Female infants are more susceptible to the effects of maternal antenatal depression; findings from the Pelotas (Brazil) Birth Cohort Study

Highlights • Antenatal depression of moderate severity did not raise the risk of adverse birth outcomes.• Severe levels of antenatal depression were associated with lower APGAR scores and being SGA.• In this group of severe antenatal depression we found an increased risk for lower weight scores, being SGA and in the lower 10th centile for length for female newborns only.• To our knowledge this is the first paper to systematically examine the potential effects of increasing severity levels of antenatal depression on birth outcomes.

https://doi.org/10.1016/j.jad.2020.02.025 Received 10 May 2019; Received in revised form 8 January 2020; Accepted 8 February 2020 Birth outcomes that may be influenced by maternal antenatal depression, such a LBW, being small for gestational age (SGA) and PTB are known predictors of poorer long-term health; increasing the risk for delayed neurodevelopment, poor linear growth (Murray et al., 2015), obesity, diabetes, hypertension and cardiovascular disease as well as entailing loss of human capital (Lawn et al., 2014). In addition, the incidence of LBW and PTB is higher in LMIC where about 1 in 5 infants are born SGA (Lawn et al., 2014).
There are many proposed mechanisms through which maternal antenatal depression may impact on foetal and child development. These include changes to the Hypothalamic Pituitary Adrenal (HPA) axis, the role of the placenta, and the contribution of genes in the heritability of stress responses and epigenetic mechanisms (Herba et al., 2016).
Research so far has predominantly relied on data from high income countries (HIC) with data from LMIC considerably under-represented. Data, such as the ones discussed in this paper, contribute significantly to the literature, not only because of the varying levels of quality of obstetric care and mental health services, but also due to the varying levels of risk factors known to raise the risk of adverse birth outcomes. Furthermore, data from settings with different confounding structures may also assist in understanding the potential causal relationship and mechanisms in play.
A number of factors may act as effect modifiers of the association between maternal antenatal depression and child birth outcomes and have not been addressed systematically in the literature. These include the persistence and severity of depression (moderate vs. severe), socioeconomic factors (for example maternal education) (Coll et al., 2017;Patton et al., 2015;Sadovsky et al., 2016;van Heyningen et al., 2016) and child sex. These have already been shown to act as moderators of the association between postnatal depression and a wide range of child development domains (Netsi et al., 2018).
There is now accumulating evidence from both HIC and LMIC (including from this cohort, Coll et al., 2017) that the likelihood of experiencing depression during pregnancy is higher in older women, women with lower levels of education (Coll et al., 2017), women reporting food insecurity and women experiencing threatening life events (van Heyningen et al., 2016). In addition, previous history of depression is one of the strongest predictors of experiencing depression during pregnancy (Coll et al., 2017;Patton et al., 2015;van Heyningen et al., 2016). These risk factors have also been associated with suboptimal birth outcomes such as LBW and intrauterine growth restriction (Sadovsky et al., 2016). Taken together this evidence demonstrates that socioeconomic and demographic inequalities impact on both perinatal mental health and birth outcomes both of which are known to affect long-term child development.
The aim of this study was to examine (i) the association between maternal antenatal depression and birth outcomes in a new prospective, longitudinal birth cohort, the 2015 Pelotas (Brazil) cohort, (ii) the role of socioeconomic and demographic moderators and (iii) the association between more severe antenatal depressive symptoms and birth outcomes.
We hypothesised that: maternal antenatal depression would be associated with LBW and PTB (<37 weeks), with the strength of the association being greater for more severe depression; and the magnitude of the association would differ across levels of socioeconomic and demographic moderators with stronger associations in the groups with lower education and income, and older maternal age groups. We also hypothesised an increased susceptibility to the effects of maternal antenatal depression for female infants.

Recruitment of participants & design
The 2015 Pelotas (Brazil) Birth Cohort study is a population-based prospective cohort of all children born from mothers living in the urban area of the city of Pelotas, Southern Brazil. (Hallal et al., 2017). All women with confirmed pregnancy and estimated delivery date in the year 2015 were considered eligible and invited to take part in the study. Eligible pregnant women were recruited from antenatal care health facilities including laboratories, ultrasound clinics, public primary health care units, university clinics and private clinics. Face-to-face interviews were conducted mid pregnancy (16-22 weeks of gestation) to collect information on several maternal health pregnancy-related aspects. Between January 1 st and December 31 st 2015, maternity hospitals in the city were visited daily and all eligible live births identified. In total, 4333 eligible births were identified in the five hospitals of the city (99% of performed deliveries). Newborns were examined and mothers interviewed during their stay. Fifty-one women refused to take part in the study and 7 were lost to follow-up. Refusal rate and losses were equivalent to 1.3% of eligible births (n = 4275). Details of the 2015 Pelotas Birth Cohort Study are available elsewhere (Hallal et al., 2017).
During the antenatal component of the study, 73.8% (n = 3199) of the mothers of children enroled in the birth cohort were identified, which forms the study population for this analysis. Analysis here is restricted to singleton births for whom maternal depression was assessed during pregnancy (n = 3046). This is due to evidence that outcomes differ substantially between singletons and twin births and because we use the INTERGROWTH 21st standards which have been calculated on singleton pregnancies. For more details on the procedures of recruitment see (Hallal et al., 2017).
The study was conducted in accordance with the ethical principles of the Declaration of Helsinki. The study was approved by the Research Ethics Committee from the Federal University of Pelotas/Superior School of Physical Education (522.064). All participants gave written informed consent.

Maternal depression
Antenatal depressive symptoms were assessed using the Edinburgh Postnatal Depression Scale (EPDS). An EPDS score of ≥13 has been previously used to indicate probable major depression in this sample (Coll et al., 2017). A score of ≥17 on the EPDS is used to indicate more severe symptoms (Murray and Cox, 1990;Netsi et al., 2018;Putnam et al., 2017). The questionnaire was collected mid pregnancy (mean 21.5 weeks). We used the EPDS Portuguese translation, which has been previously validated in the 2004 Birth Cohort Study amongst postnatal Southern Brazilian women (Santos et al., 2007).

Child outcomes
The following child outcomes were collected at birth: birth weight (g), length (cm), head circumference (cm), APGAR scores at 1 st and 5 th minutes, and gestational age at birth. Gestational age was estimated through ultrasonography or best obstetric estimate. We used the Intergrowth-21 st Project standards (Villar et al., 2014) to calculate standardised and centile scores for length, weight and head circumference by sex and gestational age (using the online platform https://intergrowth21.tghn.org/) for singleton newborns. Additionally, we included measures on mode of delivery (caesarean/vaginal), whether the infant was showing any health problems (no/yes), and whether the infant was admitted to the intensive care unit (no/yes).

Confounding variables
We controlled for the following perinatal variables: maternal age (<20 years, 20-24 years, ≥35 years old), maternal schooling (0-4 years, 5-8, 9-11, 12 or more years), family income (in quintiles), skin colour (white/other), whether mother lives with partner (yes/no), maternal parity (1 st , 2 nd , 3 rd ), self-reported history of depression before pregnancy (no/yes), maternal pre-pregnancy body mass index (18.5-24.9, 25.0-29.9, ≥30), smoking during pregnancy (no/yes) and alcohol consumption during pregnancy (no/yes), gestational hypertension (no/yes), gestational diabetes (no/yes), eclampsia (no/yes), and child sex (male/female). The information regarding obstetric risk (gestational hypertension, diabetes, eclampsia and high pre-pregnancy BMI) was selected a priori based on previous research (do Carmo Leal et al., 2016). A proportion of women in the sample (20%) participated in the PAMELA trial (Domingues et al., 2015) which examined whether women randomised to exercise during pregnancy had improved maternal and child outcomes including maternal postnatal depression. The trial did not show any effects on the primary outcomes, we nevertheless include exercise as potential confounder to account for women's participation in this study (achieving the recommended 150 min of exercise per week throughout pregnancy/ not achieving the recommended amount of exercise). The birth outcomes examined in this paper have also been associated with prenatal exposure to selective serotonin reuptake inhibitors (SSRIs) (Casper et al., 2003) we therefore considered adding the use of anti-depressants during pregnancy as a potential confounder. In this sample only 2.5% of the women reported anti-depressants use during pregnancy. Due to the low variability in this potential confounder we decided not to include it in the analysis. We also considered including drug use. Women were asked once during pregnancy (58.2% of the sample was asked between 16-24 weeks and 41.8% after 24 weeks). Due to the low reporting of drug use in this sample (1.1%) we did not include this variable as a potential confounder.

Statistical analysis
We used descriptive statistics to summarize the outcome variables and the distribution of depression in the sample, using the EPDS as a categorical score (cut-off of ≥13 to indicate depressive symptoms of at least moderate severity). We also describe birth outcomes by sex. First, we used logistic, and linear regressions to examine the association between depression and birth outcomes. Second, we examined possible interactions between depression and socioeconomic and demographic indicators (maternal education, depressive history, maternal age, multiparity, and family income) and child sex, using separate models for each potential effect modifier. We then examined the association between antenatal depression and birth outcomes including confounders which were all entered in step 2 and any observed interactions were entered in step 3. We present both the crude and adjusted models, including the interaction terms. Where the interaction term was found to be significant we also investigated the association between antenatal depression and birth outcome separately by sex. If the interaction terms were not found to be significant in the unadjusted models we did not include them in adjusted models.
We repeated the above steps to explore associations between severe depression (EPDS≥17) and child birth outcomes. For this analysis, we explored three groups using antenatal depression as an ordinal variable: women scoring <13 on the EPDS (reference group), women scoring 13-16 (moderate depression) and women scoring ≥17 on the EPDS (severe depression), using logistic regression, allowing us to investigate (i) the potential effects of progressively higher levels of depression severity and (ii) potential dose-response effects.

Results
We examined differences in key socio-demographic characteristics between the whole Pelotas birth cohort and participants with antenatal EPDS data available (n = 3046, supplementary Table 1). Women for whom no EPDS data was available were more likely to be younger (<20 years of age), to not be cohabiting with a partner, to have skin colour other than white, to have an unplanned pregnancy, to be smoking during pregnancy and have consumed alcohol during pregnancy, to have fewer years of education, to be in the lower income quintile and to be multipara. There were no differences by child sex.
The prevalence of maternal antenatal depression (defined as EPDS ≥13) in this sample was 16.3% and has been previously reported elsewhere (Coll et al., 2017). Participant demographic characteristics show differences by depression status in almost all variables (Table 1). Younger women (<20 years) were more likely to have scored above threshold on the depression scale. Women not cohabiting with their partner, of ethnic background other than white, women who hadn't planned the (current) pregnancy, who smoked and consumed alcohol during pregnancy, with fewer years of schooling (0-4 years) were also more likely to score above threshold on the depression scale. They were also more likely to be in the lowest income quintile, to have a history of depression and to be multipara. There were no differences in child sex.

Child birth outcomes
3.1.1. Moderate depression 3.1.1.1. Main effects. Table 2 presents the description of child birth outcomes for the whole sample and by maternal antenatal depression status. Table 3 presents birth outcomes by child sex. Table 4 presents the unadjusted and adjusted analyses. For neonates of women who scored above the depression threshold during pregnancy there was an increased risk of being shorter in length (standardised and centile scores, coefficient −0.12 (−0.22,−0.010) and −3.37(−6.21,−0.53)), being of LBW (standardised and centile scores −0.11 (−0.021,−0.0.2) and −3.20 (−5.93,−0.47) and SGA (OR 1.48 [1.04, 2.09]). We found a difference in type of delivery with women scoring above threshold on the depression scale more likely to have caesarean delivery compared to vaginal (OR 0.74 [0.60,0.90]). We also present the models considering confounders (as outlined in the methods section). Inclusion of confounders in the models attenuated the effects reported above (Table 4).

Interaction effects.
We did not find any interactions between maternal antenatal depression with sociodemographic factors. We found two interactions with sex; for SGA and length (<10th centile), with an increased odds ratio for females to be born SGA if their mothers had been depressed during pregnancy (OR 3.04 [1.44,6.42]) and more likely to be in the lower 10th centile for length ( We repeated the analysis to explore progressively higher levels of depression severity and present results for women scoring between 13-16 and women scoring ≥17 on the EPDS compared to women scoring <13 on the EPDS (reference group), Table 5. The analysis revealed that for neonates of women scoring 13-16 there was no increased risk for adverse birth outcomes; and increased likelihood to deliver with caeseran compared to vaginally (OR 0.74 [0.60,0.90], Table 5). Children of women scoring 17 and above were at increased risk for a number of outcomes. We found an increased risk for neonates of women scoring 17 or above to have a lower APGAR score (

Interaction effects.
We did not find interactions with sociodemographic factors. We report interactions of child sex by depression grouping for birth weight (standardised and centile scores), being SGA, LGA and being in the lower 10 th centile for length. These interactions revealed an increased risk for females compared to males. Females whose mothers had severe depression (≥17) were consistently at increased risk for these outcomes. More specifically, females whose mothers had severe depression were at increased risk of lower birth weight ( . Following the addition of confounders we included the interaction terms depression X sex for the outcomes outlined above. Similarly to the unadjusted estimates, further analysis indicated an increased risk for females but not for males. Females whose mothers scored ≥17 on the EPDS were at increased risk for having lower standardised weight scores ( The addition of confounders led to only small decreases in the strength of the estimates between crude and adjusted analysis.

Main findings
In this study, we examined the association between maternal antenatal depression and child birth outcomes in a large prospective cohort in Brazil. We examined both main effects and explored interactions with sociodemographic characteristics and child's sex. We did not find any main effects in the association of the relationship between antenatal depression and child outcomes. We did find an interaction between maternal depression and child sex; with females at increased risk for being SGA if their mothers had scored above threshold during pregnancy. Unlike previous reports we did not find an association with PTB or LBW.
We also explored this association using a higher cut-off on the EPDS to investigate higher thresholds of depression severity based on evidence that severity is associated with higher risks for child outcomes both in the antenatal and postnatal periods (Jarde et al., 2016;Netsi et al., 2018). We did find evidence of increased risk with more severe depression but overall these associations attenuated after the inclusion of confounders. Following the addition of confounders in the model we only found an increased risk of lower APGAR scores (<7) and being SGA. Depression scores between 13-16 did not raise the risk for (negative) child outcomes when examining main effects, but raised the risk when examining interactions for being in the lower 10 th centile for length and for being SGA for females only. An increased risk was consistently evident for children of women scoring 17 and above on the EPDS. For female newborns whose mothers experienced more severe depression during pregnancy, we found an increased risk of having lower weight scores, being SGA and in the lower 10 th centile for length. We did not find any evidence that sociodemographic factors moderated the association between antenatal depression and birth outcomes.

Interpretation
Our findings suggest that female infants exposed to severe depression during pregnancy are at increased risk of being SGA in the lower 10 th centile for length and for having lower weight scores. Given the evidence, that interventions in LMIC which increase birth weight and linear growth are likely to result in substantial gains in height and schooling particularly when they occur in the first 2 years of life, early identification is particularly important (Adair et al., 2013;Horta et al., 2017). Our findings highlight such a group that can be identified before birth, through screening for depressive symptoms during pregnancy, and could be prioritised for interventions early in life.
There are a number of possible explanations for our findings. First, Brazil is one the countries with the most rapid reductions in neonatal mortality (Lawn et al., 2014) and has undergone substantial changes in perinatal services in the last decades, leading to a high standard of Table 4 Crude and adjusted analysis for the association between maternal antenatal depression of at least moderate severity and birth outcomes for outcomes with. Pelotas, 2015 Birth Cohort.   Table 5 Crude and adjusted analysis for the association between maternal antenatal depression and birth outcomes by depression severity (moderate/severe) in the Pelotas, 2015 Birth Cohort.

Linear Regression
Unadjusted coeff (95%CI) Adjusted coeff (95%CI) Length for age Z-score perinatal service and improved health for all mothers. This is also reflected in some of the outcomes where we found a low variability in scores (e.g. for APGAR; at 5 min neonates in this cohort scored a mean 9.5 (standard deviation 0.8) indicating excellent health). A recent meta-analysis reported a trend for exposure to more severe depression to be associated with higher risks for PTB and LBW (Jarde et al., 2016). The nature of the depression may be important as more severe depression may have stronger effects on foetal programming but may also place women at higher risk for postnatal depression which has been shown to impact on child development via disruptions in parenting (Stein et al., 2012). More specifically, depression that is both persistent and severe seems to substantially raise the risk of offspring adverse outcomes both in early childhood and late adolescence (Netsi et al., 2018).
A number of studies have now examined potential sex specific effects in the association between antenatal depression and child outcomes and the evidence suggests there may be different programming effects for males and females. Our finding that girls are more susceptible to the effects of antenatal depression symptoms is in line with animal studies suggesting increased reactivity of the HPA axis and cardiovascular reactivity in female offspring as well as human studies suggesting increased depressive/anxious symptoms and increased HPA activity. One potential mechanism relates to increased glucocorticoids altering the development of the foetal HPA axis, and there is evidence that there may be sex specific effects from recent research where increased maternal cortisol during pregnancy was associated with higher levels of negative emotionality and lower birth weight in girls (Braithwaite et al., 2017). These effects may have long term consequences; analysis from the 2004 Pelotas birth cohort indicated birth outcomes including being SGA, smaller head circumference and lower ponderal index to be associated with attention difficulties at age four for females only (Murray et al., 2015). Furthermore a recent analysis of a large UK-based cohort, reported depression during pregnancy to be associated with increased risks of offspring depression at 18 for females only (Quarini et al., 2016), suggesting these sex dependant differences at birth may affect a number of important developmental domains. It is important to note though that in this literature, sex specific effects have not consistently identified females at being of increased risk with some reports highlighting effects on males but not females (Li et al., 2010) and some reporting no clear sex differences (O'Connor et al., 2002).
We found a high prevalence of depression in this sample (16%) and previous analysis of this data indicates similar patterns of association with known risk factors previously reported in the literature (Coll et al., 2017). However, we found a low prevalence of suicidal ideation in this community sample (1% indicated the thought of harming themselves had occurred to them quite often, or sometimes (2.4%)). Higher levels of suicidal ideation could indicate clinically more severe depression and other community samples in LMIC indicate higher level of suicidal ideation (van Heyningen et al., 2016). Future work should investigate whether this is an important consideration in relation to birth outcomes.
Our analysis included a number of known confounders for the association between antenatal depression and birth outcomes. The attenuation of association between antenatal depression and birth outcomes following the addition of these confounders further highlights that depression is likely to occur within a constellation of other risk factors. This study specifically addressed the association between antenatal depression and birth outcomes but further in-depth investigation of the relative contribution of these confounding factors and the potential underlying mechanisms through which they may confer risk requires further investigation. Most of the confounding variables included in this analysis are routinely collected at point of care and could be used in addition to screening measures for depression to identify women and infants at risk. This also highlights the need for integrated approaches to perinatal care including perinatal mental health which address the constellation of risk factors that may be present.

Strengths and limitations
This is one of the few studies examining the association between antenatal depression and child birth outcomes in LMIC and one of only a handful of studies which attempts to disentangle the effects of mild or moderate depression compared to more severe levels of depression (Campbell et al., 1995;Netsi et al., 2018;Wickramaratne et al., 2011). Future studies should further investigate the role of the severity of the depression in relation to other important outcomes known to be associated with antenatal depression such as, but not limited to, brain development (Posner et al., 2016), emotional and behavioural outcomes. Furthermore research needs to address the mechanisms which underpin the association beweeen antenatal depression and birth outcomes, particularly in LMIC where antenatal depression is more prevalent as are many risk factors known to be associated with it (Herba et al., 2016). This study has a number of strengths. We utilised a large community sample of women and their children using the Intergrowth 21 st standards to calculate standardised and centile scores. The standards take into account sex and gestational age and are drawn from an international sample of both HIC and LMIC populations of women characterised as being of low obstetric risk. Data on depression was collected prospectively using a well-established screening measure for depression which has been validated in this population. Overall studies have not adequately considered the potential confounding role of pregnancy complications or being of high obstetric risk in the association between antenatal depression and child birth outcomes (Bindt et al., 2013;Ferri et al., 2007;Rahman et al., 2007). In this sample, we include data on gestational hypertension, diabetes and eclampsia as indicators of obstetric risk.
The study has a number of limitations. Information on whether the delivery was initiated by the healthcare provider was not available and this is potentially relevant because it may have affected our variable of gestational age. Due to the high rate of caesarean sections in this sample specifically, delivery initiated by the health care provider is very likely without necessarily indicating the foetus may have been at risk or distress. Additionally, information was not available on the overall care participants received during pregnancy such as the frequency of visits or quality of care; we are therefore not able to examine whether there were differences amongst participants in this variable and whether this was associated with birth outcomes. An important limitation of this cohort is that information on antenatal depression was available only during the second trimester. Additional points of measurement in pregnancy would have allowed us to potentially disentangle timing effects as well as whether some of the increased risks seen for infants of women with more severe symptoms of depression was related only to severity of depression or also potentially chronicity of depression.

Conclusion
In this large prospective cohort, we find evidence of an elevated risk for being SGA but did not find evidence of increased risk of other variables reported elsewhere such as PTB or length at birth for children of women experiencing depression during pregnancy. We found important sex differences primarily for children of women with more severe depression adding to the literature that females may be more susceptible to the effects of antenatal depression and stress. Maternal antenatal depression remains a significant public health concern and should be addressed not only because it is distressing to the individual and their family but it also raises the risk for future episodes of depression which are known to affect children's development, through a different set of mechanisms to antenatal depression, including parenting and the mother-child relationship.

Role of the funder/sponsor
The funders had no role in the design and conduct of the study; E. Netsi, et al. Journal of Affective Disorders 267 (2020) 315-324 collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Author statement
Dr Netsi had full access to all of the data in the study and takes joint responsibility with Dr Coll for the integrity of the data and the accuracy of the data analysis. Dr Netsi contributed to the study concept and design, the acquisition, analysis and interpretation of data, drafting of the manuscript, critically revised the manuscript for important intellectual content, conducted the statistical analysis.
Dr Coll had full access to all of the data in the study and takes joint responsibility with Dr Netsi for the integrity of the data and the accuracy of the data analysis. Dr Coll contributed to the study concept and design, the acquisition, analysis and interpretation of data, drafting of the manuscript, critically revised the manuscript for important intellectual content, conducted the statistical analysis.
Professor Stein contributed to the study concept and design, the acquisition, analysis and interpretation of data, drafting of the manuscript and critically revised the manuscript for important intellectual content.
Dr Silveira contributed to the study concept and design, the acquisition, analysis and interpretation of data, drafting of the manuscript and critically revised the manuscript for important intellectual content.
Dr Bertoldi contributed to the study concept and design, the acquisition, analysis and interpretation of data, drafting of the manuscript and critically revised the manuscript for important intellectual content.
Dr Bassani contributed to the study concept and design, the acquisition, analysis and interpretation of data, drafting of the manuscript and critically revised the manuscript for important intellectual content.
Dr Wehrmeister contributed to the study concept and design, the acquisition, analysis and interpretation of data, drafting of the manuscript and critically revised the manuscript for important intellectual content.
Dr Domingues contributed to the study concept and design, the acquisition, analysis and interpretation of data, drafting of the manuscript and critically revised the manuscript for important intellectual content.
All authors contributed to and approved the final manuscript.

Declaration of Competing Interest
The authors declare no conflict of interest.