potential confounders by including confounders in the Cox model; this made the problem of confounding by indication less of an issue, although it could not be ruled out completely. Third, our study may be subject to selection bias (prevalence incidence bias) and information bias (follow-up bias). The relatively short period of occurrence for new risk factors among patients initially having a CHA2DS2-VASc score of 0 (males) or 1 (females) could have resulted from an incomplete diagnostic assessment at the baseline. Fourth, we were not able to clearly confirm the cause of ischemic stroke, which could, in some cases, have been due to AF-related thromboembolism or atherosclerosis and thrombosis of the cerebral artery. Fifth, we inadvertently excluded pathologies where the sex factor has the most interactions with the long-term outcomes.

5. Conclusions

In Korean AF patients, the female sex criterion is not a specific risk factor that contributes to the development of ischemic stroke. The CHA2DS2-VA score excluding female sex may be a more appropriate risk score for guiding anticoagulation decisions in Korean AF patients. Further discussion is warranted to apply these results to Korean patients with AF.