2015 年 79 巻 2 号 p. 432-438
Background: The incidence rate of ischemic stroke in Japanese non-valvular atrial fibrillation (NVAF) patients without anticoagulation therapy remains unclear.
Methods and Results: We performed a pooled analysis of 3,588 patients from the Shinken Database (n=1,099), J-RHYTHM Registry (n=1,002), and Fushimi AF Registry (n=1,487) to determine the incidence rate of ischemic stroke in Japanese NVAF patients without anticoagulation therapy. Average patient age was 68.1 years. During the follow-up period (total, 5,188 person-years; average, 1.4 years), 69 patients suffered from ischemic stroke (13.3 per 1,000 person-years; 95% confidence intervals [CI]: 10.5–16.8). The incidence rates of ischemic stroke were 5.4, 9.3, and 24.7 per 1,000 person-years and 5.3, 5.5, and 18.4 per 1,000 person-years in patients with low (0), intermediate (1), and high (≥2) CHADS2 and CHA2DS2-VASc scores, respectively. History of ischemic stroke or transient ischemic attack (hazard ratio [HR], 3.25; 95% CI: 1.86–5.67), age ≥75 years (HR, 2.31; 95% CI: 1.18–4.52), and hypertension (HR, 1.69; 95% CI: 1.01–2.86) were independent risk factors for ischemic stroke.
Conclusions: A low incidence rate of ischemic stroke was observed in Japanese NVAF patients except for those with CHADS2 score ≥2. In this pooled analysis, history of ischemic stroke or transient ischemic attack, advanced age, and hypertension were identified as independent risk factors for ischemic stroke. (Circ J 2015; 79: 432–438)
The prognosis of ischemic stroke associated with atrial fibrillation (AF) is extremely poor.1 Anticoagulants, however, can help prevent ischemic stroke. Warfarin can reportedly reduce the risk of ischemic stroke by 68%.2
Editorial p 307
One major issue in the warfarin era is underuse of the drug for AF.3–5 Only approximately 60% of high-risk patients (CHADS2 score ≥2) are reportedly given warfarin. With the emergence of newer oral anticoagulants in recent years, a substantial improvement in the rate of anticoagulant use is anticipated. This issue, however, should be discussed not only among cardiovascular specialists but also with other clinicians. The number of AF patients has increased with the current increase in the average life expectancy, and the number of AF patients in Japan is expected to exceed 1 million within the next 20 years.6,7 This means that general clinicians also need to take on an important role in the prevention of AF-related ischemic stroke. Compared with cardiovascular specialists,8 however, the use of anticoagulants by general physicians to prevent AF-related ischemic stroke has not gained sufficient momentum, as evidenced by the relatively lower prevalence of patients on anticoagulants in general practice.5 One reason for this may be that the risk of ischemic stroke in Japanese AF patients has not been sufficiently recognized. Based on data from the USA and Europe, Japan has accepted the CHADS2 scoring system9 and is now close to accepting the CHA2DS2-VASc scoring system.10 However, when administering anticoagulants for primary prevention of ischemic stroke in AF patients, we should know with certainty that the incidence rates of ischemic stroke based on these scores are applicable to AF patients in Japan. Therefore, the status of ischemic stroke incidence in AF patients in Japan must be determined.
To determine the incidence rate of ischemic stroke in Japanese AF patients requires the collection of a large amount of data on patients who are not receiving anticoagulants. Although anticoagulants are still considered to be underused, the current rates of anticoagulant use in Japanese AF patients is believed to range from 50% to 80%.5,8,11 Under such circumstances, there would be a certain limitation in individually analyzing the data of patients who are not on anticoagulants in the existing databases. In the present study, we used the J-RHYTHM registry,8,12–15 Fushimi AF registry,5,16 and Shinken Database,11,17 which are currently the largest AF patient registries in Japan, to collect the data of AF patients who are not on anticoagulant therapy. The acquired data were analyzed to assess the incidence rate of ischemic stroke by determining the CHADS2 and CHA2DS2-VASc scores and independent risk factors for ischemic stroke.
Currently, the J-RHYTHM Registry, Fushimi AF Registry, and Shinken Database are the largest AF patient databases in Japan. The registration criteria for each of the databases are as follows.
J-RHYTHM Registry The J-RHYTHM Registry is a prospective, observational registry with a 2-year follow-up period.15 The enrollment period was between January 2010 and July 2010. Patients ≥20 years of age who had at least 1 AF episode on a standard 12-lead electrocardiogram (ECG) were eligible for enrollment in the J-RHYTHM Registry. At each institution, a consecutive series of AF patients whose sinus rhythm was maintained for >1 year were enrolled. Patients were expected to undergo clinical observation in their outpatient clinic. Inpatients were excluded from the study. All AF patients were expected to return to their outpatient clinic for follow-up. For events that occurred during the follow-up period, final clinical data at the time closest to the event were mandatorily sent to the data center.
Fushimi AF Registry The Fushimi AF Registry is a prospective, observational registry intended to identify the current status of AF patients in a community-based clinical setting.16 Patient enrollment began in March 2011. The inclusion criterion for the registry was documentation of AF on 12-lead ECG or Holter monitor at any time. There were no exclusion criteria. A total of 76 institutions, all of which are members of Fushimi-Ishikai (Fushimi Medical Association), participated in the registry. All participating institutions enrolled consecutive AF patients under regular outpatient care or under admission at Fushimi-ku. Collection of follow-up information was mainly conducted through review of the inpatient and outpatient medical records, and additional follow-up information was obtained through contact with patients, relatives, and/or referring physicians by mail or telephone. The study protocol was approved by the ethics committees of National Hospital Organization Kyoto Medical Center and Ijinkai Takeda General Hospital.
Shinken Database The Shinken Database was established by registering all new patients visiting The Cardiovascular Institute Hospital in Tokyo, Japan.11 Patients with active cancer and any foreign travelers were excluded. The principal aim of this hospital-based database is surveillance of the prevalence and prognosis of cardiovascular diseases in the urban areas of Japan. AF was diagnosed on ECG recordings including 12-lead surface ECG and 24-h Holter recordings within 3 months after the initial visit, and medical history provided by the referring physicians. Patients diagnosed with new-onset AF (>3 months after the initial visit) were not included in the database. The registry started in June 2004 and patients continue to be registered into the database annually. Patient health status and incidence of cardiovascular events and mortality are maintained in the database through links to the hospital medical records and by study documents of prognosis sent annually to patients who stopped hospital visits or were referred to other hospitals. The ethics committee of The Cardiovascular Institute Hospital approved the study, and all patients gave written informed consent.
In the present study, we used the data of all patients registered in the J-RHYTHM Registry (non-valvular AF [NVAF], n=7,516), those with obtainable follow-up information in the Fushimi AF Registry (NVAF, n=3,189), and those registered in the Shinken Database between 2004 and 2012 (NVAF, n=2,244). We extracted and combined the data of NVAF patients who were not on anticoagulant therapy at the time of registration. In the process of combining databases, we were aware of the differences in the methods of data collecting among the 3 databases as described here. J-RHYTHM Registry was a completed study that dealt with a closed cohort. Because of its nature, the follow-up was perfect. Meanwhile, Fushimi AF Registry and Shinken Database were ongoing and open cohorts, and the enrollment of patients was continuing at the time of the present study. Therefore we should have considered that the follow-up rates were changing with time and therefore should be imperfect, as compared with J-RHYTHM Registry. To minimize the chance of missing the events and to balance the follow-up period between the 3 databases, patients were included in the analysis only when their follow-up periods were >1 year on 30 April 2014, and the follow-up data exceeding ≥2 years were excluded from the analysis. The number of patients analyzed was 3,588 (J-RHYTHM Registry, n=1,002; Fushimi AF Registry, n=1,487; Shinken Database, n=1,099). The decision to start anticoagulant therapy by the attending physicians was mainly determined according to the 2008 guidelines for pharmacotherapy of AF by Japanese Circulation Society.18
CHADS2 and CHA2DS2-VASc ScoresThe CHADS2 score was determined as previously described.9 CHA2DS2-VASc score was determined according to the 2010 European Society of Cardiology Guidelines for the management of AF,19 but “V” in the CHA2DS2-VASc score was replaced by coronary artery disease, because data regarding peripheral artery disease and aortic plaque were not available.14
Statistical AnalysisCategorical and consecutive patient data are presented as number (%) and mean±SD, respectively. The incidence rate of ischemic stroke was calculated as number per 1,000 person-years with 95% confidence intervals (95% CI) and was determined in the entire patient population and according to CHADS2 and CHA2DS2-VASc scores. Cox regression analysis was performed to estimate the unadjusted hazard ratios (HR) for each risk factor. The predictive accuracy of the CHADS2 and CHA2DS2-VASc scores was determined using a receiver operating characteristic (ROC) curve. A C-statistic of 0.5 indicates that the predictive accuracy of the score was significantly better than chance. Multivariate Cox regression analysis was performed to determine the independent risk factors for ischemic stroke. All statistical analysis was done using SPSS (SPSS, Chicago, IL, USA) for Windows (Microsoft, Redmond, WA, USA) version 19.0. Statistical significance was set at a 2-sided P<0.05.
A total of 3,588 Japanese NVAF patients who were not on anticoagulant therapy were included in the analysis (Shinken Database, n=1,099; J-RHYTHM Registry, n=1,002; Fushimi AF Registry, n=1,487). The patient characteristics of the total population are listed in Table 1. The average patient age was 68.1 years, and 1,216 patients (33.9%) were female. The number of patients with a history of hypertension; diabetes; previous ischemic stroke or transient ischemic attack (TIA); heart failure; and coronary artery disease were 1,809 (50.4%), 549 (15.3%), 305 (8.5%), 538 (15.0%), and 385 (10.7%), respectively.
| Variable | Total (n=3,588) |
Shinken Database (n=1,099) |
J-RHYTHM Registry (n=1,002) |
Fushimi Registry (n=1,487) |
|---|---|---|---|---|
| Age (years) | 68.1±13.5 | 61.4±13.3 | 67.8±11.8 | 73.2±12.5 |
| Age | ||||
| <65 | 1,340 (37.3) | 646 (58.8) | 366 (36.5) | 328 (22.1) |
| 65–74 | 1,016 (28.3) | 270 (24.6) | 328 (32.7) | 418 (28.1) |
| ≥75 | 1,232 (34.3) | 183 (16.7) | 308 (30.7) | 741 (49.8) |
| Female | 1,216 (33.9) | 274 (24.9) | 297 (29.6) | 645 (43.4) |
| Hypertension | 1,809 (50.4) | 408 (37.1) | 534 (53.3) | 867 (58.3) |
| Diabetes mellitus | 549 (15.3) | 101 (9.2) | 136 (13.6) | 312 (21.0) |
| Cerebral infarction or TIA | 305 (8.5) | 35 (3.2) | 67 (6.7) | 203 (13.7) |
| Heart failure | 538 (15.0) | 101 (9.2) | 155 (15.5) | 282 (19.0) |
| CAD | 385 (10.7) | 62 (5.6) | 94 (9.4) | 229 (15.4) |
| Antiplatelet use | 1,499 (41.8) | 397 (36.1) | 579 (57.8) | 523 (35.2) |
Data given as n (%) or mean±SD. AF, atrial fibrillation; CAD, coronary artery disease; TIA, transient ischemic attack.
During the observation period (total, 5,188 person-years; average, 1.4 years), 69 study patients suffered from ischemic stroke (13.3 per 1,000 person-years; 95% CI: 10.5–16.8). The incidence rates of ischemic stroke according to CHADS2 score were consistent between Shinken Database, J-RHYTHM Registry, and Fushimi AF Registry (Table 2).
| Variable | No. events | Person-years | Incidence rate (per 1,000 person-years) |
95% CI |
|---|---|---|---|---|
| Low risk (CHADS2 0 point) | ||||
| Shinken Database | 3 | 846 | 3.5 | 1.2–10.4 |
| J-RHYTHM Registry | 4 | 578 | 6.9 | 2.7–17.8 |
| Fushimi AF Registry | 2 | 253 | 7.9 | 2.2–28.8 |
| Intermediate risk (CHADS2 1 point) | ||||
| Shinken Database | 5 | 494 | 10.1 | 4.3–23.7 |
| J-RHYTHM Registry | 8 | 733 | 10.9 | 5.5–21.5 |
| Fushimi AF Registry | 3 | 502 | 6.0 | 2.0–17.6 |
| High risk (CHADS2 ≥2 points) | ||||
| Shinken Database | 8 | 331 | 24.1 | 12.2–47.7 |
| J-RHYTHM Registry | 15 | 628 | 23.9 | 14.5–39.4 |
| Fushimi AF Registry | 21 | 823 | 25.5 | 16.7–39.0 |
AF, atrial fibrillation; CHADS2, scoring system assigning 1 point each for congestive heart failure, hypertension, age ≥75 years and diabetes mellitus, and 2 points for prior stroke or transient ischemic attack; CI, confidence interval.
The incidence rate of ischemic stroke according to the presence or absence of each cofactor is shown in Table 3. The incidence of ischemic stroke increased with age from 6.9 per 1,000 person-years in patients aged <65 years to 9.8 per 1,000 person-years in those aged 65–74 years and 24.5 per 1,000 person-years in those aged ≥75 years. The HR in patients aged ≥75 years relative to those aged <65 years was 3.37 (95% CI: 1.83–6.22). The incidence rate was significantly higher in patients with hypertension compared with those without hypertension (18.3 vs. 8.4 per 1,000 person-years; HR, 2.12; 95% CI: 1.28–3.53). The incidence rate was also significantly higher in patients with a previous history of ischemic stroke or TIA compared with those without a history of ischemic stroke or TIA (51.2 vs. 10.2 per 1,000 person-years; HR, 4.79; 95% CI: 2.84–8.07). Although the incidence rate of ischemic stroke was higher in female patients, patients with diabetes, and patients with heart failure than in male patients, patients without diabetes, and patients without heart failure, respectively, the differences were not statistically significant. On multivariate Cox regression analysis, age ≥75 years, hypertension, and previous history of ischemic stroke or TIA were identified as independent risk factors for ischemic stroke (Table 4).
| Variable | No. events | Person-years | Incidence rate (per 1,000 person-years) |
95% CI | HR | 95% CI |
|---|---|---|---|---|---|---|
| Total | 69 | 5,188 | 13.3 | 10.5–16.8 | – | |
| Age | ||||||
| <65 years | 14 | 2,035 | 6.9 | 4.1–11.5 | Reference | |
| 65–74 years | 15 | 1,523 | 9.8 | 6.0–16.3 | 1.39 | 0.67–2.89 |
| ≥75 years | 40 | 1,630 | 24.5 | 18.0–33.4 | 3.37 | 1.83–6.22 |
| Men | 41 | 3,516 | 11.7 | 8.6–15.8 | Reference | |
| Women | 28 | 1,673 | 16.7 | 11.6–24.2 | 1.38 | 0.85–2.24 |
| Hypertension | ||||||
| No | 22 | 2,623 | 8.4 | 5.5–12.7 | Reference | |
| Yes | 47 | 2,565 | 18.3 | 13.8–24.4 | 2.12 | 1.28–3.53 |
| Diabetes mellitus | ||||||
| No | 55 | 4,448 | 12.4 | 9.5–16.1 | Reference | |
| Yes | 14 | 740 | 18.9 | 11.3–31.8 | 1.47 | 0.81–2.65 |
| Cerebral infarction or TIA | ||||||
| No | 49 | 4,797 | 10.2 | 7.7–13.5 | Reference | |
| Yes | 20 | 391 | 51.2 | 33.1–79.0 | 4.79 | 2.84–8.07 |
| Heart failure | ||||||
| No | 57 | 4,446 | 12.8 | 9.9–16.6 | Reference | |
| Yes | 12 | 742 | 16.2 | 9.3–28.3 | 1.23 | 0.66–2.29 |
| CAD | ||||||
| No | 63 | 4,680 | 13.5 | 10.5–17.2 | Reference | |
| Yes | 6 | 508 | 11.8 | 5.4–25.8 | 0.83 | 0.36–1.92 |
| Antiplatelet use | ||||||
| No | 30 | 2,890 | 10.4 | 7.3–14.8 | Reference | |
| Yes | 39 | 2,299 | 17.0 | 12.4–23.2 | 1.67 | 1.04–2.70 |
HR, hazard ratio. Other abbreviations as in Tables 1,2.
| Variable | HR | 95% CI |
|---|---|---|
| Age | ||
| <65 years | Reference | |
| 65–74 years | 1.12 | 0.53–2.37 |
| ≥75 years | 2.31 | 1.18–4.52 |
| Women | 1.07 | 0.65–1.76 |
| Hypertension | 1.69 | 1.01–2.86 |
| Diabetes mellitus | 1.18 | 0.64–2.15 |
| Cerebral infarction or TIA | 3.25 | 1.86–5.67 |
| Heart failure | 0.86 | 0.45–1.65 |
| CAD | 0.52 | 0.22–1.26 |
| Antiplatelet use | 1.42 | 0.86–2.32 |
Abbreviations as in Tables 1–3.
The incidence rate of ischemic stroke increased with increasing CHADS2 score The incidence rate of ischemic stroke was 5.4 per 1,000 person-years for a score of 0 (95% CI: 2.8–10.2), 9.3 per 1,000 person-years for a score of 1 (95% CI: 5.7–15.0), 15.4 for a score of 2 (95% CI: 9.5–25.0), 26.6 per 1,000 person-years for a score of 3 (95% CI: 15.2–46.2), and 55.3 per 1,000 person-years for a score ≥4 (95% CI: 34.0–89.8). HR was statistically significant in patients with a score ≥2 (2.75; 95% CI: 15.2–46.2; Table 5). On stratification of patients according to CHA2DS2-VASc score there was no major difference in the incidence rate of ischemic stroke between the 0 and 1 CHA2DS2-VASc score groups (HR, 5.3; 95% CI: 2.3–12.4; HR, 5.5; 95% CI: 2.5–11.9, respectively), whereas HR was significantly higher in the CHA2DS2-VASc score ≥2 group (3.28; 95% CI: 1.31–8.21; Table 5). The C-statistics calculated from the ROC curves were 0.680 for CHADS2 (95% CI: 0.614–0.746) and 0.671 (95% CI: 0.606–0.736) for CHA2DS2-VASc.
| Variable | No. patients |
No. events |
Person- years |
Incidence rate (per 1,000 person-years) |
95% CI | HR | 95% CI |
|---|---|---|---|---|---|---|---|
| CHADS2 score (consecutive) | 1.62 | 1.39–1.88 | |||||
| 0 | 1,090 | 9 | 1,677 | 5.4 | 2.8–10.2 | ||
| 1 | 1,169 | 16 | 1,729 | 9.3 | 5.7–15.0 | ||
| 2 | 754 | 16 | 1,041 | 15.4 | 9.5–25.0 | ||
| 3 | 341 | 12 | 452 | 26.6 | 15.2–46.4 | ||
| 4 | 159 | 12 | 198 | 60.5 | 34.6–105.7 | ||
| 5 | 64 | 3 | 77 | 38.9 | 13.2–114.3 | ||
| 6 | 11 | 1 | 14 | 72.4 | 12.8–410.1 | ||
| CHADS2 score (categorical) | |||||||
| Low (0) | 1,090 | 9 | 1,677 | 5.4 | 2.8–10.2 | Reference | |
| Middle (1) | 1,169 | 16 | 1,729 | 9.3 | 5.7–15.0 | 1.68 | 0.74–3.81 |
| High (≥2) | 1,329 | 44 | 1,782 | 24.7 | 18.4–33.1 | 4.36 | 2.12–8.95 |
| CHA2DS2-VASc score (consecutive) | 1.42 | 1.25–1.60 | |||||
| 0 | 617 | 5 | 944 | 5.3 | 2.3–12.4 | ||
| 1 | 702 | 6 | 1,096 | 5.5 | 2.5–11.9 | ||
| 2 | 667 | 11 | 987 | 11.1 | 6.2–20.0 | ||
| 3 | 669 | 13 | 939 | 13.8 | 8.1–23.7 | ||
| 4 | 484 | 10 | 658 | 15.2 | 8.3–28.0 | ||
| 5 | 271 | 15 | 339 | 44.3 | 26.8–73.0 | ||
| 6 | 110 | 6 | 147 | 40.7 | 18.7–88.9 | ||
| 7 | 56 | 1 | 64 | 15.6 | 2.8–88.6 | ||
| 8 | 11 | 1 | 14 | 69.5 | 12.3–393.5 | ||
| 9 | 1 | 1 | 0 | 2,109.8 | 372.2–119.5×104 | ||
| CHA2DS2-VASc score (categorical) | |||||||
| Low (0) | 617 | 5 | 944 | 5.3 | 2.3–12.4 | Reference | |
| Middle (1) | 702 | 6 | 1,096 | 5.5 | 2.5–11.9 | 1.02 | 0.31–3.35 |
| High (≥2) | 2,269 | 58 | 3,149 | 18.4 | 14.3–23.8 | 3.28 | 1.31–8.21 |
CHA2DS2-VASc, scoring system assigning 1 point each for congestive heart failure, hypertension, age 65–74 years, diabetes mellitus and female gender, and 2 points each for prior stroke or transient ischemic attack, and age ≥75 years. Other abbreviations as in Tables 2,3.
In the present study, we determined the incidence rate of ischemic stroke in Japanese NVAF patients not on anticoagulant therapy using a combination of 3 major Japanese databases. The annual incidence rate was approximately 1.3%, which is approximately half the annual incidence rate in a recent report in Western countries10 and in a previous report in Japan.20 The incidence rate of ischemic stroke increased with increasing CHADS2 score. The C-statistic was 0.680, indicating the same level of predictive accuracy of the CHADS2 score in Japan as in Western countries. The risk of ischemic stroke was significantly higher in patients with a high CHADS2 score (≥2) than in those with low and intermediate scores. The incidence rate of ischemic stroke also increased with increasing CHA2DS2-VASc score, and the C-statistic was 0.671, indicating the same level of predictive accuracy as that in Western countries. Age ≥75 years, hypertension, and previous history of ischemic stroke or TIA were identified as independent risk factors for ischemic stroke in the present cohort.
Incidence Rate of Ischemic StrokeIn the validation cohort in which the CHADS2 score was reported for the first time, 94 AF patients not on anticoagulants suffered from ischemic stroke during the 2,121-patient-years follow-up (incidence rate, 4.4%/year).9 In a report of the 1-year follow-up of the Euro Heart Survey, 25 of 1,084 patients who were not on oral anticoagulants at baseline developed ischemic stroke (2.3% by simple calculation).10 There was a decade-long difference in the study period and a 2-fold difference in the incidence of ischemic stroke between the 2 Western cohort studies. Similar differences in study period and incidence of ischemic stroke were also found between the present study and a previous study in Japan. In a multicenter study by Inoue et al, 11 NVAF patients not on warfarin therapy suffered from thromboembolic events during 483 patient-years (not provided in the paper; calculated by summing figures by groups), indicating an incidence rate of approximately 2.3% per year by simple calculation.20 In the present study, the incidence rate of ischemic stroke was determined to be approximately 1.3% per year based on 3 major databases recently established in Japan.
In general, differences in the incidence rates between the 3 different registries are assumed to be derived from differences in patient background. The distribution of low (0), intermediate (1), and high (≥2) CHADS2 scores in the first validation cohort of CHADS2 score9 was 7% (120/1,733), 27% (463/1,733), and 66% (1,150/1,733), respectively, which was obviously different from that in the Euro Heart Survey (0, 21.1%; 1, 62.1%; and ≥2, 16.8%).10 Differences in the distribution of patients with high and low CHADS2 (high and low prevalence of high CHADS2 score in the former and latter studies, respectively) may explain the differences in ischemic stroke incidence. The distribution of CHADS2 score also differed between the 2 Japanese studies. The distribution of low, intermediate, and high CHADS2 scores in the multicenter study by Inoue et al was 35% (87/246), 51% (126/246), and 13% (33/246), respectively,20 whereas the distribution in the present study was 30% (1,090/3,588), 33% (1,169/3,588), and 37% (1,329/3,588), respectively. Differences in the distribution of CHADS2 scores between the 2 Japanese cohorts (low and high prevalence of high CHADS2 score in the Inoue et al study20 and the present study, respectively), however, do not explain the differences in the incidence rate of ischemic stroke. Moreover, the prevalence of high CHADS2 score in the present study was even higher than that in the Euro Heart Survey.10
Differences in the incidence rate of ischemic stroke can be observed in different cohorts with similar CHADS2 scores. In the first validation cohort of CHADS2 score, the adjusted stroke rate was 1.9%/year in patients with a CHADS2 score of 0, 2.8%/year in those with a score of 1, and 4.0–18.2%/year in those with a score ≥2.9 In the Euro Heart Survey, the incidence rate was 1.4%/year in patients with a CHADS2 score of 0, 1.9%/year in patients with a score of 1, and 3.1%/year in patients with a score ≥2 (approximately two-thirds of that in the CHADS2 validation cohort9).10 Similar differences were observed between the 2 Japanese cohorts. In the Inoue et al study, the incidence rates were 0.6%/year in patients with a CHADS2 score of 0, 2.8%/year in patients with a score of 1, and 4.5%/year in patients with a score ≥2,20 whereas in the present study, the incidence rate was 5.4 per 1,000 person-years (approximately 0.5%/year) for a CHADS2 score of 0, 9.3 per 1,000 person-years (approximately 0.9%/year) for a CHADS2 score of 1, and 24.7 per 1,000 person-years (approximately 2.5%/year) for a CHADS2 score ≥2. The reason for the apparent difference in incidence rate remains unclear but might be due, at least in part, to the decade-long differences in medical environment, including the management of coexisting diseases.
Significance of Individual Risk FactorsIn the present study, we identified previous history of ischemic stroke or TIA, age ≥75 years, and hypertension as independent risk factors for ischemic stroke on multivariate analysis. These factors may have a different impact on ischemic stroke risk compared with the other 2 components of CHADS2 score, namely diabetes and heart failure. Neither the CHADS2 nor CHA2DS2-VASc scoring systems, however, are multivariate models composed of independent risk factors for ischemic stroke.9,10 Furthermore, in several previous medium-scale cohorts, similar differences in the role of CHADS2 components have been reported. Unlike diabetes and heart failure,21,22 previous history of ischemic stroke or TIA, age, and hypertension has been consistently identified as a significant risk factor for ischemic stroke.
Differences in the prevalence and/or intensity of the 3 significant risk factors might also explain, at least in part, the differences in the incidence rates of ischemic stroke between patients with similar CHADS2 scores. In the first validation cohort of CHADS2 score, patients with a previous history of ischemic stroke accounted for 25% of all cases,9 which was approximately 6-fold higher than that in the Euro Heart Survey (4.2%).10 The prevalence of prior stroke or TIA was 8–14% in the multicenter study by Inoue et al20 and 8.3% in the present study. The definition for hypertension has changed since 1999.23 In previous cohort studies, hypertension was defined as systolic blood pressure ≥160 mmHg.9,20 Changes in the hypertension guidelines, however, have amended this definition to ≥140 mmHg.23 Blood pressure control can have a large impact on the incidence of ischemic stroke.24,25 Therefore, it is possible that better blood pressure control may have contributed to the lower incidence rates in more recent cohorts10 than in previous cohorts.9,20 Furthermore, the number of patients diagnosed with hypertension may have differed between previous and more recent cohorts, given the change in hypertension definition.9,20 The incidence rate of ischemic stroke can vary widely depending on the number of patients aged ≥75 years included in the study population. Therefore, changes in patient distribution over time, and clinical situations associated with these important risk factors may affect the incidence rate of ischemic stroke26 even among patients with the same CHADS2 and CHA2DS2-VASc scores.
In the present study, the C-statistics of the CHADS2 and CHA2DS2-VASc scores were 0.680 and 0.671, respectively, and thus, showed good prediction ability for ischemic stroke. It is noteworthy that such simple scoring systems can have universally strong prediction abilities beyond the impact of their individual components.
Clinical ImplicationsThe present study reminds us that there are still 2 important issues in the prevention of ischemic stroke in NVAF patients in Japan. First, the incidence of ischemic stroke is higher in patients with CHADS2 score ≥2. Measures must be taken immediately to correct the current situation because many of these patients are being left untreated or placed on insufficient anticoagulant therapy.5,8 Second, previous history of ischemic stroke or TIA, advanced age, and hypertension were identified as factors that need extra attention in the CHADS2 and CHA2DS2-VASc scoring systems for the prevention of ischemic stroke. The fact that previous history of ischemic stroke or TIA was identified as a risk factor reconfirms the importance of secondary prevention. As for hypertension, it makes us recognize the importance of blood pressure control in conjunction with anticoagulant therapy. In addition, the fact that age was identified as an important risk factor indicates that the use of anticoagulant therapy in elderly patients remains a problem that has yet to be fully resolved.
Study LimitationsThe present study has several limitations. First, in the present study, using 3 Japanese large-scale databases, we intended to enhance the statistical power of the analysis by compensating for any numerical shortages in various low–high-risk patient groups through the integration of several databases consisting of patients with different backgrounds. The data, however, were not obtained from all areas of Japan in a balanced manner; therefore, the patient distribution in the present study does not represent the actual patient distribution across Japan. As yet, the resource-based bias might be minimum because, as shown in Table 2, the risk-stratified incidence rates of ischemic stroke were consistent between the 3 databases. We believe this consistency allows us to understand that the low incidence rate of ischemic stroke in the present study contains some kinds of truth. Second, selection bias was present because only subjects not on anticoagulant therapy were included. Patients not on anticoagulant therapy may have included hypertensive or diabetic patients with stable/controlled disease. Conversely, some patients might have been given anticoagulants by their attending physicians for reasons other than their risk scores alone, such as long duration of AF or left atrial dilatation. Therefore, the actual incidence rates of ischemic stroke in Japanese NVAF patients not on anticoagulant therapy may have been underestimated in the present study. Given the possible limitation regarding the underestimate of the risk of ischemic stroke, it should be emphasized that the results of the present study do not deny the significance of the risk of CHADS2 score 1 point, for which NOAC treatment is recommended in Japanese guidelines. Instead, the present results put emphasis only on the risk of CHADS2 ≥2 points and/or the risk of history of ischemic stroke or TIA, advanced age, and hypertension, which were significantly associated with incidence of ischemic stroke irrespective of the low incidence rate in the present study.
The incidence rate of ischemic stroke in Japanese NVAF patients not on anticoagulation therapy was relatively low. Ischemic stroke incidence, however, was increased in high-risk patients (CHADS2 score ≥2). In addition, we identified history of ischemic stroke or TIA, advanced age, and hypertension as independent risk factors for ischemic stroke, which is consistent with previous reports from Western countries. To our knowledge, this is the largest pooled analysis to provide the current status of ischemic stroke incidence in Japanese NVAF patients not on anticoagulation therapy.
Dr Suzuki has no conflict of interest; Dr Yamashita received research funding from Boehringer Ingelheim and Daiichi-Sankyo and remuneration from Boehringer Ingelheim, Daiichi-Sankyo, Bayer Healthcare, Pfizer, Bristol-Myers Squibb, Eisai, and Ono Pharmaceutical; Dr Okumura received research funding from Boehringer Ingelheim, Bayer Healthcare, and Daiichi-Sankyo and remuneration from Boehringer Ingelheim, Bayer Healthcare, Bristol-Myers Squibb, Pfizer, and Eisai; Dr Atarashi received research funding from Daiichi-Sankyo and Boehringer Ingelheim and lecture fees from Bayer Healthcare and Boehringer Ingelheim; Dr Akao received lecture fees from Pfizer, Bristol-Myers Squibb, Boehringer Ingelheim, Bayer Healthcare, and Daiichi-Sankyo; Dr Ogawa has no conflict of interest; Dr Inoue received research funding from Boehringer Ingelheim, Daiichi-Sankyo, Mitsubishi Tanabe Pharma, and Sumitomo Dainippon Pharma and remuneration from Otsuka Pharmaceutical, Daiichi-Sankyo, Sumitomo Dainippon Pharma, Boehringer Ingelheim, and Bayer Healthcare; the Fushimi AF Registry is supported by research funding from Boehringer Ingelheim, Bayer Healthcare, Pfizer, Bristol-Myers Squibb, Astellas Pharma, AstraZeneca, Daiichi-Sankyo, Novartis Pharma, MSD, Sanofi-Aventis, and Takeda Pharmaceutical.
