Study population
We used data obtained from the Tohoku Medical Megabank Project Birth and Three-Generation Cohort Study (TMM BirThree Cohort Study), which has been described elsewhere [15, 16]. Pregnant women and their family members were contacted in approximately 50 obstetric clinics or hospitals in Miyagi Prefecture when they scheduled their deliveries from 2013 to 2017. Tohoku University Tohoku Medical Megabank Organization established seven community support centers in Miyagi Prefecture as local facilities for voluntary admission-type recruitment and health assessment of the participants [17]. Trained genome medical research coordinators were placed in each clinic, hospital, or community support center to provide information on the TMM BirThree Cohort Study to potential participants and to receive signed informed consent forms from each participant. Of 32,986 pregnant women who were contacted, 22,493 agreed to participate. Among them, 10,288 women were excluded because of abortion or stillbirth, nonidentification of their birth status, incomplete questionnaires, no permission to transcribe their medical records, and pregnancy 1 year after delivery. Of the remaining 12,205 women, 1174 were excluded due to missing values for their drinking status 1 year after delivery, educational attainment, equivalent household income, parity, drinking status during pregnancy, work status, postpartum depression, or breastfeeding. The present study analyzed the remaining 11,031 women (Fig. 1). The Ethics Committee of Tohoku University Tohoku Medical Megabank Organization reviewed and approved the TMM BirThree Cohort Study protocol (2013-1-103-1). The characteristics of the 11,031 analyzed women and the 11,462 excluded women are shown in Table S1.
Measures
Based on the questionnaire 1 year after delivery, educational attainment was categorized as high school or lower (elementary, junior high school, or senior high school), college (2-year college or special training school), and university or higher (university or graduate school). Using the questionnaire during pregnancy, the participants were asked to select their total annual household income from seven categories: ≤1.99, 2.00–3.99, 4.00–5.99, 6.00–7.99, 8.00–9.99, 10.00–11.99, and ≥12.00 million Japanese yen (JPY). The equivalent household income was calculated as the household income divided by the square root of the number of family members [18]. The resulting value was categorized into four groups approximately corresponding to quartiles: ≤1.99, 2.00–2.99, 3.00–3.99, and ≥4.00 million JPY.
Using the questionnaire 1 year after delivery, the participants were asked to select one of the following response options for their drinking status: current drinker, past drinker, never-drinker, and constitutionally never-drinker. Current drinkers were asked to report the types of alcoholic beverages consumed as well as the frequency and amount of each consumption per day in the past year: sake (rice wine), shochu, beer, whisky, and/or wine. A score to each category of alcohol consumption frequency was assigned as follows: 7 for every day, 5.5 for 5–6 days/week, 3.5 for 3–4 days/week, 1.5 for 1–2 days/week, 0.5 for a few times/month, and 0 for almost never. The ethanol equivalent intake in grams was calculated as follows: 180 ml sake = 23 g, 180 ml shochu = 36 g, 633 ml beer = 23 g, 30 ml whisky = 10 g, and 100 ml wine = 12 g [19]. Finally, the weekly ethanol equivalent intake was estimated by multiplying the amount of ethanol by the frequency score; the daily ethanol equivalent intake was calculated by dividing these estimates by 7. Past drinkers, never-drinkers, and constitutionally never-drinkers were considered abstainers and were assigned 0 g/day. The second term of Health Japan 21, which is a 10-year national plan begun in 2013, has established the targets relating to reducing the number of people drinking quantities of alcohol that increase the risk of lifestyle-related disease onset: ethanol intake of ≥40 g/day for men and ≥20 g/day for women [20]. Following this national recommendation, the threshold of hazardous drinking was defined as ethanol intake of ≥20 g/day in the present study. This is in line with the thresholds observed in several countries [4].
As covariates, we chose age, parity, drinking status during pregnancy, work status, postpartum depression, and breastfeeding. Age at delivery and parity were ascertained from the participants’ medical records. Age at delivery was categorized as ≤29, 30–34, and ≥35 years. Parity was dichotomized into nulliparous and multiparous. Using the questionnaire during pregnancy, the participants were asked to select one of the following responses concerning their drinking status: current drinker, past drinker, never-drinker, and constitutionally never-drinker. Responses were dichotomized into drinking (current drinker) and non-drinking (past drinker, never-drinker, and constitutionally never-drinker). Information on the participants’ work status, postpartum depression, and breastfeeding was obtained from the questionnaire 1 year after delivery. The work status was dichotomized as working (permanent worker, self-employed worker, temporary worker, or part-time worker) and not working (on leave, pensioner, housewife, student, or unemployed). The participants provided responses to the Japanese version of the Edinburgh Postpartum Depression Scale, which comprises 10 items assessing any symptoms of depression in the past 7 days [21]. Each item has four possible responses with scores of 0–3, and the total score ranges from 0 to 30. Postpartum depression was defined as a score of ≥9 [22]. All participants were also asked whether their child had been breastfed.
Statistical analysis
The characteristics of non-drinkers and drinkers 1 year after delivery were compared using the chi-square test. We conducted multiple logistic regression analyses to examine the associations of education and income with hazardous drinking 1 year after delivery. The odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for education and income, adjusted for age (model 1), as well as for parity, drinking status during pregnancy, work status, postpartum depression, and breastfeeding (model 2a for education and model 2b for income). Income and education were then entered simultaneously in the same model (model 3). We also conducted sensitivity analyses restricted to only drinkers 1 year after delivery.
We entered the interaction term between education and income into the model and found that the interaction was statistically significant (p < 0.01). Based on this result, we also conducted stratified analyses of the association between education and hazardous drinking by income groups (≥3.00 vs. 0–2.99 million JPY) and the association between income and hazardous drinking by education groups (college or higher vs. high school or lower).
All analyses were conducted using the SAS version 9.4 software (SAS Institute Inc., Cary, NC, USA). For all analyses, a two-tailed p value of <0.05 was considered statistically significant.