The Japan Collaborative Cohort (JACC) study, sponsored by the Ministry of Education, Culture, Sports, Science and Technology of Japan, is a nationwide multicenter collaborative study designed to assess prospectively various risk factors and protective factors for cancer incidence and mortality. Details of the methodology used in this study have been described elsewhere. . Briefly, a baseline survey was conducted from 1988 to 1990. A total of 110,585 participants (46,395 males and 64,190 females), aged 40 to 79, residing in 45 regions in seven of Japan’s eight districts, completed self-administered questionnaires. In our study, we recruited residents living in 22 target areas. In 20 regions, registrants were those who had undergone a basic health examination carried out in accordance with the law on medical and health services for the elderly; in two regions, enrollees were identified on the basis of a health check for atomic bomb survivors.
Among the 64,190 women who took part in the study, 13,345 were excluded because they did not answer the questions concerning parity and/or level of education on the questionnaire. In addition, 7330 women were excluded because they did not provide complete information on parity and/or level of education. In addition, 2273 women were excluded due to a history of stroke, coronary artery disease or cancer. Therefore, 41,242 women were included in the analysis. There were no substantial differences in risk factors for CVD and other diseases between individuals with and without complete information regarding parity and/or education level.
Follow-up surveys were conducted every 1-2 years to determine participant survival. Causes and dates of death were identified by reviewing all death certificates from each region. According to the 10e revision of the International Classification of Diseases (ICD), mortality was classified as due to stroke (I60-I69), coronary artery disease (I20-I25), total cardiovascular disease (I01-I99), cancer ( C00-C97) and CVD and non-cancer causes (all other ICD-10 codes). These results were monitored until December 31, 2009, except in several areas where monitoring ended at the end of 1999 (6 areas), 2003 (5 areas) or 2008 (2 areas). The median follow-up period was 19.1 years (interquartile range 11.2–20.6 years).
Baseline participant characteristics were obtained using self-administered questionnaires that assessed medical history and lifestyle items, such as diet, physical activity, alcohol consumption, smoking and education level. For women, the questionnaires also assessed parity status, age at menopause, and history of sex hormone use.
Parity status categories
Reproductive history was assessed using the following self-reported question: “How many times have you given birth?” We defined the number provided in the response as the parity number. We did not further check with the women if they understood the meaning of the question. The parity number given could then be taken to include the number of stillbirths and/or miscarriages. Based on the response to this question, women with a parity number of 0 were classified as nulliparous women (nulliparity) and those with a parity number ≥ 1 were classified as parous women (parity).
Education level categories
The level of education was assessed using a single question: “How old were you when you last attended school?” For this study, responses were categorized into low education (≤ 15 years) and high education (> 16 years).
Parity and level of education were categorized as nulliparous women with a low level of education, parous women with a low level of education, nulliparous women with a high level of education and parous women with a level of education high (the reference group). Based on these categories, we calculated mean age, age-adjusted mean, and prevalence of cardiovascular disease and other covariance diseases for mean values and a logistic regression model for prevalence. Cox proportional hazards regression analyzes were then performed to calculate age-adjusted and multivariate hazard ratios (RRs) and 95% confidence intervals (95% CI) to determine associations between the four categories and mortality associated with stroke, coronary artery disease, total CVD, cancer, non-cardiovascular and non-cancer causes and all causes. All covariates were chosen based on previous research [3,4,5, 11, 20, 21], and were adjusted at the same time. We constructed multivariate risk models adjusted for baseline age (years), body mass index (BMI) (kg/m2), history of sex hormone use (yes or no), usual weekly hours of exercise and/or walking (almost never, 1–2 hrs, 3–4 hrs, and ≥ 5 hrs), daily hours of exercise and/or walking (almost never, 0 indicates the additive interaction of nulliparity and low educational attainment. The RERI reflects the excess risk due to the interaction of nulliparity and low educational attainment compared to the parity risk combined with high educational attainment. We calculated the RERI and that of the 95% confidence interval and P-value for cardiovascular disease and other deaths using the delta method (see Supplementary File 1) . Additionally, we tested age-adjusted rates for cardiovascular mortality, non-cardiovascular and non-cancer mortality, and all-cause mortality by parity and education level using the direct standardization method; this was based on deaths per 1000 persons per year and the age distribution of the national model population in 1985. All analyzes were performed using SAS software (version 9.4; SAS Institute Inc., Cary, NC, USA), and the differences were considered statistically significant at a two-sided P value