Statistical analysis

First, we described the baseline characteristics of the study participants. Next, we stratified the study participants into two groups based on the presence or absence of CVD. The groups were then compared using the Student’s t- or chi-square test. We calculated the hazard ratios and their 95% confidence intervals (CIs) for CVD onset using Cox proportional hazards models, with leisure-time and household physical activities and confounding factors. Data on the amount of leisure-time and household physical activity per day was divided into three groups: men and women, and men and women combined. In addition, the amount of leisure-time and household physical activity was divided into tertile (lowest, middle, and highest) for men and women, and men and women combined. The lowest tertile was defined as the reference group. The types of leisure-time and household physical activities were divided into two groups (0–1 type and ≥ 2 types). The group with 0–1 type of leisure-time and household physical activity was defined as the reference group. Furthermore, the types of leisure-time and household activities (0–1 type and ≥ 2 types) and the tertile of leisure-time and household physical activities (lowest, middle, and highest) were combined and evaluated. Leisure time and household activity was then stratified into two groups, and hazard ratios and 95% confidence intervals for type of leisure-time and household physical activity and CVD incidence were calculated by Cox proportional hazards model, adjusting for confounding factors. In the Cox proportional hazards models, the following factors were considered in multivariable models as confounding factors: sex and age adjusted model + adjusted for smoking (currently, past, never), drinking (currently, past, never), job (y/n), BMI, hypertension (y/n), dyslipidemia(y/n), diabetes (y/n), and physical activity other than during leisure time (sleep, occupation activity, transportation, other). Additionally, sex- and age- adjusted models were also conducted as for reference. All of these variables, with the exception of the follow-up period, used information that was assessed at baseline. All the analyses were first performed for the total population, following which sex-stratified analyses were conducted. Both the primary outcome (CVD development) and secondary outcome (stroke and AMI development) were evaluated. Sex-stratified analyses were performed separately. As the number of AMI events was small in this study, we could not analyze the relationship between leisure-time and household physical activities and AMI onset. All analyses were performed using SPSS version 22 (IBM Japan, Tokyo). The level of significance was set at 5% on both sides.