論文ID: CJ-24-0939
Background: The aim of this was to develop an atrial fibrillation (AF) risk score using items usually included in Japanese governmental health check-ups.
Methods and Results: We analyzed data from 6,476 Japanese participants registered in the Suita Study. At baseline, the participants were aged ≥30 years and were free from AF. Cox regression analysis was used to identify AF risk factors, and a 0–100 score was developed to predict AF events within 10 years. Within a median follow-up of 14.6 years, 278 participants developed AF. The risk score incorporated age [<50 years (16 points for men, 0 for women), 50–59 years (26 points for men, 14 for women), 60–69 years (41 points for men, 37 for women), ≥70 years (54 points for men, 51 for women)], current smoking with a smoking index >500 (7 points), heavy alcohol consumption (8 points), body mass index ≥25 kg/m2 (6 points), hypertension (7 points), urinary proteins (4 points), glutamic-pyruvic transaminase >50 IU/dL (4 points), and cardiovascular disease history (10 points). The 10-year AF event probabilities were 7.1%, 8.4%, 10.8%, and 15.9% for scores of 47–54, 55–58, 59–69, and ≥70, respectively.
Conclusions: The new risk score to predict AF uses items similar to those used in Japanese governmental health check-ups.
Atrial fibrillation (AF) is a significant risk factor for cardiovascular disease (CVD),1 and identifying individuals at heightened risk of developing AF is essential for effective CVD prevention.2 This need for accurate prediction of AF has led to the development of several scoring systems from population-based cohort studies such as the Framingham Heart Study (FHS), the Atherosclerosis Risk In Communities (ARIC) study, the Women’s Health Study (WHS), and the Cohorts for Aging and Research in Genomic Epidemiology (CHARGE).3–6 However, these scores were developed in Western populations with distinct lifestyles and clinical characteristics compared with those of East Asian populations.7,8
In a previous study,9 we created a risk score to predict future AF events using data from the Suita Study, a prospective cohort study reflecting the general population in Japan. Although the Suita Risk Score has been used in health counseling in some Japanese healthcare settings, its nationwide application is challenging, primarily becayse certain risk factors included in the model, such as arrhythmia and cardiac murmurs, are not routinely assessed and confirmed in all Japanese governmental health check-ups. Conversely, other potential risk factors, such as urinary proteins, are regularly evaluated during those check-ups but were not part of the original score.9,10 Japanese governmental health check-ups, such as Tokutei-Kenshin and related guidance programs, focus on preventing lifestyle-related diseases, particularly CVD, among middle-aged and older adults. They involve regular health assessments and personalized guidance to manage CVD risk factors. Insurers use health data to identify individuals at risk, provide targeted interventions, and monitor outcomes. These initiatives are key elements of Japan’s strategy to address population aging.10 Another obstacle to the nationwide use of the original risk model is its lack of validation, which could lead to inaccurate predictions, poor generalizability, misclassification of patients, and reduced clinical utility.11
To address these limitations, in this study we aimed to develop a new AF risk model based on the Suita Study data, incorporating risk factors commonly assessed in Japanese governmental health check-ups. The Hisyama Study was used to validate the results and we also created a simple risk score to predict future AF events.
As described elsewhere,12,13 the Suita Study is a population-based prospective cohort study designed to identify CVD risk factors. The study population was selected from Suita, an urban city in Osaka, Japan. It included 2 cohorts randomly selected and stratified by 10-year age groups and sex, as well as a volunteer group. The 2 cohorts comprised 7,814 participants recruited between 1989 and 1998, and the volunteer group comprised 546 participants recruited between 1992 and 2006. All participants underwent a baseline health checkup at the National Cerebral and Cardiovascular Center (NCVC), which involved gathering lifestyle information, clinical assessments, blood sampling, and electrocardiography (ECG). Follow-up visits were scheduled every 2 years.
For this analysis, participants were excluded if they did not attend the baseline assessment or any follow-up, had AF or atrial flutter at baseline, reported fasting <10 h before blood sampling, or lacked baseline data on sex, age, smoking status, alcohol consumption, body mass index (BMI), blood pressure (BP), fasting blood glucose (FBG), high-density lipoprotein-cholesterol (HDL-C), total cholesterol (TC), urinary proteins, glutamic-oxaloacetic transaminase (GOT), glutamic-pyruvic transaminase (GPT), urinary proteins, or history of CVD. Ultimately, 6,476 participants were included in the analysis. Person-years of follow-up were calculated from the date of baseline examination to the date of AF diagnosis, death, censoring, or the last health examination (up to 2015), whichever came first.
Assessment of AFAF ascertainment in the Suita Study was previously described.9 Briefly, AF was diagnosed during biennial follow-up visits at the NCVC. Two trained internists used a standard 12-lead resting ECG and adhered to the Minnesota Codes (8-3-1 to 4). Additionally, participants’ hospital records, including information on AF medication, history of catheter ablation treatment, routine check-ups, and death certificates, were thoroughly reviewed to confirm the diagnosis of AF.
Assessment of Risk FactorsInformation on risk factors was gathered during the initial health check-up. Weight and height were recorded to calculate BMI. BP was measured 3 times, and the average of the last 2 readings was used for statistical analysis. Blood samples were collected and centrifuged before conducting routine blood tests, including FBG, TC, and HDL-C levels. All participants were required to fast for a minimum of 10 h before the blood draw. Urine samples were collected and analyzed.
The following variables were investigated in the current study: age * sex (men <50 years, 50–59 years, 60–69 years, or ≥70 years and women <50 years, 50–59 years, 60–69 years, or ≥70 years), BMI: (<18.5 kg/m2 [underweight], 18.5–24.9 kg/m2 [average weight], or ≥25 kg/m2 [overweight or obesity]), smoking (never, former, current with smoking index [smoking years * number of cigarettes/day] <300, 300–499, or ≥500), alcohol consumption (never, former, <2 gou/day [4 drinks/day], or ≥2 gou/day [4 drinks/day]), BP (<120/80 mmHg [optimal BP], 120–139/80–89 mmHg [high–normal or elevated BP], or ≥140/90 mmHg or receiving medication [Hypertension]), FBG (<100 mg/dL [normal], 100–125 mg/dL [impaired], or ≥126 mg/dL or receiving medication [diabetes]), TC (<200 mg/dL, 200–239 mg/dL, or ≥240 mg/dL), HDL–C (<40 mg/dL, 40–59 mg/dL, or ≥60 mg/dL), urinary proteins (− or −+, +, ++, and +++), GOT (≤50 IU/dL or >50 IU/dL), GPT(≤50 IU/dL or >50 IU/dL), and history of CVD involving stroke and CHD (yes or no).
Statistical AnalysisSAS version 9.4 software (SAS Institute Inc., Cary, NC, US) was used for statistical analysis. The chi-squared test was used to detect differences in the distribution of risk factors between participants with and without AF. Cox regression models, adjusted for age and sex, were used to identify risk factors associated with AF incidence. Factors that showed statistically significant associations were included in a multivariable-adjusted model. The final AF risk model included variables that remained statistically significant in the later model. The model’s discrimination was evaluated by measuring the area under the receiver operating characteristic (AUROC) curve and Harrell concordance index (C-index),14 and its calibration was assessed using the Hosmer-Lemeshow statistic by comparing the observed and expected incident events across deciles of risk.15 For internal validation, we randomly stratified our sample into derivation and validation samples 2/3 to 1/3. The risk prediction model was reconstructed using the derivation sample. We then measured the AUROC of the final model in the validation sample. For external validation, we applied our final model to data from the Hisayama Study, a population-based prospective cohort study involving rural residents of Hisayama town in Fukuoka, Japan,16 and assessed the Harrell C-index of the final model. Later, we assigned weights to the risk factors based on their β-coefficients in the final model, converting these weights into risk scores ranging from zero to 100. Additionally, we calculated the 10-year predictive probability of AF for different scores.
Within a median follow-up of 14.7 years, 283 AF events were diagnosed. The group that developed AF had higher proportions of older adults (60–69 years: 15.2% vs. 12.8% in women and 28.3% vs. 11.9% in men and ≥70 years: 10.2% vs. 7.3% in women and 17.7% vs. 9.2% in men), current smoking with a smoking index >500 (10.2% vs. 6.5%), heavy alcohol consumption (21.5% vs. 14.0%), overweight or obesity (30.4% vs. 19.3%), hypertension (47.3% vs. 31.1%), urinary proteins (23.0% vs. 18.1%), GPT >50 U/dL (20.1% vs. 13.9%), and history of CVD (5.7% vs. 2.3%) (Table 1). In the multivariable-adjusted model, several risk factors were associated with a higher risk of AF. Compared with women aged <50 years, the hazard ratios (HR, 95% confidence interval [CI]) for men were 2.89 (1.19, 7.00), 5.89 (2.52, 13.81), 16.68 (7.31, 38.05), and 37.78 (16.14, 88.47) for ages <50 years, 50–59 years, 60–69 years, and ≥70 years, respectively, and for women aged 50-59 years, 60–69 years, ≥70 years, they were 2.55 (1.01, 6.46), 12.61 (5.57, 28.56), and 32.50 (13.81, 76.49) respectively. Other risk factors included current smoking with a smoking index >500 [HR 1.61 (1.00, 2.61)], alcohol consumption ≥2 gou/day [1.70 (1.15, 2.49)], BMI ≥25 kg/m2 [1.47 (1.13, 1.92)], hypertension [1.65 (1.16, 2.33)], urinary proteins [1.31 (1.00, 1.73)], GPT >50 IU/dL [1.34 (1.00, 1.81)], and history of CVD [2.02 (1.20, 3.40)] (Table 2). The model performed well in terms of discrimination and calibration. The AUROC curve (95% CI) was 0.733 (0.707, 0.760) (Figure 1) and the P value for goodness-of-fit was 0.228 (Figure 2). The corresponding results of the derivation sample were 0.739 (0.690, 0.789) for the AUROC curve (Supplementary Figure) and 0.522 for the P value for goodness-of-fit. The Harrel C-index values of the original and derivation samples were 0.808 and 0.815, respectively. When the risk model was applied to the Hisayama Study (n=3,060 and 222 AF cases within a median follow-up period of 15.2 years), the Harrel C-index was 0.768.
Risk Factors for AF Using Data From the Suita Study
| Risk factors | Overall, % (n=6,746) |
No AF, % (n=6,193) |
AF, % (n=283) |
P value | |
|---|---|---|---|---|---|
| Men | <50 years | 14.2 | 14.5 | 8.1 | <0.001 |
| 50–59 years | 11.0 | 10.9 | 13.4 | ||
| 60–69 years | 12.6 | 11.9 | 28.3 | ||
| ≥70 years | 9.6 | 9.2 | 17.7 | ||
| Women | <50 years | 19.2 | 19.9 | 2.5 | |
| 50–59 years | 13.1 | 13.5 | 4.6 | ||
| 60–69 years | 12.9 | 12.8 | 15.2 | ||
| ≥70 years | 7.4 | 7.3 | 10.2 | ||
| Smoking cigarettes | Never | 54.0 | 54.7 | 40.3 | <0.001 |
| Former | 17.3 | 16.8 | 27.2 | ||
| Smoking index ≤500 | 22.0 | 22.0 | 22.3 | ||
| Smoking index >500 | 6.7 | 6.5 | 10.2 | ||
| Alcohol consumption* | Never | 45.6 | 46.1 | 35.7 | <0.001 |
| Former | 2.6 | 2.6 | 3.9 | ||
| <2 gou/day | 37.4 | 37.3 | 38.9 | ||
| ≥2 gou/day | 14.4 | 14.0 | 21.5 | ||
| BMI (kg/m2) | <18.5 | 8.0 | 8.1 | 6.0 | <0.001 |
| 18.5–24.9 | 72.2 | 72.6 | 63.6 | ||
| ≥25 | 19.8 | 19.3 | 30.4 | ||
| BP (mmHg) | <120/80 | 35.9 | 36.7 | 18.4 | <0.001 |
| 120–139/80–89 | 32.3 | 32.2 | 34.3 | ||
| ≥140/90 or medication | 31.8 | 31.1 | 47.3 | ||
| FBG (mg/dL) | <100 | 67.8 | 68.3 | 57.6 | 0.001 |
| 100–125 | 27.3 | 26.9 | 35.3 | ||
| ≥126 or medication | 4.9 | 4.8 | 7.1 | ||
| HDL (mg/dL) | <40 | 15.2 | 15.0 | 19.1 | 0.002 |
| 40–59 | 52.1 | 51.8 | 58.6 | ||
| ≥60 | 32.7 | 33.2 | 23.3 | ||
| TC (mg/dL) | <200 | 42.0 | 41.9 | 43.8 | 0.775 |
| 200–239 | 39.8 | 39.9 | 37.8 | ||
| ≥240 | 18.2 | 18.2 | 18.4 | ||
| Urinary proteins | – | 81.7 | 81.9 | 77.0 | 0.038 |
| −+/+/++/+++ | 18.3 | 18.1 | 23.0 | ||
| GOT (IU/dL) | ≤50 | 88.5 | 88.6 | 85.9 | 0.156 |
| >50 | 11.5 | 11.4 | 14.1 | ||
| GPT (IU/dL) | ≤50 | 85.9 | 86.1 | 79.5 | 0.003 |
| >50 | 14.1 | 13.9 | 20.1 | ||
| History of CVD | No | 97.5 | 97.7 | 94.3 | <0.001 |
| Yes | 2.5 | 2.3 | 5.7 | ||
AF, atrial fibrillation; BMI, body mass index; BP, blood pressure; CVD, cardiovascular disease; FBG, fasting blood glucose; GOT, glutamic-oxaloacetic transaminase; GPT, glutamic-pyruvic transaminase; HDL, high-density lipoprotein; TC, total cholesterol. *2 gou/day=4 drinks/day.
Associations With AF Using Data From the Suita Study
| Risk factors | AF cases per 1,000 person-years |
Model I* | Model II** | |
|---|---|---|---|---|
| HR (95% CI) | HR (95% CI) | |||
| Men | <50 years | 1.60 | 4.76 (2.04, 11.08) | 2.89 (1.19, 7.00) |
| 50–59 years | 3.39 | 10.67 (4.76, 23.88) | 5.89 (2.52, 13.81) | |
| 60–69 years | 7.46 | 28.90 (13.30, 62.81) | 16.68 (7.31, 38.05) | |
| ≥70 years | 10.03 | 59.13 (26.52, 131.80) | 37.78 (16.14, 88.47) | |
| Women | <50 years | 0.34 | 1 (Ref.) | 1 (Ref.) |
| 50–59 years | 0.95 | 2.90 (1.16, 7.27) | 2.55 (1.01, 6.46) | |
| 60–69 years | 4.01 | 15.45 (6.93, 34.46) | 12.61 (5.57, 28.56) | |
| ≥70 years | 7.10 | 39.71 (17.23, 91.51) | 32.50 (13.81, 76.49) | |
| Smoking cigarettes | Never | 2.23 | 1 (Ref.) | 1 (Ref.) |
| Former | 5.38 | 1.15 (0.79, 1.67) | 1.06 (0.73, 1.54) | |
| Smoking index ≤500 | 3.21 | 1.17 (0.80, 1.70) | 1.16 (0.80, 1.71) | |
| Smoking index >500 | 5.59 | 1.71 (1.07, 2.75) | 1.61 (1.00, 2.61) | |
| Alcohol consumption† | Never | 2.48 | 1 (Ref.) | 1 (Ref.) |
| Former | 5.75 | 1.62 (0.85, 3.07) | 1.45 (0.76, 2.78) | |
| <2 gou/day | 3.18 | 1.11 (0.82, 1.49) | 1.15 (0.85, 1.55) | |
| ≥2 gou/day | 4.73 | 1.87 (1.28, 2.73) | 1.70 (1.15, 2.49) | |
| BMI (kg/m2) | <18.5 | 2.50 | 1.00 (0.60, 1.64) | 1.09 (0.66, 1.81) |
| 18.5–24.9 | 2.72 | 1 (Ref.) | 1 (Ref.) | |
| ≥25 | 4.99 | 1.68 (1.30, 2.18) | 1.47 (1.13, 1.92) | |
| BP (mmHg) | <120/80 | 1.46 | 1 (Ref.) | 1 (Ref.) |
| 120–139/80–89 | 3.27 | 1.47 (1.05, 2.07) | 1.35 (0.95, 1.91) | |
| ≥140/90 or medication | 5.41 | 1.94 (1.39, 2.72) | 1.65 (1.16, 2.33) | |
| FBG (mg/dL) | <100 | 2.60 | 1 (Ref.) | 1 (Ref.) |
| 100–125 | 4.21 | 1.10 (0.85, 1.41) | – | |
| ≥126 or medication | 5.53 | 1.39 (0.87, 2.22) | – | |
| HDL (mg/dL) | <40 | 4.10 | 0.92 (0.67, 1.25) | – |
| 40–59 | 3.47 | 1 (Ref.) | 1 (Ref.) | |
| ≥60 | 2.20 | 0.80 (0.60, 1.07) | – | |
| TC (mg/dL) | <200 | 3.27 | 1 (Ref.) | 1 (Ref.) |
| 200–239 | 2.98 | 0.81 (0.62, 1.05) | – | |
| ≥240 | 3.19 | 0.83 (0.59, 1.16) | – | |
| Urinary proteins | – | 2.91 | 1 (Ref.) | 1 (Ref.) |
| −+/+/++/+++ | 4.28 | 1.40 (1.06, 1.85) | 1.31 (1.00, 1.73) | |
| GOT (IU/dL) | ≤50 | 3.00 | 1 (Ref.) | 1 (Ref.) |
| >50 | 4.39 | 1.25 (0.89, 1.75) | – | |
| GPT (IU/dL) | ≤50 | 2.90 | 1 (Ref.) | 1 (Ref.) |
| >50 | 4.71 | 1.53 (1.14, 2.05) | 1.34 (1.00, 1.81) | |
| History of CVD | No | 3.01 | 1 (Ref.) | 1 (Ref.) |
| Yes | 11.15 | 2.24 (1.34, 3.75) | 2.02 (1.20, 3.40) | |
*Adjusted for age and sex. **Adjusted for variables significant in Model I. †2 gou/day=4 drinks/day. CI, confidence interval; IU, international units. Other abbreviations as in Table 1.
Receiver-operating characteristic curve. (C-statistics=0.733; 95% confidence interval (CI): 0.707, 0.760.)
Hosmer-Lameshow goodness-of-fit test. (P value for goodness-of-fit=0.228.)
The risk scores of the variables contributing to AF risk were as follows: age [<50 years (16 points in men), 50–59 years (14 points in women and 26 points in men), 60–69 years (37 points in women and 41 points in men), and ≥70 years (51 points in women and 54 points in men)], current smoking with smoking index >500 (7 points), heavy alcohol consumption (8 points), BMI ≥25 kg/m2 (6 points), hypertension (7 points), urinary proteins (4 points), GPT >50 IU/dL (4 points), and CVD history (10 points) (Table 3). The 10-year predicted probability of AF events was 0.3%, 0.7%, 1.1%, 2.3%, 3.2%, 4.1%, 7.1%, 8.4%, 10.8%, and 15.9% for scores <10, 10–17, 18–24, 25–33, 34–42, 43–46, 47–54, 55–58, 59–69, and ≥70, respectively (Figure 3).
Risk Score for AF
| Risk factors | β | Score* |
|---|---|---|
| Men | ||
| <50 years | 1.06 | 16 |
| 50–59 years | 1.77 | 26 |
| 60–69 years | 2.81 | 41 |
| ≥70 years | 3.63 | 54 |
| Women | ||
| 50–59 years | 0.94 | 14 |
| 60–69 years | 2.53 | 37 |
| ≥70 years | 3.48 | 51 |
| Current smoking (smoking index >500) | 0.48 | 7 |
| Heavy alcohol consumption (≥2 gou/day)** | 0.53 | 8 |
| BMI ≥25 kg/m2 | 0.38 | 6 |
| BP ≥140/90 mmHg or medications | 0.50 | 7 |
| Urinary proteins (−+/+/++/+++) | 0.27 | 4 |
| GPT >50 U/dL | 0.29 | 4 |
| History of CVD | 0.70 | 10 |
*Highest possible score is 100. **2 gou/day = 4 drinks/day. Abbreviations as in Table 1.
New risk score prediction of the 10-year probability of atrial fibrillation.
Based on the results from this study, we introduce a new AF risk model using variables similar to those collected in Japanese governmental health check-ups. Within a median follow-up of 14.7 years, male sex, older age, heavy smoking, heavy drinking, overweight or obesity, hypertension, urinary proteins, elevated liver enzymes, and history of CVD were associated with an increased risk of AF. The risk model performed well in terms of discrimination, calibration, and internal and external validations. For health education and counseling purposes, we developed a risk score to calculate the 10-year predicted probability of AF. This study complements our recent study that used data from the Suita Study, to develop new risk scores for stroke, CHD, and atherosclerotic CVD.17
The new model shares the following risk factors with the previous model:9 age, sex, hypertension, overweight or obesity, smoking, excessive drinking, and CVD history. However, certain differences were observed between the models. First, the previous model included arrhythmia and cardiac murmurs, which were removed from the current model because they are not routinely assessed in Japanese governmental health check-ups. Second, the previous model assessed chronic kidney disease (CKD) using the estimated glomerular filtration rate (eGFR) and found no association with AF. In contrast, the current study assessed proteinuria, which is routinely measured in Japanese health check-ups, and detected a strong association with AF risk. A retrospective cohort study using Canadian administrative healthcare databases showed a positive association between proteinuria and the risk of AF in patients with intact kidney function.18 Similarly, the Korean National Health Insurance Service showed a higher risk of AF among those with proteinuria even after adjusting for eGFR.19 Proteinuria may indicate kidney damage, leading to systemic inflammation and endothelial dysfunction, contributing to atrial remodeling and increased AF risk.20,21 Overactivation of the renin-angiotensin-aldosterone system (RAAS), a common condition with proteinuria, promotes hypertension and left ventricular hypertrophy, creating a substrate for AF.22,23 Third, the previous model did not examine the role of liver enzymes in AF development, whereas the current model showed a positive association between elevated GPT and AF risk. Some have investigated the association between liver enzymes and AF risk and reached inconclusive findings.24 Elevated liver enzymes may be associated with AF because they reflect underlying metabolic disturbances, such as non-alcoholic fatty liver disease (NAFLD) and systemic inflammation. NAFLD is often associated with obesity, insulin resistance, and metabolic syndrome, all of which are risk factors for AF. The liver plays a crucial role in regulating systemic inflammation and oxidative stress, promoting atrial remodeling, fibrosis, and electrophysiological changes, increasing susceptibility to AF.25–27 Fourth, unlike this new model, the previous one was neither internally nor externally validated.
Study StrengthsApplying a prospective cohort design with a long follow-up period and biennial check-ups, investigating a randomly recruited sample from the general population, and using standardized methods to ascertain AF risk factors are among the main strengths of this study. We also identified AF using ECGs, hospital records, and death certificates, improving on studies that used only ECGs. Most participants who developed CVD during follow-up were hospitalized, and almost all of them were examined by Holter ECG, allowing for the detection of paroxysmal AF. Furthermore, the Suita Study reflects an urban general population, and the Hisayama Study represents a rural population, indicating that our findings may be generalizable to the broader Japanese population.
Study LimitationsFirst, the risk factors were only assessed at baseline. Participants with chronic diseases might have been closely monitored, reducing their contribution to the AF risk. For instance, our results showed that heavy smoking was associated with a higher risk of AF, whereas lighter smoking was not. Over half of the Suita Study participants quit smoking during follow-up, potentially weakening the association between smoking and AF. Second, some lifestyle habits that could influence AF, such as physical activity, diet, and sleep,28,29 are not routinely assessed in Japanese governmental health check-ups and were therefore excluded from this risk model. Third, Holter ECG monitoring was not conducted for all participants, and even if it had been, it may have failed to detect cases of paroxysmal AF. To identify incident AF as accurately as possible, we relied on multiple sources, including ECG results from medical examinations, medical records, and causes of death (accounting for 26% of the total cases). Fourth, some participants who were censored due to death may have had incomplete or imprecise medical records. Fifth, age ≥70 years contributed to more than half of the score, suggesting that older adults may not derive the same level of benefit from lifestyle changes as their younger and middle-aged counterparts. Therefore, our study underscores the significance of lifestyle modifications, particularly for young and middle-aged individuals.
In conclusion, from the Suita Study we developed a new risk model to predict AF in Japanese individuals using traditional CVD risk factors. The model identifies modifiable risk factors, such as heavy smoking, heavy alcohol consumption, BMI, BP, urinary proteins, and GPT. The early detection and management of these conditions can potentially reduce AF risk. Unlike the previous Suita model, the new model is more practical for nationwide health guidance because it incorporates variables routinely assessed in Japanese governmental health check-ups and because its validation was established.
We thank the Suita Medical Association, Suita City Health Center Staff, and all cohort members.
The Suita study was supported by the Intramural Research Fund for CVD of the NCVC (23-B-9 and 20-4-9), Grants-in-Aid for Scientific Research (16H05252, 24K02718, and 24K23771), Japan Health Research Promotion Bureau (JH1-1, 2024-B-05), Japan Science and Technology Agency (JPMJPF2018), a Grant-in-Aid from the Japanese Ministry of Health, Labour and Welfare (23FA0501), the Meiji Yasuda Research Institute, Inc., and the Meiji Yasuda Life Insurance Company.
The Hisayama Study was supported by the Ministry of Education, Culture, Sports, Science and Technology of Japan (JP22K07421, JP22K17396, JP23K09692, JP23K09717, JP23K16330, JP23K06787, and JP23K09060), Health and Labour Sciences Research Grants of the Ministry of Health, Labour and Welfare of Japan (JPMH23FA1006, JPMH23FA1022, and JPMH24GB1002), the Japan Agency for Medical Research and Development (JP24dk0207053, JP24 km0405209, and JP24tm0524003), and the Japan Science and Technology Agency (JPMJPF2210).
The authors have no conflicts of interest to declare.
The protocols of the Suita Study and the Hisayama study were approved by the corresponding institutional review boards (M25-043-5, R21024-3, and 2022-151, respectively). This study was conducted according to the Declaration of Helsinki. All participants provided informed consent.
For ethical and administrative reasons, the data cannot be shared.
Please find supplementary file(s);
https://doi.org/10.1253/circj.CJ-24-0939
