Maternal exposure to childhood maltreatment and adverse birth outcomes

Exposure to traumatic events during pregnancy may influence pregnancy and birth outcomes. Growing evidence suggests that exposure to traumatic events well before pregnancy, such as childhood maltreatment (CM), also may influence the course of pregnancy and risk of adverse birth outcomes. We aimed to estimate associations between maternal CM exposure and small-for-gestational-age birth (SGA) and preterm birth (PTB) in a diverse US sample, and to examine whether common CM-associated health and behavioral sequelae either moderate or mediate these associations. The Measurement of Maternal Stress (MOMS) Study was a prospective cohort study that enrolled 744 healthy English-speaking participants ≥ 18 years with a singleton pregnancy, who were < 21 weeks at enrollment, between 2013 and 2015. CM was measured via the Childhood Trauma Questionnaire (CTQ) and participants above the moderate/severe cut-off for any of the five childhood abuse and neglect scales were assigned to the CM-exposed group. Common CM-associated health (obesity, depressive symptoms, hypertensive disorders) and behavioral (substance use) sequelae were obtained from standardized questionnaires and medical records. The main outcomes included PTB (gestational age < 37 weeks at birth) and SGA (birthweight < 10%ile for gestational age) abstracted from the medical record. Multivariable logisitic regression was used to test associations between CM, sequeale, and birth outcomes, and both moderation and mediation by CM-related sequelae were tested. Data were available for 657/744 participants. Any CM exposure was reported by 32% of participants. Risk for SGA birth was 61% higher among those in the CM group compared to the non-CM group (14.1% vs. 7.6%), and each subsequent form of CM that an individual was exposed to corresponded with a 27% increased risk for SGA (aOR 1.27, 95% CI 1.05, 1.53). There was no significant association between CM and PTB (9.3% vs. 13.0%, aOR 1.07, 95% CI 0.58, 1.97). Of these sequelae only hypertensive disorders were associated with both CM and SGA and hypertensive disorders of pregnancy did not mediate the association between CM and SGA. Our findings indicate that maternal CM exposure is associated with increased risk for SGA birth and highlight the importance of investigating the mechanisms whereby childhood adversity sets the trajectory for long-term and intergenerational health issues.

www.nature.com/scientificreports/ cut off. Finally, recognizing that these individual domains are highly correlated, we assessed cumulative CM, defined as the sum of the number of domains at or above the "moderate to severe" cutoff (range of 0-5), as our exposure variable. Covariates: The analyses included covariates that might provide alternative explanations for an observed association between CM and either SGA or PTB. Age in pregnancy has a bi-modal association with risk for adverse pregnancy outcomes 42 , and was categorized as ≤ 20, 21-34, and ≥ 35 years. Prevalence of CM as well as adverse pregnancy outcomes are more frequent among Black and Hispanic/Latine individuals 43 , so we assessed ethnicity and race as White, Black, Hispanic/Latine, and 'Other' . Because in the US CM 44 , PTB 45 , and SGA 46 are more prevalent in families of low socioeconomic status, we included measures of both childhood socioeconomic status and current socioeconomic status: Childhood poverty, indicating whether the participant's family received public assistance before the participant's 18th birthday; and current educational attainment, which is highly correlated with adult income and was coded as high school or less, some college or associates, and bachelor's degree or more.
CM-associated sequelae as potential mediators or moderators: Potential mediators/moderators included variables from 4 categories that have been reported as common health and behavioral CM sequelae. Mental health was indexed by presence of depressive symptoms in mid pregnancy based on a score of 16 or greater on the CES-D, a widely used tool designed to evaluate depressive symptomatology 32,47,48 . Hypertensive disorders of pregnancy included gestational hypertension, pre-eclampsia, or eclampsia diagnosed during pregnancy and documented in the medical record 25,49 . Any substance use during pregnancy was defined as any alcohol consumption, recreational drug use (marijuana, cocaine, heroin), or smoking since becoming pregnant. (4) Metabolic risk was indexed by an individual's pre-pregnancy body mass index (BMI, kg/m 2 ), categorized as underweight (< 18.5), normal weight (18.5 to < 25), or overweight/obese (≥ 25) 50 .

Statistical analysis.
The associations between the three CM exposure variables and birth outcomes (SGA, PTB) were examined via multivariable logistic regression in separate models, which included ethnicity/race, age, education, and childhood poverty as covariates. Analyses were not adjusted for any of the above-listed CMassociated sequelae because these may be on the potential causal pathway.
Moderation analyses, adjusting for the same covariates listed above, were conducted to evaluate whether CM-associated sequelae modified risk for adverse birth outcomes. For these analyses an interaction term was created between the cumulative CM exposure score, as it best captures variation in CM severity, and each of the 4 potential CM-related sequelae which was included in multiple regression models. Because the moderating influence of the different sequelae may increase if several sequelae are present, we also created a total sequelae score by assigning one point for each CM-associated sequela reported by the participant (range 0-4) and used it in the moderation analyses.
For those CM-associated sequelae that were significantly associated with both CM and SGA or CM and PTB, their mediating effect was tested using PROC CAUSALMED in SAS, which uses a counterfactual approach for mediation analysis 51,52 . Each mediator was considered separately and models were adjusted for ethnicity/race, childhood poverty, age, education, and the other CM-associated sequelae not evaluated as mediators. As CM exposure was evaluated in three different ways (any, cumulative, and domains), a Bonferroni correction was used to account for multiple testing and a p-value of 0.0167 (0.05/3) was used to determine statistical significance. Analyses were conducted using SAS Software Version 9.4. Ethics approval and consent to participate. Institutional Review Boards at each site approved the protocol (Overall IRB Northwestern University #STU00039484 approved 2/13/14), and all participants provided consent.

Results
CM and delivery data were available for 657 of the 744 MOMS study participants. The sociodemographic characteristics of these 657 participants were not significantly different from those in the full cohort ( Table 1). The distribution of key study variables and the crude associations between key variables and birth outcomes are summarized in Table 2. 14.6% of study participants reported one form of moderate-to-severe CM, 6.1% reported 2, 6.1% reported 3, 3.0% reported 4, and 1.7% reported all 5 forms of CM. In total, 31.5% reported at least one form of moderate-to-severe CM, which is consistent with the estimated prevalence of CM in the US 53 . The most commonly-reported forms of CM included sexual abuse (18.4%), emotional abuse (16.1%), and physical abuse (13.1%), followed by emotional neglect (9.7%) and physical neglect (8.2%).
Participants with any history of CM were more likely to have experienced childhood poverty and to have lower educational attainment and income compared to those without exposure to CM. These participants were also more likely to be overweight or obese, use recreational drugs during pregnancy, smoke cigarettes, report symptoms of depression, and develop hypertensive disorders of pregnancy (Table 2).
Sixty three participants (9.6%) delivered SGA infants. Participants who delivered an SGA infant were more than twice as likely to have developed hypertensive disorders during the index pregnancy, and were more likely to have experienced childhood poverty compared to those who did not deliver SGA (Table 2).
Fifty-four participants (8.2%) had a PTB. PTB was more prevalent among participants who were from lowincome households (< $15 k/year), identified as Non-Latine Black, had a diagnosis of HDP, and were overweight/ obese ( Table 2).
Maternal childhood maltreatment and offspring preterm birth. Having any maternal CM exposure was not associated with PTB in the current study. The PTB rate among participants with CM was 9.7% compared to 7.6% for participants without CM, and this difference was not statistically significant in either crude  Table 2). Each of the 5 domains of CM were significantly associated with increased odds of SGA delivery (see Table 2) and cumulative CM exposure was also significantly associated with SGA in both crude and adjusted analyses (OR 1.36, 95% CI 1.14, 1.62/aOR 1.27, 95% CI 1.05, 1.53) ( Table 3), even after accounting for multiple testing. In general, as the number of CM domains increased, the proportion of SGA births increased ( Fig. 1). When models were adjusted for race/ethnicity, age at delivery, education, and childhood poverty, each subsequent domain of moderate or severe CM endorsed by a participant was associated with a 27% increased likelihood of delivering an SGA infant, compared to subjects who did not report any severe/moderate CM (aOR 1.27, 95% CI 1.05, 1.53) ( Table 3). Table 4 none of the 4 CM-associated sequelae or the sequelae sum score moderated the association between CM and SGA, nor that between CM and PTB.

CM-associated sequelae as potential mediators of the association between CM and SGA.
Of the CM-associated sequelae, only hypertensive disorders was significantly associated with CM and SGA (Table 2) and was therefore tested as a potential mediator. As shown in Table 5, hypertensive disorders did not significantly mediate the relationship between CM and SGA.

Discussion
The present results replicate the previously-observed association between CM and risk of an SGA birth, and indicate that CM is associated with a 61% overall increased risk of SGA. CM may be a significant clinical risk factor for SGA birth, as the association was comparable to that of established risk factors 54  www.nature.com/scientificreports/ this study was slightly lower than the national average for that same period (9.6%), and the 9.6% SGA rate was similar to that of the US overall (10%) 58 . In this healthy sample that was not enriched for CM exposure, elevated risk for SGA did not appear to stem from depression, substance use, obesity, or hypertension associated with CM exposure. We were unable to replicate a significant association between CM and PTB which has been demonstrated, albeit inconsistently, in prior studies 13,59 . Unlike those studies, the MOMS cohort was inherently at lower risk for PTB due to the exclusion of individuals on progesterone treatment for prior preterm birth, so this may in part explain our findings. These findings extend the literature in a number of ways. First, we replicated CM and growth associations previously detected in small relatively disadvantaged cohorts 15,16 in a large US sampleStevens-Simon and McAnarney 1994 included 127 low-income Black adolescent individuals from Rochester NY 16 , and Gavin et al. 2011 examined a cohort of 136 deliveries from Seattle, WA with greater than 50% childhood and adult poverty rates 15 .
Second, while SGA and PTB incidence overlap, our results suggest that CM may impact fetal growth and length of gestation differentially. Only three prior studies have investigated fetal growth as an outcome, and these   www.nature.com/scientificreports/ studies examined low birthweight but not growth percentile [14][15][16] , which does not identify growth abnormalities per se, since it combines consequences of PTB with those of growth restriction. By differentiating growth effects from length of gestation in our outcomes variables, our analysis indicates that these outcomes may result from different pathways. Third, our results suggest a CM and SGA association independent of CM-related sequelae. In concordance with previous studies we observed that depression, substance use, obesity, and gestational hypertensive disorders were significantly more common among individuals exposed to any form of CM, compared to those with no CM exposure. Of these, only gestational hypertensive disorders was more common among those delivering SGA infants, but gestational hypertensive disorders did not mediate nor moderate the association between CM and SGA. Substance use has previously been reported to mediate associations between CM and both birthweight and PTB 15,16 . However, these study populations had considerably higher rates of prenatal tobacco, alcohol, and recreational drug use (29% in Simon and McAnarney 1994 and 21% in Gavin et al. 2011) than our cohort 15,16 . The prevalence of substance use in our cohort may have been insufficient to exhibit a significant association with either birthweight or length of gestation.
Our results add to a growing literature evidencing the importance of early life exposures on intergenerational health outcomes. Adverse experiences during this sensitive developmental period may induce permanent or long-term alterations in neural, endocrine and immune systems which may be adaptive from the perspective of increased chances of survival in the short term. Over time these alterations may, however, have deleterious or unfavorable effects on the health and well-being of an individual 72 and future generations 11 . CM-associated variation in maternal-placental-fetal (MPF) stress biology may underlie the observed association between CM and SGA 60 . CM has been associated with variation in cortisol concentrations, steeper increases in placental CRH and increased systemic inflammation during pregnancy (reviewed in Moog et al. 11 ), biological alterations that Table 4. Results for models of depression, gestational hypertensive disorders, substance abuse, and obesity as moderators of the association between childhood maltreatment (sum score of moderate/severe domains) and adverse outcomes. a Models adjusted for ethnicity/race, maternal age, education, childhood poverty. b BMI = Body Mass Index weight/height 2 .  www.nature.com/scientificreports/ may impair fetal growth by leading to maternal vascular underperfusion and/or chronic placental inflammatory lesions that reduce placental function critical for fetal growth 61 . Reduced intrauterine growth in association with maternal CM may contribute to higher cross-disease morbidity in offspring 62 , starting with the effects on fetal growth and development 1-3,14-16 and extending to subsequent infant, child (behavioral problems, autism), and even adult health and disease risk 63,64 . While prevalence rates for the other sequelae were in concordance with reported national rates 65-67 , our sample size (and consequent number of SGA deliveries in this sample) was too small to identify mediators of the association between CM and this complex phenotype with multifactorial etiology. We suggest that future studies take advantage of cohorts enriched for SGA births to replicate the higher prevalence of CM among individuals delivering an SGA offspring and investigate the role of these sequelae as potential mediators.
In the context of a healthy population not enriched for CM, our findings point to the potential clinical significance of screening for CM to detect risk for growth restriction and other adverse outcomes. The CTQ has been validated in multiple iterations that may be clinically useful 68 , and by identifying individuals exposed to CM, may provide opportunity for therapeutic intervention during pregnancy. Moreover, these findings indicate the need to identify effective treatments, such as those that may reduce stress and normalize stress physiology, which as a consequence may reduce SGA risk among those with CM histories.
This study has some limitations. First, CM was based on retrospective self-report using the CTQ. Retrospective reports in adulthood of childhood adverse experiences are subject to problems like non-awareness, nondisclosure, simple forgetting, and reporting biases due to mood states. However, while the recall of subtle details of early traumatic experiences may depend upon personal interpretations and more prone to recall bias, major experiences (i.e., abuse) are better recalled. In general, false negatives are more frequent than false positives 69 . This downward bias should, therefore, result in an observed estimate of the association between CM and SGA during pregnancy 70 that is more conservative than the one that actually exists. Additionally, the CTQ demonstrates high internal consistency, good test-retest reliability, and convergence with other instruments 65 . The CTQ does not collect the date of exposure to CM, so we were unable to examine whether the timing of abuse is relevant to the association of CM and obstetric outcomes. We were also unable to evaluate PTSD symptomology or diagnosis, another highly prevalent consequence of CM, which in future studies should be examined as a potential mediator or moderator. Smoking and substance use during pregnancy was also based on self report, which is typically underreported 66 and could have obscured this pathway in our study. Finally, we are not able to comment on resilience factors that may have contributed to these findings.
Our findings extend those from previous studies of childhood adversity and birth outcomes 14,34,67 to show that CM is associated with increased risk for SGA birth independent of substance use, depression, obesity, and hypertensive disorders during pregnancy. These findings support the notion that adversity in childhood sets a trajectory for long-term and intergenerational patterns of health, and may underlie persistent population disparities 71 .

Data availability
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.