Study cohorts
Analyses included women enrolled in one of two pregnancy cohorts based in Boston, USA and New York City, USA with similarly derived information on fetal sex and maternal depressive symptoms.
Asthma Coalition on Community Environment and Social Stress (ACCESS). English- or Spanish-speaking women (≥ 18 years) with a singleton pregnancy were enrolled from Brigham & Women’s Hospital (BWH), Boston Medical Center (BMC), and affiliated prenatal clinics from 2001 to 2010. N = 970 mothers delivered a live born infant and remained eligible for study follow-up. Procedures were approved by the human studies committees at BWH and BMC; written informed consent was obtained in the mother’s preferred language.
Programming of Intergenerational Stress Mechanisms (PRISM). Pregnant English- or Spanish-speaking women (≥ 18 years) with a singleton pregnancy were recruited from prenatal clinics at the East Boston Neighborhood Health Center, the BWH, and the Beth Israel deaconess Medical Center (BIDMC) in Boston and from the Mount Sinai Hospital in New York City. Exclusion criteria included HIV positive status and drinking > 7 alcoholic drinks per week prior to pregnancy recognition or any alcohol after pregnancy recognition. Recruitment began in 2011 and is ongoing; the present study considered N = 987 eligible women with a live born infant enrolled through December 2019. Procedures were approved by the institutional review boards at BWH and the Icahn School of Medicine at Mount Sinai; BIDMC relied on BWH for review and oversight of the protocol. Written informed consent was obtained in the mother’s preferred language.
In ACCESS, women excluded due to missing depression assessments were more likely to be Black/Black-Hispanic (34% vs. 28%), and less likely to be Hispanic, non-Black (49% vs. 59%) or to have breastfed for more than 6 months (47% vs. 64%) (Supplemental Table 1). Similarly, in PRISM, excluded women were more likely to be Black/Black-Hispanic (51% vs. 38%), and less likely to be White, non-Hispanic (12% vs. 20%), to be over age 35 years (15% vs. 20%) or to have breastfed for more than 6 months (26% vs. 57%) (Supplemental Table 2). The differences in age and breastfeeding duration between excluded and included women likely reflects differences in race/ethnicity, as Black/Black-Hispanic women in the cohorts were younger and less likely to breastfeed for more than 6 months.
Maternal perinatal depression symptoms
We measured maternal depression symptoms in-person or by telephone during the prenatal (mean ± SD, ACCESS: 27.3 ± 8.2 weeks gestation, PRISM: 29.0 ± 8.0 weeks gestation) and postnatal (ACCESS: 5.0 ± 1.6 months postpartum, PRISM: 6.6 ± 0.8 months postpartum) periods using the Edinburgh Postpartum Depression Scale (EPDS). The EPDS was developed as a screening assessment for identifying women with perinatal depression (29). The validated tool is available in English and Spanish and has sensitivity and specificity of 95% and 93% compared to Diagnostic and Statistical Manual of Mental Disorders, Third Edition criteria, respectively (29–31). Women rated each of 10 questions on a four-point scale (0–3) reflecting how often she experienced various symptoms associated with depression during the previous seven days. Consistent with prior studies, we defined elevated depressive symptomology as an EPDS score > 10 (32). Of the 961 women enrolled in ACCESS, 718 (75%) completed the EPDS during pregnancy and 346 (36%) additionally completed the EPDS during the postnatal period (Fig. 1). Of the 987 eligible women enrolled in PRISM, 693 (70%) completed the EPDS during pregnancy, and 528 (53%) additionally completed the EPDS during the postnatal period (Fig. 1). Prenatal EPDS scores did not significantly differ between women with versus without a postnatal measure in either cohort (Wilcoxon rank sum: ACCESS: Z = 0.58, p = 0.56; PRISM: Z = 1.13, p = 0.26). Only women with complete EPDS measures at both periods were included in the study.
Mothers were additionally asked a series of questions about depressive feelings, depression diagnosis, and the use of prescription medication for depression (PRISM only) before and during pregnancy. Specifically, mothers were asked: “Before this pregnancy, was there ever a period of time when you were feeling depressed or down or when you lost interest in pleasurable activities most of the day, nearly every day, for at least 2 weeks?”, “Before this pregnancy, did you ever see a health care professional who said that you were depressed?”, and “During your pregnancy, did you take prescription medication for depression or to help you sleep?”
Child sex
Child sex was reported by the mother at a study visit conducted approximately one month following delivery. In the PRISM cohort, self-reported newborn sex was confirmed by review of the baby’s medical record.
Covariates
Data on sociodemographic characteristics (race/ethnicity, age, education, relationship status) and lifestyle factors (parity, pre-pregnancy body mass index [BMI], smoking during pregnancy) were collected using questionnaires administered in-person by trained study staff during pregnancy. Information on breastfeeding duration was assessed by in-person or telephone questionnaires administered through age 6 months. In PRISM, information on admittance to the Neonatal Intensive Care Unit (NICU) and gestational age at birth was abstracted from the newborn’s medical record. Gestational age was based on best obstetrical estimate derived from routine prenatal ultrasounds; if no obstetrical estimate was available, gestational age was calculated from the date of birth and maternal reported last menstrual period.
Statistical analyses
We calculated descriptive statistics for all sociodemographic and lifestyle characteristic considered. We examined the distribution of maternal EPDS scores using histograms and boxplots and calculated Spearman correlations between prenatal and postnatal scores. We created a four-level categorical variable that indicated the presence of depressive symptoms (defined as an EPDS score > 10) during the prenatal and/or postnatal period as follows: 1) no depressive symptoms during either period, 2) depressive symptoms during the prenatal period only, 3) depressive symptoms during the postnatal period only, and 4) depressive symptoms during both periods. In each cohort considered separately, we used logistic regression to examine the odds of maternal depressive symptoms during the prenatal period only, postnatal period only, or both periods versus no depressive symptoms during either period among women carrying a male versus female fetus. We used the Firth penalized maximum likelihood estimation method to reduce potential bias introduced by sparse cases across the four levels of prenatal versus postnatal depressive symptoms (33). Given that there are few known determinants of fetal sex, the relationship between fetal sex and maternal depressive symptoms is unlikely susceptible to confounding by shared antecedents of the exposure and outcome. Here, we examined both unadjusted models, as well as models adjusted for a core set of sociodemographic variables (maternal race/ethnicity [White, non-Hispanic/other vs. Black/Black-Hispanic vs. Hispanic, non-Black], age [< 35 vs. ≥35 years], education [< high school vs. ≥high school]) and lifestyle or health-related precision variables that have previously been associated with depression, including: relationship status (married or living with partner vs. other), parity (nulliparous vs. multiparous), pre-pregnancy BMI (continuous in kg/m2), cigarette smoking during pregnancy (any vs. none), and breast feeding duration (≤ 6 months vs. >6 months). For regression models, women who self-reported their race/ethnicity as “other” (approximately 5%) were collapsed with white women due to small numbers across levels of the depression variable. In adjusted models, we used multiple imputation with chained equations to impute values for missing covariates (ACCESS: n = 47, 14%; PRISM: n = 26, 5%).
To understand potential misclassification related to our use of the EPDS, we used separate multinomial logistic regression models to examine maternal self-reported feelings of depression before pregnancy, diagnosis with depression (ever) before pregnancy, or use of depression medication during pregnancy (PRISM only) in relation to this 4-level EPDS variable.
Evidence suggests that boys may be more susceptible to neonatal complications, which in turn may lead to increased risk for maternal depression. To evaluate this possibility, we examined bivariate associations between neonatal complications, defined as premature birth (< 37 weeks gestation) or admittance to the NICU (yes/no) and maternal EPDS scores in the PRISM cohort. All statistical analyses were perform using SAS v9.4.