A Novel Risk Stratification System for Ischemic Stroke in Japanese Patients With Non-Valvular Atrial Fibrillation

Abstract

Background: Recently, identification of independent risk factors for ischemic stroke in Japanese non-valvular atrial fibrillation (NVAF) patients was made by analyzing the 5 major Japanese registries: J-RHYTHM Registry, Fushimi AF Registry, Shinken Database, Keio interhospital Cardiovascular Studies, and the Hokuriku-Plus AF Registry.

Methods and Results: The predictive value of the risk scheme in Japanese NVAF patients was assessed. Of 16,918 patients, 12,289 NVAF patients were analyzed (mean follow up, 649±181 days). Hazard ratios (HRs) of each significant, independent risk factor were determined by using adjusted Cox-hazard proportional analysis. Scoring system for ischemic stroke was created by transforming HR logarithmically and was estimated by c-statistic. During the 21,820 person-years follow up, 241 ischemic stroke events occurred. Significant risk factors were: being elderly (aged 75–84 years [E], HR=1.74), extreme elderly (≥85 years [EE], HR=2.41), having hypertension (H, HR=1.60), previous stroke (S, HR=2.75), type of AF (persistent/permanent) (T, HR=1.59), and low body mass index <18.5 kg/m² (L, HR=1.55) after adjusting for oral anticoagulant treatment. The score was assigned as follows: 1 point to H, E, L, and T, and 2 points to EE and S (HELT-E₂S₂ score). The C-statistic, using this score, was 0.681 (95% confidence interval [CI]=0.647–0.714), which was significantly higher than those using CHADS₂ (0.647; 95% CI=0.614–0.681, P=0.027 for comparison) and CHA₂DS₂-VASc scores (0.641; 95% CI=0.608–0.673, P=0.008).

Conclusions: The HELT-E₂S₂ score may be useful for identifying Japanese NVAF patients at risk of ischemic stroke.

The CHA₂DS₂-VASc score is widely used in Guidelines or Consensus Statements in Europe, in the United States and in the Asia-Pacific area for thromboembolic risk stratification in patients with non-valvular atrial fibrillation (NVAF),^1,2 whereas the CHADS₂ score is mainly used in Japan,³ although there are some arguments about its generalization. We previously showed that both scoring systems are useful for risk stratification for ischemic stroke in Japanese NVAF patients,⁴ but we also reported that all of the CHADS₂ score factors may not necessarily represent risks for thromboembolism.⁵ This may be related, in part, to relatively lower annual ischemic stroke rates of 0.5%, 0.9%, 1.5%, and 2.7% in Japanese NVAF patients with CHADS₂ scores of 0, 1, 2, and 3, respectively, compared with those shown in initial data.^5,6 Also, there may be some differences in risk factor schemes for thromboembolism according to race and region. Thus, appropriate risk stratification for Asian or Japanese NVAF patients still remains to be established.

Editorial p ????

By combining the individual data of patients registered in the 5 major Japanese registries (J-RHYTHM Registry,⁷ Fushimi AF Registry,⁸ Shinken Database,⁹ Keio interhospital Cardiovascular Studies,¹⁰ and Hokuriku-Plus AF Registry),¹¹ and analyzing them (n=12,289), we recently identified independent risk factors for ischemic stroke in Japanese NVAF patients; these being elderly (75–84 years), extreme elderly (≥85 years), hypertension, previous stroke, persistent/permanent type of AF, and low body mass index (BMI) (<18.5 kg/m²).¹² However, the predictive value of these risk factors and their relative weights are unknown. In the present study, we assessed their predictive value by assigning scores to each risk factor based on the logarithmically transformed hazard ratio (HR) and compared it with those of CHADS₂ and CHA₂DS₂-VASc scores in Japanese NVAF patients.

Methods

Study Patients

Detailed profiles of the study patients were reported, as described previously.¹² Briefly, the data from 5 major AF prospective registries in Japan, J-RHYTHM Registry (n=7,937),⁷ Fushimi AF Registry (n=3,749),⁸ Shinken Database (n=2,957),⁹ Keio interhospital Cardiovascular Studies (n=783),¹⁰ and Hokuriku-Plus AF Registry (n=1,492)¹¹ were collected and integrated in March 2016. The data from Keio interhospital Cardiovascular Studies were updated in April 2018. To balance the follow-up period among the registries, event data from individuals up to 730 days were collected. Of all patients, 819 with valvular AF and 3,810 with lack of the data were excluded, and finally, 12,289 NVAF patients (mean age, 70.2±11 years; female, 31%; mean CHADS₂ score, 1.7±1.3; mean CHA₂DS₂-VASc score, 2.9±1.7; hypertension, 73%; previous stroke, 14%; persistent/permanent AF, 55%; use of warfarin and direct oral anticoagulants [OAC], 64% and 10%, respectively) were analyzed. During the 21,820 person-years follow up (mean follow-up period 649±181 days), 241 ischemic stroke events occurred.

This study was approved by the ethics committees of the Hirosaki University Graduate School of Medicine (2015-117, 2017-1051), the National Cerebral and Cardiovascular Center (M27-092-4), the National Hospital Organization Kyoto Medical Center (15-101), the Cardiovascular Institute (279), the Kanazawa University Graduate School of Medical Science (2035-1, 2460-1), and the Keio University School of Medicine (20120029), and was performed within ethics committee-approved research protocols at other institutes.

Propensity Score Matching Analysis

Propensity score matching using the nearest-neighbor matching method was constructed by logistic regression modeling, adjusted for age, sex, congestive heart failure, hypertension, diabetes, previous stroke, type of AF, and BMI. Matching was performed in a 1 : 1 ratio without replacements, with 0.0001 as a caliper. Absolute standard mean differences were calculated.

Statistical Analysis

Data were expressed as mean±standard deviation or event number/patient number (%). The CHADS₂ and CHA₂DS₂-VASc scores were assessed as a categorical variable (0, 1, and ≥2). Kaplan-Meier analysis was conducted with numbers at risk in each score at 0, 200, 400, and 600 days. Cox proportional hazards models were used to calculate HR and 95% CI for incidence of ischemic stroke events. The score was assigned to each significant risk factor based on the coefficient calculated by logarithmically transformed HR, and a novel scoring system was generated. To assess the model performances of CHADS₂ and CHA₂DS₂-VASc scores and the novel scoring system, concordance-statistic (c-statistic) and 95% CI were calculated with 200 iterations in the bootstrap method. In order to check the risk of over-fitting in the model performance, a cross-validation procedure was performed, as outlined by Newson.¹³ Absolute standard mean differences were calculated. The modified Hosmer-Lemeshow χ² statistic test was used to determine the goodness-of-fit of the novel scoring system. Statistical analyses were performed using STATA16 (College Station, TX, USA). A P value <0.05 was considered statistically significant.

Results

Prediction of Ischemic Stroke Event by CHADS₂ and CHA₂DS₂-VASc Scores

When the study patients were stratified by CHADS₂ score, patients with the score ≥2 (high risk) showed significantly higher HR=3.25 (95% CI=1.98–5.35, P<0.001) compared with those with a score of 0 (low risk) as a reference (Figure 1A,B). C-statistic stratified by CHADS₂ score (full scoring: 0–6) was 0.647 (95% CI=0.614–0.681). When stratified by CHA₂DS₂-VASc score, patients with a score ≥2 showed a higher HR=1.81 (95% CI=0.99–3.33, P=0.05) compared with those with a score of 0 as a reference (Figure 1C,D). C-statistic stratified by CHA₂DS₂-VASc score (full scoring: 0–9) was 0.641 (95% CI=0.608–0.673).

Figure 1.

(A,C) Kaplan-Meier curves for the incidence of ischemic stroke stratified by CHADS₂ and CHA₂DS₂-VASc scores. (B,D) Hazard ratio (HR) and 95% confidence interval (CI) were calculated.

Significant Risk Factors Associated With Ischemic Stroke and Score Assignment

Our recent study showed that elderly age (75–84 years [E], HR=1.74), extreme elderly age (≥85 years [EE], HR=2.41), hypertension (H, HR=1.60), low BMI<18.5 kg/m² (L, HR=1.55), type of AF (persistent/permanent) (T, HR=1.59), and previous stroke (ischemic stroke + transient ischemic attack) (S, HR=2.75) were significant risk factors associated with ischemic stroke after adjusting for OAC administration at enrollment (Table 1).¹²

Table 1. Risk Factors Analyzed by Cox Proportional Hazards Model, and Score Assignment

	HR (95% CI)	Coefficient	Score
Age (<75 years)	Reference
75–84 years (Elderly)	1.74 (1.32–2.30)	0.6	1
≥85 years (Extreme elderly)	2.41 (1.63–3.56)	0.9	2
Hypertension	1.60 (1.15–2.23)	0.5	1
Low BMI <18.5 kg/m2	1.55 (1.05–2.29)	0.4	1
Type of AF (persistent/permanent)	1.59 (1.21–2.10)	0.5	1
Previous stroke	2.75 (2.09–3.62)	1.0	2
No oral anticoagulant	1.86 (1.40–2.47)	0.6

Scores were assigned based on the coefficient (0.4, 0.5, and 0.6=score 1, 0.9 and 1.0=score 2). Previous stroke indicates previous ischemic stroke+transient ischemic attack. AF, atrial fibrillation; BMI, body mass index; CI, confidence interval; HR, hazard ratio.

Based on this analysis, coefficients were calculated and a score was assigned simply by multiplying 2 to the coefficients and rounding off the numbers as follows: 1 point to each of age 75–84 years, hypertension, type of AF (persistent/permanent), and BMI <18.5 kg/m², and 2 points to each of age ≥85 years and previous stroke (Table 1). Thus, we propose a novel risk scoring system, the HELT-E₂S₂ score (H: hypertension; E: elderly age 75–84 years; L, low BMI (<18.5 kg/m²); T: type of AF (persistent/permanent); E₂, extreme elderly, age ≥85 years; S₂: previous stroke).

Prediction of Ischemic Stroke by Using the HELT-E₂S₂ Score

When the study patients were stratified by HELT-E₂S₂ score, the number of ischemic stroke events during the follow-up period was 6/1,077 (0.6%), 35/3,301 (1.1%), 55/3,983 (1.4%), 57/2,190 (2.6%), and 88/1,738 (5.1%) for scores of 0, 1, 2, 3, and ≥4, respectively (Table 2). Kaplan-Meier curves for each score are shown in Figure 2A, which indicated that a cumulative incidence of ischemic stroke increased with the increase in score. In agreement with this, patients with HELT-E₂S₂ scores of 2, 3, and ≥4 showed significantly higher HRs of 2.34 (95% CI=1.01–5.43; P=0.04), 4.51 (1.95–10.46; P<0.001), and 9.23 (4.04–21.11; P<0.001), respectively, compared to those with a score of 0 as a reference (Figure 2B). C-statistic, using the HELT-E₂S₂ score (full scoring: 0–7), was 0.681 (95% CI=0.647–0.714), which was significantly higher than that found when CHADS₂ and CHA₂DS₂-VASc scores were used (Table 3).

Table 2. Ischemic Stroke Events Stratified by the Risk Scoring Systems and Oral Anticoagulants

Registry	Total	J-RHYTHM	Fushimi	Shinken	Keio	Hokuriku
No. of patients	12,289	5,928	2,971	1,534	553	1,303
Ischemic stroke events	241 (2.0)	88 (1.5)	101 (3.4)	23 (1.5)	5 (0.9)	24 (1.8)
CHADS2 score
0	17/1,965 (0.9)	5/893 (0.6)	4/287 (1.4)	6/469 (1.3)	1/157 (0.6)	1/159 (0.6)
1	46/3,920 (1.2)	21/2,041 (1.0)	13/794 (1.6)	10/533 (1.9)	1/190 (0.5)	1/362 (0.3)
≥2	178/6,404 (2.8)	62/2,994 (2.1)	84/1,890 (4.4)	7/532 (1.3)	3/206 (1.5)	22/782 (2.8)
CHA2DS2-VASc score
0	11/890 (1.2)	2/372 (0.5)	3/89 (3.4)	5/297 (1.7)	1/85 (1.2)	0/47 (0)
1	11/1,830 (0.6)	5/901 (0.6)	1/290 (0.3)	4/370 (1.1)	0/121 (0)	1/148 (0.7)
≥2	219/9,569 (2.3)	81/4,655 (1.7)	97/2,592 (3.7)	14/867 (1.6)	4/337 (1.2)	23/1,108 (2.1)
HELT-E2S2 score
0	6/1,077 (0.6)	1/360 (0.3)	2/151 (1.3)	3/423 (0.7)	0/50 (0)	0/93 (0)
1	35/3,301 (1.1)	14/1,516 (0.9)	10/628 (1.6)	8/637 (1.3)	1/200 (0.5)	2/320 (0.6)
2	55/3,983 (1.4)	27/2,144 (1.3)	15/903 (1.7)	6/330 (1.8)	3/180 (1.7)	4/426 (0.9)
3	57/2,190 (2.6)	17/1,136 (1.5)	28/615 (4.6)	2/98 (2.0)	1/69 (1.5)	9/272 (3.3)
≥4	88/1,738 (5.1)	29/772 (3.8)	46/674 (6.8)	4/46 (8.7)	0/54 (0)	9/192 (4.7)
OAC
OAC (−)	78/3,197 (2.4)	22/745 (3.0)	39/1,373 (2.8)	12/783 (1.5)	2/107 (1.9)	3/189 (1.6)
OAC (+)	163/9,092 (1.8)	66/5,183 (1.3)	62/1,598 (3.9)	11/751 (1.5)	3/446 (0.7)	21/1,114 (1.9)

Data provided are from five atrial fibrillation (AF) registries in Japan: J-RHYTHM Registry (J-RHYTHM), Fushimi AF Registry (Fushimi), Shinken Database (Shinken), Keio interhospital Cardiovascular Studies (Keio), and Hokuriku-Plus AF Registry (Hokuriku) were collected. Values are expressed as n (%) or event number/patient number (%). BMI, body mass index; OAC, oral anticoagulant.

Figure 2.

(A) Kaplan-Meier curves for the incidence of ischemic stroke stratified by the HELT-E₂S₂ score. (B) Hazard ratio (HR) and 95% confidence interval (CI) were calculated.

Table 3. C-Statistics and Their Comparison Among Risk Stratification Schemes

	C-statistic (95% CI)	P value (Comparison with HELT-E2S2 score)
All patients (n=12,289)
HELT-E2S2 score	0.681 (0.647–0.714)	–
CHADS2 score	0.647 (0.614–0.681)	0.027
CHA2DS2-VASc score	0.641 (0.608–0.673)	0.008
Patients treated without OAC (n=3,197)
HELT-E2S2 score	0.703 (0.645–0.762)	–
CHADS2 score	0.657 (0.589–0.726)	0.108
CHA2DS2-VASc score	0.655 (0.594–0.716)	0.052
Patients treated with OAC (n=9,092)
HELT-E2S2 score	0.685 (0.643–0.726)	–
CHADS2 score	0.655 (0.615–0.695)	0.077
CHA2DS2-VASc score	0.646 (0.607–0.684)	0.027

C-statistic was evaluated by using the following classifications: score 0–7 in the HELT-E₂S₂ score, score 0–6 in the CHADS₂ score, and score 0–9 in the CHA₂DS₂-VASc score. CI, confidence interval; OAC, oral anticoagulants.

In order to check over-fitting of our model, we created a training dataset and a test dataset with a 5 : 5 ratio.¹³ After fitting a Cox model with scores from the present risk stratification scheme to the training set and creating the inverse hazard rate variables, we estimated the c-statistic in the test dataset for model fit to the training set using the bootstrap method. The C-statistic for its model was 0.689 (95% CI=0.640–0.739), which was similar to the c-statistic of the original dataset (0.681; 95% CI=0.647–0.714).

Analyses of Patients Without and With OAC

Patients without OAC (n=3,197) and with OAC (n=9,092) were separately analyzed. Clinical characteristics of each patient group are summarized in Supplementary Table. Kaplan-Meier curves in each score patients without OAC are shown in Figure 3A, which indicated that a cumulative incidence of ischemic stroke increased with the increase in score. In agreement, patients with HELT-E₂S₂ scores of 2, 3, and ≥4 showed a higher HR of 2.41 (95% CI=0.89–6.48; P=0.08), 4.50 (1.65–12.30; P=0.003), and 7.84 (3.00–20.49; P<0.001), respectively (Figure 3B). C-statistic using the HELT-E₂S₂ score was 0.703 (95% CI=0.645–0.762), which was higher than that found when using CHADS₂ (c-statistic=0.657; 95% CI=0.589–0.726, P=0.108 for comparison) and CHA₂DS₂-VASc scores (c-statistic=0.655; 95% CI=0.594–0.716, P=0.052 for comparison) (Table 3), although this was not found to be statistically significant.

Figure 3.

(A,C) Kaplan-Meier curves for the incidence of ischemic stroke stratified by the HELT-E₂S₂ score in patients without and with oral anticoagulants (OAC). (B,D) Hazard ratio (HR) and 95% confidence interval (CI) were calculated.

Kaplan-Meier curves for each score patient with OAC are shown in Figure 3C, which indicated that again, a cumulative incidence of ischemic stroke increased with the increase in score. In agreement, patients with HELT-E₂S₂ scores of 2, 3, and ≥4 showed a higher HR of 6.01 (95% CI=0.82–43.82; P=0.08), 12.09 (95% CI=1.66–87.87; P=0.01), and 25.54 (95% CI=3.54–184.11; P=0.001), respectively (Figure 3D).

C-statistic using the HELT-E₂S₂ score was 0.685 (95% CI=0.643–0.726), which was higher than that using CHADS₂ score (c-statistic=0.655; 95% CI=0.615–0.695, P=0.077 for comparison; close to being significant) and was significantly higher than that using the CHA₂DS₂-VASc score (c-statistic=0.646; 95% CI=0.607–0.684, P=0.027 for comparison) (Table 3).

Furthermore, cumulative incidence of ischemic stroke was compared between patients without and with OAC in each score separately. Although there were no significant differences in the patient group with HELT-E₂S₂ scores of 0 and 1, the incidence was significantly lower in patients with OAC in the other groups with HELT-E₂S₂ scores of 2 (HR=0.46; 95% CI=0.26–0.80; P=0.005), 3 (HR=0.49; 95% CI=0.27–0.87; P=0.01), and ≥4 (HR=0.58; 95% CI=0.37–0.93; P=0.02) (Figure 4). Incidence rate of ischemic stroke calculated by person-years ×100 also showed similar results (Figure 5).

Figure 4.

(A–E) Kaplan-Meier curves for the incidence of ischemic stroke in each score of the HELT-E₂S₂ scoring system in patients without oral anticoagulants (No OAC) and with OAC (OAC (+)). Hazard ratio (HR) was calculated and compared with No OAC as a reference. CI, confidence interval.

Figure 5.

Incidence rates of ischemic stroke stratified by each HELT-E₂S₂ score in patients without and with oral anticoagulants (OAC).

The propensity score-matched pairs (n=2,566) were created by matching clinical characteristics between patients without and with OAC (Supplementary Table). Kaplan-Meier analysis showed that the incidence of ischemic stroke was significantly lower in patients with OAC than in those without OAC (HR=0.67; 95% CI=0.47–0.96, P=0.029) (Supplementary Figure 1).

Evaluation of the Fitness of the HELT-E₂S₂ Score

Patients were divided into quartiles for the HELT-E₂S₂ score and then the modified Hosmer-Lemeshow test was performed. The test showed that the predicted number of ischemic stroke events using the HELT-E₂S₂ score suitably fitted the actual number of the events (χ²-statistic=3.146, P=0.37) (Supplementary Figure 2).

Prediction of Ischemic Stroke by Using a Non-Integer-Based Scoring System

A non-integer-based score was assigned by multiplying 10 to the coefficients shown in Table 1 as follows: 4 points to low BMI<18.5 kg/m², 5 points to hypertension and type of AF (persistent/permanent), 6 points to age 75–84 years, 9 points to age ≥85 years, and 10 points to previous stroke. C-statistic using this scoring system was 0.691 (95% CI=0.658–0.725), which was similar to that using the HELT-E₂S₂ score (c-statistic=0.681; 95% CI=0.647–0.714, P=0.25 for comparison).

Discussion

In the present study, we assigned a score to significant risk factors associated with ischemic stroke in Japanese NVAF patients as follows: 1 point to each of hypertension (H), age 75–84 years (E), BMI <18.5 kg/m² (L), and type of AF (persistent/permanent) (T), and 2 points to each of age ≥85 years (EE) and previous stroke (S). Thus, we generated a novel risk stratification scheme, the HELT-E₂S₂ score. The C-statistic using this novel score was significantly higher than that obtained when using CHADS₂ and CHA₂DS₂-VASc scores. The HELT-E₂S₂ score may be useful for identifying Japanese NVAF patients at risk for ischemic stroke.

A Novel Scoring System for Japanese NVAF Patients

Incidence rates of ischemic stroke associated with NVAF are different according to race and region.^5,6 Although CHADS₂ and CHA₂DS₂-VASc scores are widely used for risk assessment of ischemic stroke, a scoring system that includes region-specific risk factors may predict its risk more effectively in the region than globally recognized CHADS₂ and CHA₂DS₂-VASc scores. In this context, we proposed a novel risk scoring system, the HELT-E₂S₂ score, and assessed its predictability of the development of ischemic stroke in the Japanese NVAF patients. Ischemic stroke events during the follow-up period increased progressively from 0.6% in the score 0 group to 1.1%, 1.4%, 2.6%, and 5.1% in score groups 1, 2, 3, and ≥4, respectively. As HRs of the scores of ≥2 were significantly higher than those of score 0 (reference), the Japanese NVAF patients with a HELT-E₂S₂ score ≥2 are considered to be at increased risk for ischemic stroke. In this study, we further assessed the predictability of the novel score separately in patients without and with OAC at enrollment. As shown in Table 1, the absence of OACs was an independent risk factor for ischemic stroke, with a HR of 1.86. Propensity score matching analysis also supported this result. Accordingly, the number of ischemic stroke events increased to a greater degree in patients without OAC group than in those with OAC as the score increased (Figure 5). This was supported by the results showing that cumulative incidence of ischemic stroke in the groups with a HELT-E₂S₂ score ≥2 was significantly lower in patients with OAC than in those without OAC (Figure 4). However, as the number of patients without and with OAC (3,197 and 9,092, respectively) might not be sufficient to confirm the results conclusively, a higher number of patients need to be studied.

In the present study, c-statistic stratified by the HELT-E₂S₂ score in patients without OAC was 0.703, which was numerically higher than that found by using the CHADS₂ score in Taiwanese patients (0.659 or 0.683),^14,15 but lower than when the CHADS₂ score was used in Korean populations (0.79).¹⁶ Importantly, all risk factors composing the CHADS₂ score in Taiwanese and Korean NVAF patients were significantly associated with ischemic stroke events, whereas this was not the case for all risk factors in Japanese NVAF patients. Data resource also differed between our study from registries and other studies that used nationwide claim data. All these differences may affect the difference in c-statistic among Japanese, Taiwanese and Korean NVAF patients. Notably, recent analysis of the Korean National Health Insurance Service database showed that low body weight (<50 kg) of NVAF patients with OAC had a higher incidence of ischemic stroke than those weighting 50–60 kg.¹⁷ Thus, analysis by using the HELT-E₂S₂ scoring system in these countries is of great interest.

Characteristics of the HELT-E₂S₂ Score

The HELT-E₂S₂ score includes novel region-specific risk factors, which are extreme elderly (≥85 years), persistent/permanent AF, and low BMI, in addition to the 3 risk factors that are widely used for the CHADS₂ score.¹² Among them, the extreme elderly (≥85 years) is characterized as a most region-specific factor because Japan is one of the super aging countries where the prevalence of AF is high. Advanced age is known as a definite global risk factor for ischemic stroke in NVAF patients; the CHADS₂ score described age ≥75 years as a risk of stroke (one point), whereas the CHA₂DS₂-VASc score described age ≥65 years as a risk (one point), and that age ≥75 years has a higher risk (2 points) than age ≥65 years. In the present Japanese cohort, we also analyzed the effect of age on the risk of ischemic stroke, and found that age 65–74 years was not an independent risk factor (data not shown). We further found that the extreme elderly was associated with a higher risk compared with age ≥75 years, and its risk for ischemic stroke was as high as that of a previous stroke (Table 1). Thus, the inclusion of such a high age (E₂) as an independent factor is one of the most important regional characteristics of the present scoring system.

Although persistent/permanent AF was shown to be associated with an increased risk of ischemic stroke in the ROCKET-AF and ENGAGE AF-TIMI 48 trials,^18,19 this globally recognized risk factor has not been included in the widely used CHADS₂ and CHA₂DS₂-VASc scores. Furthermore, low body weight (≤50 kg) or low BMI (<18.5 kg/m²) is a risk factor for ischemic stroke in Japanese and Asian NVAF patients; these are region-specific factors.^17,20,21 All these considerations support that the fact that the HELT-E₂S₂ score consists of a combination of globally recognized risk factors and region-specific ones, which is the most important characteristic of this scoring system.

Comparison With CHADS₂ and CHA₂DS₂-VASc Scores

A Cox proportional hazards model revealed that the components of the risk factors for ischemic stroke in Japanese NVAF patients were not the same as those for the CHADS₂ and CHA₂DS₂-VASc risk factors.¹² Therefore, it was not surprising that the potentials for prediction of ischemic stroke events were different between the present and the preceding scoring systems. To characterize the present scoring system, the HELT-E₂S₂ score, we estimated comparatively the c-statistic in predicting the development of ischemic stroke among the 3 scoring systems. The results demonstrated a significantly greater c-statistic for the HELT-E₂S₂ score than for the CHADS₂ and CHA₂DS₂-VASc scores. In support of this result, analysis performed only in patients without OAC further showed a higher c-statistic in the HELT-E₂S₂ score than in CHADS₂ and CHA₂DS₂-VASc scores, which was close to significance. Furthermore, the ratio of ischemic stroke events during the follow-up period was lower in patients with a HELT-E₂S₂ score=0 (0.6%) than in those with a CHADS₂ score=0 (0.9%) and a CHA₂DS₂-VASc score=0 (1.2%) (Table 2). All these findings indicate that the present HELT-E₂S₂ score is likely to risk-stratify the Japanese NVAF patients more effectively, maybe because the present score includes region-specific risk factors found in Japanese NVAF patients.

Study Limitations

There are several limitations in the present study. First, the present results were not prospectively validated by external data. Although the modified Hosmer-Lemeshow test revealed that the HELT-E₂S₂ score suitably predicted the actual event data, and internal validation analysis confirmed the validity of this novel scoring system, an external validation study is warranted. It may be pointed out that such a validation study targeting the NVAF patients without OAC treatment would be preferable in order to risk-stratify the patients accurately, but this may not be possible in the real world where the necessity of prophylaxis with OAC in high-risk patients has been established. Second, the risk scoring system should be ideally developed among patients who were not anticoagulated. However, a relatively small sample size and a low number of ischemic stroke events in patients without anticoagulation limited the development of the risk scoring system. Third, c-statistic using a non-integer-based score was slightly higher than but statistically similar to that using the HELT-E₂S₂ score. Although the HELT-E₂S₂ score, which is based on integers, is likely to be more readily applied at the bedside, a non-integer-based score may also be useful for web-based scoring or algorithms embedded in the electronic health record. Fourth, the definition of hypertension is slightly different among the registries, as described in our previous paper.¹² The J-RHYTHM Registry and Keio interhospital Cardiovascular Studies included a history of hypertension, but the other registries did not, which may have affected our results. Fifth, although a significantly lower cumulative incidence of ischemic stroke in the group with a HELT-E₂S₂ score ≥2 was found in patients with OAC compared with those without OAC, indication of OAC remained to be elucidated. Net clinical benefit analysis, including major bleeding events, will be required. Sixth, the addition of other reported risk factors or markers such as the plasma brain natriuretic peptide level and echocardiographic values may have a potential to further increase the predictive value.^11,22 Finally, although the HELT-E₂S₂ score was developed for Japanese NVAF patients, partly based on their region-specific risk factors, it is still uncertain whether this scoring system is applicable to other races, countries, or regions.

Conclusions

We developed a novel risk stratification system, the HELT-E₂S₂ score, to determine the incidence of ischemic stroke in Japanese NVAF patients. This score may be more useful than the CHADS₂ and CHA₂DS₂-VASc scores for Japanese NVAF patients.

Acknowledgment

We thank Sachiko Tandai for her excellent administrative support. This study was supported by the Japan Agency for Medical Research and Development under Grant Number 19ek0210082 h0003.

Disclosures

Dr. Okumura received Speakers’ Bureau/Honorarium from Boehringer Ingelheim, Bayer, Daiichi-Sankyo, Bristol-Myers Squibb, and Pfizer; Dr. Tomita received research funding from Boehringer Ingelheim, Bayer, Daiichi-Sankyo, and Pfizer, and Speakers’ Bureau/Honorarium from Boehringer Ingelheim, Bayer, Daiichi-Sankyo, and Bristol-Myers Squibb; Dr. Kodani received remuneration from Daiichi-Sankyo, Bristol-Myers Squibb, and Ono Pharmaceutical; Dr. Akao received lecture fees from Pfizer, Bristol-Myers Squibb, Boehringer Ingelheim, Bayer Healthcare, and Daiichi-Sankyo; Dr. Suzuki received research funding from Daiichi-Sankyo, and Mitsubishi-Tanabe, and Speakers’ Bureau/Honorarium from Daiichi-Sankyo; Dr. Hayashi received Speakers’ Bureau/Honorarium from Bayer, Daiichi-Sankyo, and Bristol-Myers Squibb; Dr. Sawano received lecture fees from Boehringer Ingelheim, Bristol-Myers Squibb, Astellas Pharma, Sanofi, and research funding from Takeda Pharmaceutical; Dr. Goya received Speakers’ Bureau/Honorarium from Daiichi-Sankyo, Abbott, and Japan Life Line; Dr. Yamashita received research funding from Daiichi-Sankyo and Bristol-Myers Squibb, and Speakers’ Bureau/Honorarium from Daiichi-Sankyo, Bristol-Myers Squibb, Bayer, Ono Pharmaceutical, Boehringer Ingelheim, and Toa Eiyo; Dr. Isobe received Speakers’ Bureau/Honorarium from Daiichi-Sankyo, Otsuka Pharmaceutical, Chugai Pharmaceutical, and Pfizer; Dr. Toyoda received Speakers’ Bureau/Honorarium from Boehringer Ingelheim, Bayer, Daiichi-Sankyo, and Bristol-Myers Squibb; and Dr. Okamura received Speakers’ Bureau/Honorarium from Bayer and Daiichi-Sankyo.

Dr. Yamashita and Dr. Fukuda are Associate Editors for Circulation Journal.

IRB Information

This study was approved by the ethics committees of the Hirosaki University Graduate School of Medicine (2015-117, 2017-1051), the National Cerebral and Cardiovascular Center (M27-092-4), the National Hospital Organization Kyoto Medical Center (15-101), the Cardiovascular Institute (279), the Kanazawa University Graduate School of Medical Science (2035-1, 2460-1), and the Keio University School of Medicine (20120029), and was performed within ethics committee-approved research protocols at other institutes.

Data Availability

The deidentified participant data will not be shared.

Supplementary Files

Please find supplementary file(s);

http://dx.doi.org/10.1253/circj.CJ-20-1075

References

• 1.   Kirchhof P, Benussi S, Kotecha D, Ahlsson A, Atar D, Casadei B, et al. 2016 ESC Guidelines for the management of atrial fibrillation developed in collaboration with EACTS. Europace 2016; 18: 1609–1678.
• 2.   January CT, Wann LS, Calkins H, Chen LY, Cigarroa JE, Cleveland JC Jr, et al. 2019 AHA/ACC/HRS focused update of the 2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society. Heart Rhythm 2019; 16: e66–e93.
• 3.   JCS Joint Working Group. Guidelines for pharmacotherapy of atrial fibrillation (JCS 2013): Digest version. Circ J 2014; 78: 1997–2021.
• 4.   Okumura K, Inoue H, Atarashi H, Yamashita T, Tomita H, Origasa H. Validation of CHA₂DS₂-VASc and HAS-BLED scores in Japanese patients with nonvalvular atrial fibrillation: An analysis of the J-RHYTHM Registry. Circ J 2014; 78: 1593–1599.
• 5.   Suzuki S, Yamashita T, Okumura K, Atarashi H, Akao M, Ogawa H, et al. Incidence of ischemic stroke in Japanese patients with atrial fibrillation not receiving anticoagulation therapy: Pooled analysis of the Shinken Database, J-RHYTHM Registry, and Fushimi AF Registry. Circ J 2015; 79: 432–438.
• 6.   Gage BF, Waterman AD, Shannon W, Boechler M, Rich MW, Radford MJ. Validation of clinical classification schemes for predicting stroke: Results from the National Registry of Atrial Fibrillation. JAMA 2001; 285: 2864–2870.
• 7.   Atarashi H, Inoue H, Okumura K, Yamashita T, Kumagai N, Origasa H. Present status of anticoagulation treatment in Japanese patients with atrial fibrillation: A report from the J-RHYTHM Registry. Circ J 2011; 75: 1328–1333.
• 8.   Akao M, Chun YH, Wada H, Esato M, Hashimoto T, Abe M, et al. Current status of clinical background of patients with atrial fibrillation in a community-based survey: The Fushimi AF Registry. J Cardiol 2013; 61: 260–266.
• 9.   Suzuki S, Yamashita T, Otsuka T, Sagara K, Uejima T, Oikawa Y, et al. Recent mortality of Japanese patients with atrial fibrillation in an urban city of Tokyo. J Cardiol 2011; 58: 116–123.
• 10.   Ikemura N, Kohsaka S, Kimura T, Ueda I, Katsumata Y, Nishiyama T, et al. Assessment of sex differences in the initial symptom burden, applied treatment strategy, and quality of life in Japanese patients with atrial fibrillation. JAMA Netw Open 2019; 2: e191145.
• 11.   Hayashi K, Tsuda T, Nomura A, Fujino N, Nohara A, Sakata K, et al. Impact of B-type natriuretic peptide level on risk stratification of thromboembolism and death in patients with nonvalvular atrial fibrillation: The Hokuriku-Plus AF Registry. Circ J 2018; 82: 1271–1278.
• 12.   Okumura K, Tomita H, Nakai M, Kodani E, Akao M, Suzuki S, et al. Risk factors associated with ischemic stroke in Japanese patients with nonvalvular atrial fibrillation. JAMA Netw Open 2020; 3: e202881.
• 13.   Newson RB. Comparing the predictive powers of survival models using Harrell’s C or Somers’ D. Stata J 2010; 10: 339–358.
• 14.   Chao TF, Liu CJ, Tuan TC, Chen SJ, Wang KL, Lin YJ, et al. Comparisons of CHADS₂ and CHA₂DS₂-VASc scores for stroke risk stratification in atrial fibrillation: Which scoring system should be used for Asians? Heart Rhythm 2016; 13: 46–53.
• 15.   Lin LY, Lee CH, Yu CC, Tsai CT, Lai LP, Hwang JJ, et al. Risk factors and incidence of ischemic stroke in Taiwanese with nonvalvular atrial fibrillation: A nation wide database analysis. Atherosclerosis 2011; 217: 292–295.
• 16.   Kang SH, Choi EK, Han KD, Lee SR, Lim WH, Cha MJ, et al. Risk of ischemic stroke in patients with non-valvular atrial fibrillation not receiving oral anticoagulants: Korean nationwide population-based study. Circ J 2017; 81: 1158–1164.
• 17.   Lee SR, Choi EK, Park CS, Han KD, Jung JH, Oh S, et al. Direct oral anticoagulants in patients with nonvalvular atrial fibrillation and low body weight. J Am Coll Cardiol 2019; 73: 919–931.
• 18.   Steinberg BA, Hellkamp AS, Lokhnygina Y, Patel MR, Breithardt G, Hankey GJ, et al. Higher risk of death and stroke in patients with persistent vs. paroxysmal atrial fibrillation: Results from the ROCKET-AF Trial. Eur Heart J 2015; 36: 288–296.
• 19.   Link MS, Giugliano RP, Ruff CT, Scirica BM, Huikuri H, Oto A, et al. Stroke and mortality risk in patients with various patterns of atrial fibrillation: Results from the ENGAGE AF-TIMI 48 trial (Effective Anticoagulation with Factor Xa Next Generation in Atrial Fibrillation-Thrombolysis in Myocardial Infarction 48). Circ Arrhythm Electrophysiol 2017; 10: e004267.
• 20.   Hamatani Y, Yamashita Y, Esato M, Chun YH, Tsuji H, Wada H, et al. Predictors for stroke and death in non-anticoagulated Asian patients with atrial fibrillation: The Fushimi AF Registry. PLoS One 2015; 10: e0142394.
• 21.   Hamatani Y, Ogawa H, Uozumi R, Iguchi M, Yamashita Y, Esato M, et al. Low body weight is associated with the incidence of stroke in atrial fibrillation patients: Insight from the Fushimi AF Registry. Circ J 2015; 79: 1009–1017.
• 22.   Hamatani Y, Ogawa H, Takabayashi K, Yamashita Y, Takagi D, Esato M, et al. Left atrial enlargement is an independent predictor of stroke and systemic embolism in patients with non-valvular atrial fibrillation. Sci Rep 2016; 6: 31042.