Maternal prenatal stress during the first trimester and infant birthweight in Soweto, South Africa

INTRODUCTION

Being born low birthweight (LBW, BW<2500g at birth) increases one’s risk for a wide range of poor physical and mental health outcomes that can persist across the lifecourse, and can even have effects that impact health in the next generation (Barker 1998; Kuzawa & Sweet 2009; Levitt et al. 2000). It is well-known that neonates born small or preterm face greater risk of perinatal morbidity and mortality (Bernstein et al. 2000). Additionally, growing evidence suggests that LBW status also shapes health outcome across development into adulthood, affecting risk of a wide range of diseases such as diabetes (Harder et al. 2007; Whincup et al. 2008), hypertension (Mu et al. 2012), and neuroendocrinological (Jones et al. 2006; Wu□st et al. 2005) and inflammatory dysregulation (McDade et al. 2009). In turn, some of the long-term sequelae of being born small can increase the risk of giving birth to lighter babies in the next generation (Kuzawa & Sweet 2009; Thayer et al. 2012). These major public burdens caused by LBW have long impacted low- and middle-income countries (LMICs) at disproportionate levels (UNICEF-WHO 2019), which foreshadows a concerning public health burden caused by the increased risk of diseases associated with both LBW status.

A growing body of research has reported that gestating mothers who experience psychosocial stress are at heightened risk of giving birth to neonates with lower birthweights (Ae-Ngibise et al. 2019; Patil et al. 2020; Therrien et al. 2020; Zhu et al. 2010). Experiences of maternal prenatal stress are understood to contribute to LBW as a result of reduced gestational duration and restricted fetal growth, and often times, a complex combination of both gestational responses. Previous studies have shown that pregnant mothers with greater psychosocial stress tend to give birth to children with shorter gestational ages (Bussières et al. 2015; Coussons-Read 2012), who are preterm (Rosa et al. 2019; Wadhwa et al. 2011; Zhu et al. 2010), who are small-for-gestational age (SGA) (Class et al. 2011; Khashan et al. 2014), and in some cases, infants who are growth restricted (Berkowitz et al. 2003; Glynn et al. 2001; Resnik 2002). Additionally, previous studies have documented a wide variety of stressors that also predict LBW, which include pregnancy-related stress (Hobel et al. 2008), poverty (Lee & Lim 2010), racial discrimination (Collins et al. 2004), and maternal depression (Grote et al. 2010) and anxiety (Ding et al. 2014). Conversely, some researchers have found no evidence for impacts of stress on birthweight (Engel et al. 2005; Ramchandani et al. 2010).

While the mechanisms underlying the impacts of maternal cortisol on fetal development are only partially understood (Gitau et al. 2001), psychosocial stress exposure during earlier windows of gestation may lead to decreases in birthweight due to neuroendocrine-mediated fetal growth restriction and earlier parturition (D’Anna-Hernandez et al. 2012; Glynn et al. 2001). Three potential mechanisms have been suggested: maternal cortisol exposure through the placenta, downregulation of cortisol-shielding enzymatic mechanisms, and reduced stress reactivity across pregnancy. First, stress-initiated elevations in circulating glucocorticoids (GCs) may lead to greater levels of cortisol to move through the placenta and reach the fetus (O’Donnell et al. 2009). Greater maternal cortisol levels during early periods of pregnancy have been associated with being SGA (Goedhart et al. 2010) and LBW (Bolten et al. 2011; D’Anna-Hernandez et al. 2012; Goedhart et al. 2010). Contrary to the negative feedback loop between adrenal cortisol exposure and hypothalamic corticotropin-releasing hormone (CRH) production, higher cortisol exposure increases CRH production in the placenta, which subsequently can upregulate the production of pituitary adrenocorticotropic hormone (ACTH) and in turn, cortisol in both the mother and the fetus (Majzoub & Karalis 1999). Growing evidence also suggests that greater prenatal stress exposure may result in maternal immune dysregulation during pregnancy, which has been associated with adverse gestational development and birth outcomes, including restricted fetal growth, preterm birth, and LBW (Beijers et al. 2014; Nazzari et al. 2019).

Second, early increases in maternal cortisol during gestation may also downregulate fetal buffering mechanisms against maternal cortisol. Placental levels of 11β-hydroxysteroid dehydrogenase (11β-HSD-2) act as a natural shield against maternal cortisol in utero by converting cortisol into biologically inactive cortisone. Maternal cortisol, however, is still able to penetrate the fetus in small amounts particularly when circulating maternal cortisol reaches high concentrations or when 11β-HSD-2 is downregulated (O’Donnell et al. 2012). For instance, recent evidence reports that high levels of maternal GCs can consequently downregulate 11β-HSD-2 production, thus allowing greater passage of GCs, including cortisol, into the womb (O’Donnell et al. 2012). Placental 11β-HSD-2 levels are naturally and relatively low during early gestation (Hobel et al. 2008), and during this time cortisol may more easily pass through the placenta during earlier stages of pregnancy to slow fetal growth rate and shorten GA.

Finally, the impacts of early prenatal stressors on fetal development and birthweight may be heightened as behavioral and biological stress sensitivity decreases among mothers over the course of pregnancy. The sensitivity of both major stress-regulatory systems, the HPA and the sympathetic-adrenal-medullary (SAM) axes, declines as pregnancy advances. For example, mothers in advanced stages of pregnancy show decreased psychological sensitivity (Glynn et al. 2001), HPA axis reactivity (Kammerer et al. 2002; Obel et al. 2005), and decreased blood pressure responses to stress (Nisell et al. 1985) relative to those in earlier stages.

While the mechanisms by which maternal GCs contribute to intrauterine growth restriction and preterm birth are unclear, high levels of stress (Glynn et al. 2001; Lederman et al. 2004) and cortisol during the first trimester have consistently been shown to predict LBW (Bolten et al. 2011; D’Anna-Hernandez et al. 2012; Goedhart et al. 2010). This stands in contrast to the potential impacts of second or third trimester stress on LBW risk, which may play less of a role compared to first trimester stress due to the relative dampening of stress reactivity among late-stage pregnant women (Glynn et al. 2001; Kammerer et al. 2002) as well as the higher levels of 11β-HSD-2 buffering as pregnancy progresses. Thus, during times of low 11β-HSD-2 buffering, normal psychological and physiological stress reactivity, and high levels of stress-initiated maternal cortisol and placental CRH secretions in utero, cortisol may more easily pass through the placenta during the first trimester to affect the developing fetus to possibly slow fetal growth rate and shorten gestational length. We hypothesize that greater exposure to maternal stress during earlier periods of gestation will correspond with lower birthweights.

Notably, studies of prenatal stress and birthweight predominantly draw from samples in North America and Europe (Bussières et al. 2015), which leads to a research bias that overemphasizes psychosocial experiences and social-environmental conditions affecting mothers among high-income, Western populations. Consequently, it also overlooks LMICs that face greater burdens of LBW, such as South Africa where rates of LBW are relatively high (14.2%) and surprisingly close to the rates observed in more poverty-stricken countries in the sub-Saharan region (UNICEF-WHO 2019). In South Africa, LBW is the second leading cause of death in children under 5 years of age (Bradshaw et al. 2003). These trends have been tied to a long history of institutionalized segregation, poverty, and social marginalization characteristic of the apartheid regime, which actively discriminated against people of color, including but not limited to individuals categorized as Black, and continues to affect these communities. To address these gaps, here we examine the timing-specific effects of maternal stress during pregnancy on birthweight in a sample of women living in Soweto, South Africa.

METHODS

RESULTS

This study included 657 mother and infants (Table 1). Participants included in the analytical sample were similar to those excluded (n = 173) with respect to first trimester prenatal stress, birthweight, fetal gender, maternal BMI, and GA (p > 0.05). Late pregnancy prenatal stress, primiparity, and maternal age were significantly different from those excluded from the sample (p < 0.05). Mothers in the analytical sample exhibited slightly lower average levels of stress, reported marginally higher rates of primiparity, and were older by about one year. Fetal gender, maternal age, maternal BMI, primiparity, and GA were significantly associated with birthweight.

Table 3 presents the results of the OLS regression analyses of maternal and fetal factors that predict infant birthweight. The unadjusted model (Model l) predicting birthweight on first trimester prenatal stress, while insignificant, displays a negative relationship as expected. This relationship between prenatal stress and birthweight remains insignificant after adjusting for fetal gender, maternal age, and maternal BMI (Model 2 & 3). The association between prenatal stress approaches significance (β = −12.4, p = 0.079, 95% CI [-26.3, 1.44]) after controlling for GA, which strongly predicts birthweight and accounts for a large portion of the variation (46%). Finally, prenatal stress remains inversely related to and a modest predictor of birthweight after adjusting for primiparity. A one-point increase in prenatal stress scores corresponded with a - 12.7 g decrease in birthweight (β = −12.7, p = 0.071, 95% CI [-26.4, 1.10]). Gender, maternal BMI, GA, and primiparity significantly relate to birthweight and maternal age serves as a modest predictor of birthweight. After adding maternal smoking to the model (Model 7), the effect of prenatal stress strengthens but becomes insignificant. The interaction between prenatal stress and gender was not significant (p = 0.752, 95% CI [-32.1, 23.2]) suggesting that there were no gender differences in the effects of prenatal stress on birthweight in our sample. An identical logistic model predicting LBW was also analyzed and found that the effect of prenatal stress on LBW was non-significant yet trended in the expected direction (p = 0.673, OR = 1.03, 95% CI [0.90, 1.17]).

The effect of first trimester maternal stress on GA was also examined to determine whether shorter gestational periods contributed to lower birthweights. Both unadjusted and fully adjusted models show that the effect of first trimester prenatal stress is not associated with GA (results not shown). Specifically, these analyses also reported a positive but non-significant relationship between maternal stress and GA: women with higher levels of maternal stress during their first trimester seemed to have longer gestations.

DISCUSSION

We report an inverse relationship between maternal stress severity and infant birthweight: first trimester maternal stress is associated with lower birthweights in our sample of women in Soweto, South Africa. There was also no evidence that the relationship between stress and birthweight was modified by maternal depression, anxiety, and perceived social support. These results point to the importance of the first trimester as an important developmental period for infant birth outcomes.

Previous studies have also reported the inverse relationship between higher first trimester prenatal stress and lower birthweights (Coussons-Read et al. 2012; Dancause et al. 2011; Ryu 2019; Vrijkotte et al. 2009; Zhu et al. 2010). For example, in a large sample of expectant mothers in Hefei, China, self-reported stressful life events that occurred during the first trimester predicted lower birthweights, which were unrelated to stress during the second or third trimester (Zhu et al. 2010). Additionally, in a study of the impacts of a severe ice storm, GA and birthweights were lower among women exposed to the ice storm during early and mid-pregnancy (Dancause et al. 2011). Conversely, a number of studies have also reported no evidence for the effects of first trimester stress on birthweight (Rondó et al. 2003). A wide range of severity and type of maternal stress exposure have been examined in past literature, (e.g. pregnancy-related distress, natural disasters, domestic violence, poverty, trauma, etc.) and the downstream biological pathways linking certain forms of maternal experience and LBW may vary, such as the case of maternal depression (Field et al. 2006). Nevertheless, the larger relationship between maternal prenatal stress that occurred during the first trimester and infant birthweight is consistently documented across study populations.

Our final model (Model 7) also uncovered the deleterious effect of maternal smoking during pregnancy on lowering infant birthweights. While the effect of maternal stress approached significance after controlling for key covariates, the prenatal stress variable became statistically insignificant after including the smoking variable into the model, which itself was not a significant predictor of birthweight. Maternal smoking during pregnancy is a well-known risk factor for LBW (Magee et al. 2004; Steyn et al. 2006) and is understood to lower birthweight through various pathways, including fetal hypoxia (Abel 1980), uterine vasoconstriction (Quigley et al. 1979), and oxidative stress-mediated alterations in placental function (Stone et al. 2014), all of which can contribute to fetal growth restriction. Further analyses indicate that prenatal stress and smoking are collinear with one another (results not shown), suggesting a possible bi-directional relationship between the two variables: smokers may exhibit higher stress levels than non-smokers, or highly stressed women may be more likely to smoke as a coping mechanism (Lobel et al. 2008). Because of this collinearity, our measure of prenatal stress alone may adequately capture the negative impacts of both stress and smoking on infant birthweight.

While GA was a significant and expected predictor of birthweight and accounted for 46% of the variation in our models, contrary to the larger literature (Glynn et al. 2008; Zhu et al. 2010), prenatal stress does not appear to lower birthweight as a result of shortened GA in this sample. Our data reported a positive but statistically insignificant relationship between the severity of stress exposure during early pregnancy and GA. It is only after controlling for gestational duration that the negative association between early prenatal stress and birthweight approaches significance (Model 6). This pattern suggests that fetal growth is likely slowed in relation to first trimester stress, but that this stress-linked delay in fetal growth is masked by non-significantly longer gestations among women in our sample. The lack of an association between stress and GA may be explained by prior global epidemiological patterns of birth outcomes by location, which show that in high-income countries, the majority of LBW babies are the result of shortened gestation rather than growth restriction (Mattison et al. 2001). In contrast, in LMICs, LBW persists irrespective of GA as well as an increased prevalence of SGA due to environmental and nutritional conditions such as greater infectious disease burden and undernutrition, which compromise maternal metabolism and the overall development of the fetus (Pike 2005).

While the prenatal stress scale was operationalized as a marker of maternal stress at the time of collection, the timing effects may reasonably reflect exposures experienced up to six months prior to data collection as the scale assessed retrospective maternal exposures within the past six months. Because the stress assessment was administered before 14 weeks of pregnancy, the measure may represent experiences up to three to six months before pregnancy. Pre-pregnancy maternal stress has also predicted LBW in past studies (Khashan et al. 2008). In the present study, there is also no evidence for the protective effects of social support as the main effects of social support and the interaction between stress and social support was not significant.

In summary, women who experienced psychosocial stress early in gestation gave birth to modestly lower birthweight babies in our sample of South African mothers and infants. Women who experienced greater levels of stress appeared to have non-significantly longer gestations, and the negative impact of maternal stress on infant birthweight only appeared after adjusting for this pattern. These results suggest that delays in fetal growth associated with first trimester maternal stress may contribute to lowering birthweights in our sample. Our findings emphasize the importance of better understanding the sources of stress experienced among women of reproductive age in Soweto in order create social interventions aimed at reducing maternal stress, especially early in gestation. Furthermore, this study informs larger public health campaigns, such as earlier antenatal visits, earlier ultrasound screenings, and better monitoring of maternal psychosocial states during pregnancy, designed to ameliorate and prevent the high morbidity of poor birth outcomes across South Africa and other LMICs.