Maternal stress modifies the effect of exposure to lead during pregnancy and 24-month old children's neurodevelopment

Abstract

Background

Lead and psychosocial stress disrupt similar but not completely overlapping mechanisms. Exposure during the prenatal period to each of these insults singularly has been found to alter normal neurodevelopment; however, longitudinal associations with stress modifying the effect of lead have not been sufficiently analyzed in epidemiologic studies.

Objective

To evaluate prenatal stress as an effect modifier of gestational lead neurotoxicity.

Methods

We used a structural equations modeling approach with a trivariate response to evaluate cognitive, language and motor scores of the Bayley Scales of Infant Development-III in 24 month-old children (n=360). Maternal blood lead levels were measured at the 2^nd and 3^rd trimester and psychosocial stress during pregnancy was assessed using a negative life events (NLE) scale derived from the CRYSIS questionnaire.

Results

3^rd trimester lead (mean 3.9 ± 3.0 SD μg/dL) and stress (median=3 NLE) were negatively associated with Bayley III scores. Using the model's results we generated profiles for 0, 2, 4 and 6 NLE across lead levels (up to 10 μg/dL) and observed a dose-response for the developmental scores when lead levels were below 2 μg/dL. Each NLE curve had a different shape across increasing lead levels. Higher stress (NLE=6) resulted in lower cognitive scores for both sexes, in lower language scores in girls but not boys. In the absence of stress we saw a negative association with lead for all scores, however for language and motor scores, higher stress seemed to mask this association.

Conclusions

Our work examined and confirmed prenatal stress exposure as a modifier of the well-known neurotoxic effects of prenatal lead. It adds to the existing evidence pointing at the importance of studying the co-exposure of chemical and non-chemical exposures, specifically of considering the emotional environment of children at early developmental stages of life.

1. Background

Prenatal exposures to lead and to psychosocial stress have been individually associated with infant neurodevelopment (Bellinger 2013; Bock et al. 2015; Hu et al. 2006; Laplante et al. 2008; O'Donnell et al. 2009; Talge et al. 2007). Both of these exposures are likely to co-occur, higher lead exposures being more common in communities with lower socioeconomic status, which in turn can imply chronic stress exposure (Cory-Slechta et al. 2004). In Mexico, lead exposure remains common at a population level, particularly through the use of lead-glazed pottery (Caravanos et al. 2014; Diaz-Ruiz et al. 2016) and stress exposure has also been documented (Medina-Mora Icaza et al. 2004). Studies combining these exposures have been performed mainly in animals and research in humans remain scarce (Schneider and Cory-Slechta 2016).

Lead can interfere with normal brain development by disrupting several critical processes such as neuron migration, synaptogenesis, myelination and selective synaptic pruning (Faustman et al. 2000). It can also affect the release of neurotransmitters and stimulation of neurotransmission (Bressler et al. 1999; Bressler and Goldstein 1991). Even at low levels, lead can induce neuronal cell death (Dribben et al. 2011). All of these processes are inter-related and their regulation by environmental signaling mold the underlying architecture of the brain. The hippocampus and nucleus accumbens are particularly susceptible to the toxic effects of lead, consequently memory and behavior/executive functions are affected (Nihei et al. 2001; Verina et al. 2007). Studies have found a detrimental effect of prenatal lead exposure with neurodevelopment (Boucher et al. 2014; Hu et al. 2006; Lin et al. 2013) that can persist in adulthood (Mazumdar et al. 2012).

Maternal psychosocial stress during pregnancy has been associated with different types of adverse neurodevelopmental outcomes in the child, like reduced cognitive abilities and language development (Bock et al. 2015; Laplante et al. 2008; Tarabulsy et al. 2014). Animal studies using the same spatial memory tests showed similar results when rats were exposed to lead as when they were exposed to stress; lead has been found to disrupt cognition through its effects on the mesocorticolimbic dopamine pathway and stress hormones act on this same pathway via the hypothalamic-pituitary-adrenal (HPA) axis (Cory-Slechta et al. 2008). The HPA axis responds to stress by producing cortisol, which is necessary for normal brain development in infants (Ng 2000). However, chronic stress can result in an altered production of cortisol (McEwen 2004), which can cross the blood-brain-barrier and bind to the highly abundant glucocorticoid receptors in the hippocampus, known to be involved in learning, memory, and emotional processing (Preston and Eichenbaum 2013). A study examining cortisol in amniotic fluid and prenatal maternal psychosocial stress found that higher cortisol levels predicted lower cognitive Bayley scale scores, this was independent of the stress measures and these were also negatively associated child's cognition (Bergman et al. 2010).

Epidemiological studies have looked at maternal self-esteem and it's simultaneous exposure with lead in humans (Surkan et al. 2008; Xu et al. 2015), however to our knowledge none have studied prenatal exposure to lead and the potential modifying effect of stress on neurodevelopment, particularly in a prospective design.

In this study we examine if the exposure to stress in the prenatal period modifies the effect of lead on cognitive, language and motor development in 24 month-old infants.

2. Methods

2.1 Study population

Study participants are part of the ongoing PROGRESS (Programming Research in Obesity, GRowth, Environment and Social Stressors) birth cohort in Mexico City. Women attending a prenatal consult were approached in 4 clinics belonging to the Mexican Social Security System (IMSS) between July 2007 and February 2011; if in their first trimester, they were invited to participate in the study, and answered a screening questionnaire. Inclusion criteria considered: being <20 weeks pregnant, ≥18 years old (Mexican legal voting age), being heart or kidney disease free, having access to a telephone, planning to reside in Mexico City for the next 3 years, no use of steroids (including glucocorticoids) or anti-epilepsy drugs, and not consuming alcohol on a daily basis (Braun et al. 2014). PROGRESS study protocols were approved by the institutional review boards of the Icahn School of Medicine at Mount Sinai, Harvard T. H. Chan School of Public Health, the National Institute of Public Health Mexico, the Mexican Social Security System, and the National Institute of Perinatology, Mexico. At each visit the study protocol was explained to women, who provided informed consent before any procedure was carried out.

Of the 760 mother-infant pairs ensued, 541 children had a Bayley III assessment at 24 months of age. For this analysis, we excluded 2 children who were very premature (≤32 weeks of gestation) and 2 that where both premature and had extremely low birth weight (≤1500 g).

2.2 Lead measurements

Prenatal lead exposure was assessed using maternal venous blood samples drawn during the second (2T BPb, between the 16^th and 20^th pregnancy week) and third (3T BPb, between the 30^th and 34^th pregnancy week) trimesters. We used royal blue trace metal Vacutainer (Becton-Dickinson and Company, Franklin Lakes, New Jersey) tubes containing EDTA to collect the samples that were kept at 4°C until they were shipped to the Trace Metals Laboratory at the Harvard T.H Chan School of Public Health (HSPH), where they were stored at −20°C until analyzed. Lead concentrations were measured using a dynamic reaction cell inductively-coupled plasma mass spectrometer (Elan 6100; PerkinElmer, Norwalk, CT). Five replicate measurements of each sample were taken and averaged. The recovery of the analysis quality control standards and spike samples was 90%-110%, and the limit of detection for the procedure was 0.02 μg/dl.

2.3 Psychosocial Stress assessment

We measured maternal stress in the third trimester using the Crisis in Family Systems-Revised (CRISYS-R). The Spanish version of the survey had a test-retest reliability of .86 over 2 weeks among Spanish speaking subjects (Berry et al. 2006). Berry and collaborators also verified the construct validity of the survey finding strong associations between reported life stressors and covariates in their study population (greater depressive symptomatology, poorer physical and mental health function and lower household income). This questionnaire has been previously used in studies examining stress (Peters et al. 2012; Suglia et al. 2010; Tse et al. 2012). The 64-item questionnaire assesses the occurrence of a series of life events during the previous 6 months. Women answer if an event occurred during the past semester and if so, whether the experience was positive, negative, or neutral. Life events experienced are assessed across 11 domains: financial, legal, career, relationships, community and home violence, medical problems, other home issues, discrimination/prejudice, and difficulty with authority. We generated an additive score that ranged from 0 to 11 which summarized the number of domains with at least one negative life events (NLE). Higher scores indicate greater psychosocial stress: a larger number of domains with negative life events reported indicates a greater diversity of stress experienced by a woman, which is more likely to overwhelm her coping resources (Schreier et al. 2015).

2.4 Infant cognitive, language and motor development

Cognitive and language development scales were assessed when children were 24 months old using the Bayley Scales of Infant Development III (BSID-III). The scales are based on US norms (with a range from 40-160 points) and could be different for our study population (Cromwell et al. 2014), therefore we used the raw scores obtained by children and standardized them to the expected mean of 100 and SD of 15. Study personnel blinded to the infant's lead or maternal stress level administered the tests using a standard protocol.

Information on covariates included: infant sex, birth weight, gestational age (based on last menstrual period and by a standardized physical examination (Capurro test) to determine gestational age at birth) and maternal age, which were collected at time of delivery, maternal IQ (using the Wechsler Adult Intelligence Scale, Spanish version) and Home Observation for Measurement of the Environment (HOME) score (assessed at 24 months postpartum). All psychometric tests, including the Bayley and Crysis were applied by psychologists with training specifically for these questionnaires.

2.5 Statistical analyses

We began by performing an exploratory data analysis of exposure and outcome variables as well as covariates. BPb levels showed a right skewed distribution therefore were log transformed to reduce the asymmetry. For our final analysis 360 out of 536 mother-infant pairs had complete information all covariates. Most of the missing data was from the HOME score however we decided to include it as it has shown to be a good predictor for both scores; besides, a comparison analysis between pairs with and without HOME showed no significant differences for the rest of the variables.

In order to account for the longitudinal structure between the exposures and the outcome, as well as for the complex association pattern among covariates and outcomes, we used the structural equation model (SEM) illustrated in Figure 1. Cognitive, language and motor scores were jointly considered as a trivariate outcome, to account for the correlation between them. Assuming that some of the 2T BPb will be carried over into the 3T BPb (Lamadrid-Figueroa et al. 2006) we decided to include both measurements in the proposed model. Based on biological plausibility and previous knowledge on predictors of neurodevelopment, we adjusted for child's sex and included 3 equations for 3T BPb, birth weight and HOME score as intermediate variables of 2t PBb, maternal age at delivery and gestational age and maternal IQ respectively. Using a nonparametric lowess fit, we detected a quadratic association between stress and cognitive scores but not for language or motor scores, hence we included a quadratic term of stress when modeling cognitive score. We also detected a quadratic relationship between birth weight and the outcomes. Based on results of the modeling process, interaction terms between BPb's and stress, HOME score and sex, and stress and sex were also included, the two latter terms leading only to language scores.

Figure 1.

Structural Equation Model for the association between prenatal exposure lead and stress and 24 month-old children's Bayley III Scores.

We conducted all statistical analyses using STATA 13 (StataCorp. 2013. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP) and Stata 14 (StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP).

3. Results

Women in our study had a mean 2T BPb of 3.6 ± 2.6 μg/dL and of 3.9± 3.1 μg/dL in the 3T BPb with a Pearson correlation between the log of these measures of 0.78. The median stress score was 3 (domains with at least one NLE) and 6% reported experiencing no NLE. Characteristics were similar between study participants (n=536) and nonparticipants (n=224) and differences were not statistically significant (Table 1). Correlations between covariates are reported in the supplemental material.

Table 1.

Characteristics of study participants and nonparticipants (mean ± SD or %).

Characteristic	Participantsa	Nonparticipantsb
Sex (% boys)	53	51.6
Birth weight (kg)	3.1 ± 0.4	3.0 ± 0.5
Gestational age (weeks)	38.4 ± 1.5	38.3 ± 1.9
2nd trimester blood lead (μg/dL)	3.7 ± 2.6	3.6 ± 2.6
3rd trimester blood lead (μg/dL)	3.9 ± 3.0	3.6 ± 2.7
Stress Score ( median, % NLE=0)	3, 6	3, 9
HOME score	31.7 ± 5.4	31.5 ± 5.2
Maternal IQ	85.5 ± 12.5	83.8 ± 12.8
Maternal education (years)	12.9 ± 2.7	11.7 ± 2.9
Maternal age at delivery (years)	26.9 ± 5.5	27.2 ± 5.3

^an=536 except for: 2T BPb n=534, 3T BPb n=484, Stress n=502, HOME n=415, Mom IQ n=525.;

^bn=224 except for: 3T BPb n=85, Stress n=194, HOME n=82, Mom IQ n=206.

Overall the estimated coefficients for the SEM total effects (Table 2) reflected the expected negative direction for the association between lead as well as stress and developmental scores; they also indicated that the most relevant BPb exposure was the third trimester and that girls had higher scores than boys.

Table 2.

Total Effects for SEM of the association between prenatal exposure lead and stress and 24 month old children's neurodevelopment (n=360).

	Cognitive		Language
	β	p value	β	p value	β	p value
log 3T BPb	−6.60	0.06	−6.00	0.09	−11.014	0.00
log 2T BPb	0.76	0.73	0.97	0.66	1.97	0.39
Stress	1.72	0.13	−1.37	0.09	−1.04	0.19
Stress2	−0.23	0.04
2T BPb * Stress	−1.14	0.23	−1.24	0.19	−2.51	0.01
3T BPb *Stress	1.02	0.27	1.66	0.07	2.93	0.00
Child Sex (girl)	4.19	0.01	−8.45	0.26	4.89	0.00
Birth weight	40.95	0.04	37.16	0.07	39.93	0.06
Birth weight2	−5.93	0.07	−5.10	0.11	−5.48	0.10
Gestational Age	4.13	0.05	3.75	0.07	4.03	0.67
Mom IQ	0.13	0.00	0.08	0.01	0.03	0.12
Mom age at delivery	0.24	0.19	0.22	0.21	0.23	0.21
HOME	0.82	0.00	0.52	0.00	0.22	0.11
HOME * Sex			0.35	0.11
Stress*Sex			1.02	0.08

The SEM included more than one equation as well as interaction terms therefore, in order to simplify its interpretation and to illustrate the results better we used profile graphs for 3T BPb by child's sex. We considered 4 different profiles corresponding to specific stress levels: NLE=0, NLE=2, NLE=4 and NLE=6. For each case, predictions of expected scores were obtained using the estimated coefficients and evaluating the equations at the mean of the other covariates, generating a profile curve. We graphed these different profiles separately for the three developmental scores (Figure 2) with respect to third trimester BPb for each of these four profiles.

Figure 2.

Adjusted means curves for 4 stress levels based on the SEM (n=360)

3.1 Cognitive score results

Comparing the profiles for girls, when 3T BPb is less than 2μg/dL there is a decrease of approximately 9 cognitive points on average for the 0 NLE profile compared to that of 6 NLE. The mean difference would be of about 1 point and 5 points less for the profiles corresponding to an NLE score of 2 and 4 respectively. The mean cognitive score would decrease for each stress profile with increasing 3T BPb, however the steepest decline would be for the profile with 0 NLE, coming close to the profile with 6 NLE=6 (which always had the lowest cognitive scores and remained almost flat regardless of the lead level). In the case of boys, the interpretation for the profiles is the same as for girls as they have the same shape, except all are shifted downwards on the y axis by 2 and a half points with the same significance levels for each profile (Figure 2).

3.2 Language score results

These results were different for girls and boys as is shown in Figure 2. For girls, when 3T BPb was less than 2 μg/dL the same pattern was observed for the 4 stress profiles as that seen for cognitive scores: the highest language score would be expected for the 0 NLE profile, and about a 5, 15 and 25 points less difference for profiles with a NLE score of 2, 4 and 6 respectively. For boys the differences were not as large, with a 3 point decrease in the language score as the NLE profile increases. As the lead level increases, the shape of the different profile curves is different depending on sex. For girls we see a descending pattern for the 0 and 2 NLE profiles and although the former almost reaches the same language scores as the latter, their profiles don't overlap like they do in the case of the profiles for boys with a 0 and 2 NLE scores. On average the profile for 4 NLE had very similar results for both sexes, remaining flat at any lead level. The profile for 6 NLE showed an increasing pattern for both sexes. However, the number of women with a NLE score of 6 was small for girls (n= 16) and boys (n= 14) and these numbers were even smaller for lead levels ≥5 μg/dL (girls n = 10, boys n=9).

3.3 Motor score results

At lead levels <2μg/dL the adjusted means curves showed the same dose-response seen for cognitive and language scores. As the lead level was increased the curves showed a different pattern. In the absence of stress the lowest motor scores were observed, however as stress increased motor scores increased as lead increased, particularly for the higher stress curve. We did not see a difference between boys and girls.

Lastly, we evaluated the final model's goodness of fit which resulted in an overall R² = 0.89. When comparing the null model (which assumes there is no association with the observed variables) to the proposed we obtained a statistic of 681.75 which compares with a χ² of 26 d.f. These tests therefore indicating that the model had a reasonable adjustment.

4. Discussion

Our results are in line with previous research that has found inverse associations with Bayley scores for increased prenatal lead (Bellinger 2013; Braun et al. 2012; Hu et al. 2006; Lin et al. 2013) and prenatal stress (Bergman et al. 2010; Bock et al. 2015; Buss et al. 2012; Laplante et al. 2008; Talge et al. 2007; Tarabulsy et al. 2014) separately. In our models adjusting simultaneously for both, we saw statistically significant negative coefficients for the main effect of lead on both cognitive and language scores, as well as for the coefficients of the main effect of the quadratic term of stress for cognitive scores and for the stress term for language scores. Furthermore our model suggests that prenatal exposure to stress modifies the association between prenatal lead exposure and neurodevelopment. Although evidence of interaction is modest, the graphs in figure 2 are not parallel which is consistent with interaction. All converge to lower Bayley cognitive scores as blood Pb and stress levels increase but do so at different rates and with different starting points. The lower starting points for each incresased stress profile suggest that while there are some modest interactions, the effects of stress and lead are mostly additive.

We observed a different response to lead for each of the stress profiles and cognitive scores. At lead levels below 2μg/dL a dose-response relationship can be observed where a girl born to women with a profile of no reported stress (NLE=0) had higher cognitive scores compared to the profiles for NLE scores of 2, 4 and 6.

In the absence of stress we can see the steepest decrease in cognitive scores as lead level is increased. In comparison, the high stress profile has an almost flat slope, indicating that irrespectively of the lead exposure, a child born to a woman experiencing higher stress would have the lowest cognitive scores. This is likely a “floor” effect. That is, in the presence of very high stress, its neurotoxicity is maximized. In this situation low level lead poisoning does not add additional toxicity. This finding is also consistent with lead and stress being additive in their toxicity rather than multiplicative (i.e. synergistic). It may suggest that high stress and low level lead poisoning work via similar additive mechanisms that have a floor.

Mild prenatal stress (“eustress”) exposure has been previously documented to result in higher cognitive scores, indicating that not all stress exposure is detrimental, which is in line with our results when comparing the profiles of 2 and 4 NLE to that of no stress (Glover 2015). Nonetheless these profiles showed a decrease in cognitive scores as lead levels increased.

Boy's cognitive scores were lower than girls which is seen by a downward shift of the curves and the model's coefficient for sex. The shape for the profiles was identical for girls and boys, illustrating no further effect modification by sex. However, this was not the same for language, for which the profiles had different shapes depending on the child's sex.

For girls, the stress profiles for language scores showed a dose-response pattern throughout the lead levels. Therefore, having no prenatal exposure to stress in girls had the highest language scores (also comparing to boys), followed by the 2 NLE profile, the 4 NLE profile and lastly the 6 NLE profile (which was the lowest throughout the different lead levels and also comparing to the lowest language scores for boys). Additionally, the gap for language scores between the profile for no stress and highest stress is greater than the one seen for cognitive scores (almost 25 language points compared to 10 cognitive points at lead levels below 2 μg/dL). For boys this dose-response pattern is also observed for low lead levels, the gap between the language scores for the different NLE profiles is not as wide as for girls.

This sex-dependent effect of the lead-stress association has been mostly investigated by animal studies where females show worse outcomes than males and the suggested mechanism involves the role of oestrogen in the mediation of pre- and post-natal effects of this association (Cory-Slechta et al. 2012).

For both sexes the profile pattern changes completely with increasing lead levels. The profiles for no stress and 2 NLE showed a decreasing shape, the profile for 4 NLE remained almost flat and the curve for 6 NLE increased. In the case of boys the profile for no stress results in the lowest language scores as lead is increased, even lower than the girls profile for 4 NLE; by contrast, the highest stress profile (NLE=6) showed the highest language scores as lead increased. For girls there was no overlap with the lower stress profiles.

Motor results did not show further modification by sex however we saw increasing patterns for the scores with the higher stress profile. These higher motor scores are surprising and difficult to explain. But we note that motor function at age 2 years has not generally been associated with lead exposure (Mason et al. 2014). These results, as well as the increasing pattern for language scores in the higher stress profiles may be due to unmeasured confounding or it may be due to chance, and need to be interpreted with caution, by no means should this imply recommending higher lead levels for pregnant women experiencing higher stress levels.

In trying to explain the ascendant pattern for the higher stress and higher lead profile we hypothesized a scenario where high levels of these exposures in utero could have a programming effect whereby an overwhelming of the hypothalamic-pituitary-adrenal (HPA) axis and the hippocampus sets-off a compensatory mechanism and postnatal language development is enhanced, similar to the findings of the Dutch famine effects on later obesity (Roseboom et al. 2011). When examining the characteristics of the women included in the analyses we noticed that those reporting 6 NLE had a median IQ of 3 points (for boys) and 8 points (for girls) higher than those reporting no NLE, perhaps resulting in women with a higher IQ considering and reporting more life events as negative (correlations between covariates are reported in the supplemental table 1). Engaged parenting could also be associated with a higher IQ and would likely be reflected in Bayley scale scores. Higher scores were expected with increased HOME scores, furthermore, we saw sex-differences showing that girls had higher language scores with higher HOME scores. Despite this, higher stress in girls resulted in the lowest language scores.

The use of structural equation models enabled us to adjust simultaneously for prenatal lead exposure during the second and third trimester recognizing the order in time. This is an important advantage since it allowed us to identify the third trimester as a more relevant window of exposure. Previous studies on sensitive periods to lead exposure and neurodevelopment have pointed at the first trimester (Hu et al. 2006) or the third trimester (Schnaas et al. 2006) as the most relevant. Other studies have used cord blood (Boucher et al. 2014) (which in our study is highly correlated to the third trimester BPb (Spearman 0.81 p< 0.05)) and found significant results.

With respect to stress in pregnancy, studies have shown that earlier prenatal exposure may be most relevant. Lazinski et al. comment that many of these studies have focused on different types of exposure to stress (acute vs. chronic) and these have been associated with a variety of birth outcomes, not only neurodevelopment (Lazinski et al. 2008). LaPlante et al. found that acute stress resulting from an ice storm during early gestation was associated with lower mental development index Bayley scores that persist into early childhood (Laplante et al. 2008). A recent meta-analysis has found the evidence inconclusive about the role of timing in stress measurement, and could not point to any particular time during gestation as the most sensitive one with respect to child neurodevelopment (Tarabulsy et al. 2014).

Animal studies on simultaneous prenatal lead and stress exposures have focused on the gestational period as a whole and not on specific time points (Cory-Slechta et al. 2008). We assessed maternal stress once during the third trimester and the questionnaire captures information on the previous six months, covering the second trimester. Therefore we accounted for stress experienced throughout pregnancy which, in combination with lead support late pregnancy as more relevant for exposure.

To our knowledge, only one epidemiologic study by Surkan et al has evaluated a psychosocial exposure using maternal self-esteem as a modifier of the detrimental effects of lead on development indexes of the Bayley scales (Surkan et al. 2008). In their study, infants of women with higher self-esteem had better scores compared to those with mothers of low self-esteem. We cannot directly compare our results to that study since having a low NLE score might not correspond to having high self-esteem. Additionally our study used the Bayley III scores which have separate scales for cognitive and language scores, whereas Surkan et al used a previous Bayley examination with different scales. Despite these differences, our study complements their findings, is in line with a non-chemical exposure modifying the association between lead and neurodevelopment, and highlights the importance of considering the co-exposure to the emotional environment of the developing child, including the gestational period.

Our results may be most important not at the individual level as the impact on each child's Bayley score by either lead or stress is small. However, it may have implications at a population level shifting the distribution of scores and reflecting a larger numbers of infants with lower neurodevelopmental scores (Claus Henn et al. 2012).

4.1 Limitations

Our study sample is very homogenous in terms of race and SES and our results may not be generalizable to all populations, however we do not believe that there are any specific differences in the type of stress or lead that these Mexican women and their infants may be exposed to that could not be similar to those of other populations. Our final analytic sample was reduced mainly because of the absence of the HOME evaluation, mostly due to the logistic difficulties this evaluation represents (a psychologist visited the participant's house). When comparing participants with and without HOME we found no statistically significant differences, but our results might have lost statistical power. We were unable to control for lead in early pregnancy (1st trimester), since our participants were enrolled from the 2^nd trimester. However, finding a specific window of susceptibility was not an aim of this study. Information on smoking during pregnancy and breastfeeding was collected, however none of the study participants reported smoking and breastfeeding had very little variability 95% of the study participants reported breastfeeding (ever/never) by 12 months. These are potential confounders that could not be accounted for in our analyses.

4.2 Strengths

Our study is among the first to prospectively assess lead and stress in pregnancy and determine whether stress modifies the neurotoxic effect of lead on subsequent child neurodevelopment (our participants are followed from the gestational stage onwards). We were also able to include the HOME evaluation in our analyses which captures the environment where the child is developing much better than any other measure.

5. Conclusions

Our results add to the existing evidence pointing at the importance of studying non-chemical exposures as potential effect modifiers of chemical exposures. This study sets a precedent to consider psychosocial stress, an exposure that is common to all human populations, as an effect modifier of associations of known neurotoxic substances that impact child development. These processes share common mechanistic pathways and might be acting synergistically on the same biologic pathways in producing neurotoxicity.

Highlights.

• Lead and stress were negatively associated with Bayley III neurodevelopment scores.

• For lead levels <2 μg/dL there was a dose-response for stress with both scores.

• Prenatal stress exposure modified the neurotoxic effects of prenatal lead.

• Higher prenatal stress resulted in lower cognitive scores.

• Further modification by sex was seen for language scores.