Edited by: Jordan Marrocco, The Rockefeller University, United States
Reviewed by: Dustin Scheinost, Yale University, United States; María I. Cordero, Manchester Metropolitan University, United Kingdom
†These authors have contributed equally to this work
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
There is strong evidence that prenatal maternal stress (PNMS) or anxiety during pregnancy has profound and long-lasting adverse consequences on offspring development (see Van den Bergh et al.,
The amygdala is an important limbic structure involved in processing of biologically relevant stimuli, emotional learning and memory (Davis and Whalen,
Structural properties of the amygdala are also associated with psychopathology and subclinical psychopathological symptoms. For example, AGV is associated with psychiatric problems including depression (MacMillan et al.,
The associations between perinatal stressors and/or naturally occurring maternal cortisol levels during pregnancy and AGV can explain psychopathological outcomes in the offspring. Interestingly, rodent studies suggest that PNMS may influence offspring development and subsequent behavioral outcomes by altering amygdala maturation as measured by AGV (Salm et al.,
While these findings link perinatal maternal depression and prenatal maternal cortisol to structural changes in the amygdala, and subsequent psychopathological symptoms in exposed offspring, these studies have their limitations. The use of maternal psychopathology as the stressor is confounded by potential genetic transmission from mother to child. Moreover, while the Buss et al. (
This gap can be addressed in our PNMS study of children exposed to a natural disaster just prior to, and during gestation: Project Ice Storm. In January 1998, a series of ice storms resulted in electrical power failures for more than 3 million individuals for periods varying from 6 h, to more than five weeks, during the coldest months of the year. Mothers of the youth in our prospective, longitudinal study were at various stages of pregnancy and were quasi-randomly exposed to varying degrees of storm-related hardship. The ice storm serves as a “natural experiment,” affecting women regardless of their personal characteristics while allowing for the women's disaster experience to be divided into various components (i.e., objective hardship exposure, cognitive appraisal of the event, and subjective distress from the event). As well, because the ice storm had a sudden onset, the timing of the onset of maternal stress can be assessed with great accuracy. Among the findings from Project Ice Storm, we have shown that increased objective hardship predicts delayed cognitive and language development in childhood (King and Laplante,
The goals of the present study were to test whether varying aspects of the PNMS experience affect AGV in 11½ year-old youth using a prospective longitudinal design, and whether AGV mediates the association between varying aspects of the PNMS experience and child internalizing and externalizing problems. We also hypothesized that the timing of the prenatal maternal stressor would moderate those associations.
This study was carried out in accordance with the recommendations of The Code of Ethics of the World Medical Association, and approved by the Douglas Mental Health University Institute Research Ethics Board. All subjects gave written informed consent or written informed assent for participants under the age of 18 years, in accordance with the Declaration of Helsinki.
The initial Project Ice Storm cohort consisted of 178 children whose mothers were pregnant during the 1998 Quebec ice storm or became pregnant within 3 months of the ice storm and who responded to the questionnaire “reaction to the storm” sent on June 1, 1998. At 11½ years, 100 families were approached about participating in a standardized assessment and a structural brain magnetic resonance imaging (MRI) study. Of those, 90 children (male = 47; female = 43) underwent the cognitive and behavioral assessment, and 71 agreed to undergo the MRI protocol (male = 35; female = 36). Two children (both female) refused to undergo the scanning upon seeing the scanner. The scan of one girl was unusable because of excessive movement. An additional five participants (male = 4; female = 1) who were born preterm (before 37 weeks) with a low birth weight were excluded from the analyses as these birth outcomes have been shown to affect brain structure and integrity (Buss et al.,
There were 2 left-handed boys, and 3 left-handed girls. Thirty-three percent (21/63) of the participants were exposed to the storm prior to conception (boys: 11/31; girls: 10/32), 22.2% (14/63) during the first trimester (boys: 7/31; girls: 7/32), 23.8% (15/63) during the second trimester (boys: 8/31; girls: 7/32) and 20.6% (13/63) during the third trimester (boys: 5/31; girls: 8/32).
Objective hardship was assessed in June 1998 using Storm32 (Laplante et al.,
Mothers' subjective distress was assessed in June 1998, using a validated French version (Brunet et al.,
Mothers' cognitive appraisal of the ice storm crisis was assessed in the June 1998 questionnaire using the following question: “Overall, what were the consequences of the ice storm on you and your family?” Response options were on a 5-point Likert scale: “Very negative” (1), “Negative” (2), “Neutral” (3), “Positive” (4), and “Very positive” (5). This item was recoded into “negative” (0) or “neutral/positive” (1). We opted to treat cognitive appraisal as binary for two reasons. First, very few mothers reported that the crisis was either “very negative” (2) or “very positive” (1). Second, we wanted to isolate the “negative” appraisal from “neutral” and “positive” appraisals, because we believe negative cognitive appraisal is one of the components of the maternal stress experience that can influence child development.
Timing of
Behavioral problems were assessed using the mother-rated 113-item Child Behavior Checklist (CBCL), the gold-standard for behavioral research, which yields scores on several subscales that combine to create Internalizing and Externalizing scales. The standardized scores with a mean of 50 and standard deviation of 10 were used in the analyses (Achenbach and Ruffle,
Socioeconomic status (SES) was measured with the Hollingshead scale (Hollingshead,
The Life Experience Survey (LES) (Sarason et al.,
Mothers' concurrent psychological symptoms were assessed when the children were aged 11½ using the General Health Questionnaire (GHQ-28; Goldberg,
The number of obstetric complications (e.g., hypertension, preeclampsia, cold or flu) was determined by maternal recall when their children were 6 months of age using an adaptation of the checklist used by Jacobsen and Kinney (
Smoking and alcohol habits in pregnancy were assessed in the 6 month postpartum questionnaire in order to capture the entire pregnancy. Women were asked to indicate the number of cigarettes smoked per day, and the number of drinks consumed per week.
Child handedness (e.g., right- or left-handed) was determined at age 11½ by observing which hand the child used to complete a visual-motor integration task.
Anatomical MRI was performed at the
The amygdala was automatically segmented using the multiple automatically generated templates (MAGeT) Brain pipeline (Chakravarty et al.,
All analyses were carried out on boys and girls separately to account for the difference between male and female neurodevelopmental trajectories, and given the sex-specific effects reported in the literature (Buss et al.,
To test the effects of PNMS on normalized AGV, the data for right and left normalized AGV were subjected to hierarchical regression analyses. Child handedness was controlled for in all analyses. Given the evidence that timing of exposure to the stressor might influence the response to PNMS, timing of exposure to the storm
To assess the moderating effects of timing of exposure on the relationship between PNMS and right and left normalized AGV, interaction terms were added in the last step of the hierarchical regression, with either Objective hardship × Timing of exposure, Subjective distress × Timing of exposure or Cognitive appraisal × Timing of exposure as the interaction terms, entered in separate analyses.
In order to determine the confounding effect of postnatal environment on the effect of PNMS, concurrent SES at 11½ years postpartum, as well as postnatal maternal life events, were then added to the models if they were significantly correlated with AGV. Even though concurrent maternal psychological symptoms are postnatal, this variable was included in all models if significant, to address a potential responder bias issue.
Given the sample size, the equations were trimmed of any non-significant variables (
To test the effects of left or right AGV on internalizing and externalizing problems partial correlations were run, controlling for covariates that were significantly associated with the behavioral outcomes. As per the regression analyses, handedness and timing of exposure to the storm were always included as covariates, and analyses were run with and without postnatal measures to determine their confounding effect.
To test the extent to which changes in normalized AGV explained the effect of PNMS on behavior, simple mediation or moderated mediation analyses were used. Because the indirect effect is more likely to be significant if the two paths forming it are strongly associated, the mediation or moderated mediation was only tested when both paths of the indirect effect (path 1: PNMS or PNMS by timing interaction to normalized AGV from the regressions, and path 2: normalized AGV to behavior from the partial correlations) were significant or marginally significant. Covariates that were significantly correlated with either normalized AGV or the behavioral outcome were entered in the model, and then trimmed out if not-significant in the final model. Again, the models were run with and without postnatal measures.
For the moderation analyses, the SPSS PROCESS macro (Hayes and Preacher,
Descriptive statistics for outcome, predictor, and control variables are presented for boys and girls separately in
Descriptive statistics of variables, and results of
Right normalized AGV | 0.070 | 0.007 | 0.072 | 0.008 | 0.274 | 0.278 |
Left normalized AGV | 0.072 | 0.007 | 0.072 | 0.007 | 0.922 | 0.025 |
Objective stress storm32 | 11.839 | 4.810 | 11.031 | 4.816 | 0.508 | 0.168 |
Subjective stress IES-R | 12.032 | 13.824 | 8.650 | 8.870 | 0.472 | 0.182 |
Number of days of pregnancy when ice storm happened (timing of PNMS) | 78.450 | 102.849 | 94.590 | 103.788 | 0.538 | 0.156 |
Gestational age at birth (weeks) | 40.226 | 0.976 | 39.754 | 0.927 | 0.054 | 0.495 |
Maternal psychological symptoms GHQ-28 | 0.106 |
0.155 | 0.060 |
0.115 | 0.192 | 0.338 |
Socioeconomic status (SES) Hollingshead scale (perinatal) | 29.839 | 12.718 | 27.094 | 12.105 | 0.384 | 0.221 |
Socioeconomic status (SES) Hollingshead scale (concurrent) | 34.31 |
15.05 | 27.67 |
14.60 | 0.091 | 0.448 |
Number of cigarettes/day | 2.419 | 5.622 | 1.750 | 4.930 | 0.617 | 0.127 |
Number of glasses of alcohol/week | 0.047 | 0.184 | 0.127 | 0.421 | 0.330 | 0.246 |
Maternal life events (perinatal) | 6.516 | 4.098 | 5.468 | 2.961 | 0.248 | 0.294 |
Maternal life events (postnatal) | 1.29 |
0.89 | 0.9470 | 0.661 | 0.092 | 0.440 |
Obstetric complications | 4.81 | 3.02 | 3.970 | 2.192 | 0.211 | 0.319 |
Maternal age at birth | 29.000 | 5.320 | 30.196 | 4.311 | 0.330 | 0.247 |
Internalizing problems CBCL (T-scores) | 54.73 |
10.106 | 49.47 | 11.376 | 0.059 | 0.488 |
Externalizing problems CBCL (T-scores) | 48.03 |
8.732 | 45.81 | 8.716 | 0.320 | 0.255 |
As shown in
Pearson's and Spearman's rho correlation coefficients between all variables split by sex.
1- Right normalized AGV |
– | 0.796 |
0.055 | 0.045 | 0.144 | 0.256 | −0.354 |
0.243 | −0.408 |
−0.425 |
−0.204 | 0.072 | 0.118 | 0.410 |
−0.131 | – | 0.251 | 0.432 |
2- Left normalized AGV |
0.765 |
– | 0.089 | 0.135 | 0.077 | 0.203 | −0.322 |
0.434 |
−0.355 |
−0.445 |
−0.288 | 0.182 | 0.238 | 0.477 |
−0.266 | – | 0.330 |
0.457 |
3- Objective hardship |
0.087 | 0.146 | – | 0.373 |
−0.406 |
−0.108 | 0.198 | 0.409 |
−0.206 | −0.135 | 0.012 | −0.278 | 0.060 | 0.197 | −0.106 | – | 0.053 | 0.133 |
4- Subjective distress |
0.229 | 0.266 | 0.212 | – | −0.140 | −0.066 | 0.016 | 0.269 | 0.263 | 0.264 | 0.158 | −0.034 | 0.330 |
0.277 | 0.172 | – | 0.126 | 0.124 |
5- Cognitive appraisal | −0.207 | −0.216 | −0.303 |
−0.021 | – | 0.286 | −0.196 | 0.059 | 0.045 | 0.149 | 0.135 | 0.074 | 0.032 | 0.282 | −0.189 | – | 0.072 | 0.223 |
6- Timing of PNMS |
0.441 |
0.285 | −0.259 | 0.000 | −0.184 | – | −0.087 | 0.095 | −0.189 | −0.180 | −0.048 | −0.031 | −0.061 | 0.197 | −0.128 | – | 0.079 | −0.097 |
7- Gestational age at birth |
−0.008 | −0.155 | 0.037 | −0.243 | 0.265 | 0.074 | – | 0.033 | −0.240 | −0.326 |
0.116 | −0.066 | 0.048 | −0.290 | 0.159 | – | 0.153 | −0.154 |
8- Maternal psychological, GHQ |
−0.127 | −0.063 | 0.001 | 0.373 |
0.265 | −0.130 | −0.286 | – | −0.243 | −0.191 | 0.031 | 0.112 | 0.429 |
0.634 |
−0.109 | – | 0.529 |
0.390 |
9- Socioeconomic status (perinatal) |
0.310 |
0.209 | 0.303 |
0.379 |
−0.084 | 0.091 | 0.076 | 0.157 | – | 0.843 |
0.353 |
−0.003 | −0.272 | −0.146 | 0.079 | – | −0.261 | −0.129 |
10- Socioeconomic status (concurrent) |
0.280 | 0.193 | 0.412 |
0.604 |
−0.023 | −0.037 | 0.019 | 0.198 | 0.883 |
– | 0.324 |
−0.002 | −0.108 | 0.051 | 0.293 | – | −0.306 | −0.075 |
11- Smoking |
0.176 | 0.150 | 0.213 | 0.400 |
0.111 | −0.232 | −0.300 |
0.109 | 0.327 |
0.524 |
– | 0.032 | −0.135 | −0.130 | 0.413 |
– | −0.073 | −0.274 |
12- Alcohol usage |
−0.100 | −0.185 | 0.031 | −0.175 | −0.324 |
−0.148 | −0.324 |
−0.018 | −0.031 | −0.106 | 0.105 | – | 0.195 | 0.087 | 0.108 | – | 0.397 |
0.391 |
13- Maternal life events (perinatal) |
0.090 | 0.171 | −0.045 | 0.289 | −0.170 | 0.291 | −0.260 | 0.101 | 0.049 | 0.051 | 0.131 | −0.185 | – | 0.428 |
0.132 | – | 0.326 |
0.269 |
14-Maternal life events (postnatal) |
−0.171 | 0.074 | 0.264 | 0.322 |
−0.079 | −0.341 |
−0.004 | 0.393 |
0.007 | 0.156 | −0.107 | −0.239 | 0.224 | – | 0.055 | – | 0.394 |
0.389 |
15-Obstetric complications |
−0.036 | −0.166 | 0.028 | −0.011 | 0.136 | −0.093 | 0.183 | 0.158 | 0.008 | 0.143 | 0.236 | −0.020 | −0.269 | −0.049 | – | – | 0.050 | −0.154 |
16- Maternal age at birth |
−0.107 | −0.246 | −0.285 | 0.137 | 0.072 | −0.046 | −0.027 | −0.006 | −0.199 | −0.134 | 0.059 | 0.049 | −0.200 | 0.007 | 0.307 |
– | 0.207 | 0.109 |
17- Internalizing problems |
0.116 | 0.092 | −0.039 | 0.291 | −0.051 | −0.048 | −0.260 | 0.189 | 0.107 | 0.236 | −0.036 | −0.120 | 0.131 | 0.496 |
0.134 | 0.048 | – | 0.583 |
18- Externalizing problems |
0.403 |
0.378 |
0.070 | 0.396 |
−0.174 | −0.045 | −0.215 | 0.215 | 0.150 | 0.281 | 0.210 | −0.009 | 0.350 |
0.378 |
0.151 | −0.254 | 0.636 |
– |
In boys, the results of the hierarchical regressions for right normalized AGV are presented in
Summary of hierarchical regression analyses for (a) the normalized right amygdala volume in boys and for (b) the normalized right and (c) left amygdala volume in girls at 11 years of age.
Step 1 | 0.256 | 0.065 | 2.027 | |||||
Timing | 0.256 | 1.76E-05 | 1.20E-05 | |||||
Step 2 | 0.303 | 0.092 | 0.027 | 1.417 | 0.820 | |||
Timing | 0.294 | 2.02E-05 | 1.30E-05 | |||||
Preferred hand | 0.167 | 0.005 | 0.005 | |||||
Step 3 | 0.471 | 0.222 | 0.130 | 2.562 |
4.498 |
|||
Timing | 0.258 | 1.78E-05 | 1.20E-05 | |||||
Preferred hand | 0.126 | 0.004 | 0.005 | |||||
SES (perinatal) | −0.363 |
−2.02E-04 |
9.50E-05 | |||||
Step 4 | 0.474 | 0.225 | 0.003 | 1.882 | 0.099 | |||
Timing | 0.255 | 1.76E-05 | 1.20E-05 | |||||
Preferred hand | 0.145 | 0.004 | 0.005 | |||||
SES (perinatal) | −0.376 |
−2.10E-04 |
1.00E-04 | |||||
Objective stress | −0.060 | −8.78E-05 | 2.79E-04 | |||||
Step 5 | 0.523 | 0.273 | 0.049 | 1.880 | 1.677 | |||
Timing | 0.257 | 1.77E-05 | 1.20E-05 | |||||
Preferred hand | 0.173 | 0.005 | 0.005 | |||||
SES (perinatal) | −0.480 |
−2.68E-04 |
1.08E-04 | |||||
Objective stress | −0.193 | −2.84E-04 | 3.14E-04 | |||||
Subjective stress | 0.263 | 0.002 | 0.001 | |||||
Step 6 | 0.634 | 0.402 | 0.129 | 2.687 |
5.159 |
|||
Timing | 0.095 | −4.40E-05 | 2.90E-05 | |||||
Preferred hand | 0.096 | 0.003 | 0.005 | |||||
SES (perinatal) | −0.526 |
−2.93E-04 |
1.01E-04 | |||||
Objective stress | −0.049 | −7.27E-05 | 3.05E-04 | |||||
Subjective stress | 0.072 | −0.002 | 0.002 | |||||
Subjective stress |
0.443 |
2.57E-05 |
1.10E-05 | |||||
0.705 | 0.497 | 0.095 | 3.100 |
4.057 |
||||
Timing | −0.074 | −3.16E-05 | 2.88E-05 | |||||
Preferred hand | 0.235 | 0.007 | 0.005 | |||||
SES (perinatal) | −0.444 |
−2.47E-04 |
1.00E-04 | |||||
Objective stress | −0.141 | −2.09E-04 | 3.03E-04 | |||||
Subjective stress | 0.028 | −0.001 | 0.002 | |||||
Subjective stress |
0.329 | 1.88E-05 | 1.13E-05 | |||||
Postnatal life events | 0.368 |
0.003 |
0.001 | |||||
Step 1 | 0.441 | 0.194 | 7.227 |
|||||
Timing | 0.441 |
3.19E-05 |
1.20E-05 | |||||
Step 2 | 0.463 | 0.214 | 0.020 | 3.954 |
0.742 | |||
Timing | 0.443 |
3.21E-05 |
1.20E-05 | |||||
Preferred hand | 0.142 | 0.004 | 0.004 | |||||
Step 3 | 0.543 | 0.295 | 0.081 | 3.904 |
3.202 |
|||
Timing | 0.528 |
3.82E-05 |
1.20E-05 | |||||
Preferred hand | 0.259 | 0.007 | 0.004 | |||||
Objective stress | 0.318 |
4.96E-04 |
2.77E-04 | |||||
Step 1 | 0.285 | 0.081 | 2.648 | |||||
Timing | 0.285 | 1.82E-5 | 1.10E-05 | |||||
Step 2 | 0.358 | 0.128 | 0.047 | 2.137 | 1.575 | |||
Timing | 0.289 | 1.84E-5 | 1.10E-05 | |||||
Preferred hand | 0.218 | 0.005 | 0.004 | |||||
Step 3 | 0.491 | 0.241 | 0.113 | 2.969 |
4.168 |
|||
Timing | 0.389 |
2.48E-5 |
1.10E-05 | |||||
Preferred hand | 0.357 |
0.008 |
0.004 | |||||
Objective stress | 0.376 |
0.001 |
2.54E-04 |
We probed this interaction and found, as illustrated in
Moderation of subjective stress' effect on normalized right AGV by timing of exposure in boys. Low and high stress lines are represented at the 10 and 90th sample percentile of subjective distress (measured with the Impact of Events Scale-Revised) respectively, which are at a log-transformed level of 0 and 3.33 (0 and 26.94 in original scale), respectively. Following a significant interaction between subjective stress and timing, probing the interaction revealed that when mothers were exposed to the storm from day 157 (week 22.42) onwards, the greater their mothers' subjective stress, the larger their right normalized AGV; the region of significance is represented by the vertical line. When mothers were exposed to the storm before day 157, which includes the mothers exposed to the storm during preconception, there was no significant effect of subjective stress on right normalized AGV. Additionally, when subjective stress scores were equal to or greater than a log value of 2.64 (original subjective stress scale 13.01), there was a significant (
When adjusting the model for postnatal measures, concurrent SES was no longer significant, so it was trimmed out of the model. Postnatal life events became marginally significant (
For the left normalized AGV in boys, the hierarchical regression models revealed no significant main effects of PNMS measures, and no significant PNMS-by-timing interactions (data not shown).
When testing the second path of the indirect effect we found that right and left normalized AGV in boys were associated with externalizing, but not internalizing, problems when controlling for timing of exposure, handedness, maternal psychological functioning, and maternal alcohol usage during pregnancy (
Partial correlation coefficients between amygdala volume (AGV) and behavioral problems controlling for handedness, timing of exposure to the storm and additional covariates (see footnotes).
Boys | Normalized right AGV | 0.167 |
0.452 |
Normalized left AGV | 0.154 |
0.412 |
|
Girls | Normalized right AGV | 0.177 | 0.561 |
Normalized left AGV | 0.144 | 0.384 |
|
Boys | Normalized right AGV | 0.082 |
0.355 |
Normalized left AGV | 0.075 |
0.313 |
|
Girls | Normalized right AGV | 0.203 | 0.568 |
Normalized left AGV | 0.074 | 0.344 |
Because a significant subjective distress by timing interaction effect was obtained on right normalized AGV, and that partial correlations showed that right normalized AGV was significantly associated with externalizing problems, a moderated mediation effect was tested in boys, moderating the first path by timing of exposure, and adjusting the model for handedness, SES, maternal psychological functioning, and maternal alcohol usage. However, since neither path in the model showed a significant effect of maternal psychological functioning, this covariate was removed from the final model. The results indicated a significant moderated mediation effect (index of moderated mediation = 0.0135 and 95% confidence interval = [0.0006; 0.0369]), such that the mediation effect from higher subjective distress to more externalizing problems via larger right AGV was only significant for boys exposed on gestational day 212 (30.29 weeks) or later. However, since the moderation effect was no longer significant when adjusting for number of yearly postnatal life events, the mediation adjusting for postnatal measures was not tested.
As shown in
In girls, the results of the hierarchical regressions for right normalized AGV are presented in
Results of the hierarchical regression for left normalized AGV in girls are presented in
When testing the second path of the indirect effect we found that larger right normalized AGV predicted more externalizing problems in girls when controlling for timing, handedness, and maternal life events (
Because objective hardship was marginally associated with left and right normalized AGV, which were associated with externalizing problems, we tested for mediation effects, adjusting for timing of exposure, handedness, and maternal life events. However, in the model with left normalized AGV, neither path showed a significant maternal life events effect, so this covariate was removed from the final model. The analyses revealed that more objective hardship predicted larger left normalized AGV which in turn was associated with more externalizing problems (indirect effect = 0.286 and 95% confidence interval [0.0166; 0.9643]). The same mediation effect was also observed for the right normalized AGV (effect: 0.3831 and confidence interval [0.0106; 1.2727]). When adjusting for postnatal measures, the yearly number of postnatal life events was no longer significant, so it was trimmed out of the models. As such, the mediations through left and right AGV remain significant.
The first aim of this study was to determine whether there is an association between disaster-related PNMS and normalized AGV in boys and girls, and to investigate the extent to which timing of the ice storm
Our first hypothesis was that higher levels of maternal objective hardship and/or subjective distress, and/or a negative cognitive appraisal of the crisis would explain variance in child AGV as normalized by total brain volume. No significant direct associations between our disaster-related PNMS measures and normalized AGV in boys were detected. From our correlational analyses we found that higher SES was associated with larger AGV, and when controlling for SES and objective PNMS exposure, an interaction between subjective PNMS and timing of exposure emerged. When exposure to the ice storm occurred at or after gestational week 22, higher maternal subjective distress predicted larger normalized right AGV. When subjective PNMS was >13 (which is not very severe), the later the exposure, the larger the AGV, such that AGV in those exposed earlier were below the sample's average while AGV in those exposed later were above the sample's average (
These data add to a growing body of literature showing that the amygdala is vulnerable to early life adverse experiences and prenatal factors, and suggest that the timing of exposure may be important. Maternal cortisol levels and prenatal maternal depression have been associated with structural and functional changes in the amygdala, and some have shown that the effects are specific to girls (e.g., Favaro et al.,
Prenatal factors can also differentially influence structural properties of the amygdala at different points in postnatal development. For example, in one rodent study that considered only males, mild prenatal environmental stressors (starting on gestational day 14/21) was associated with smaller lateral, basolateral, and central nuclei on postnatal day 25 (i.e., early pubertal period), but not at postnatal days 7, 45, or 60. Interestingly, in the basolateral amygdala, but not the lateral or central nuclei, prenatal stress was associated with a shorter nucleus at postnatal day 25, no difference on postnatal day 45, but longer nucleus on postnatal day 60 (Kraszpulski et al.,
In boys, higher maternal subjective distress predicted higher externalizing scores through larger normalized right AGV, but only when the ice storm occurred at or after gestational week 30. This suggests that although maternal subjective distress can influence right AGV as of 22 weeks of gestation (as described above), the behavioral consequences of increased AGV are more likely to result from even later exposure (i.e., week 30). Whereas, there was no main effect of PNMS on internalizing or externalizing problems in boys, the partial correlations from both normalized left and right AGV predicted more externalizing problems, explaining 20% of the variance. Although the bivariate correlation suggested a tendency for left AGV to be associated with internalizing problems in boys, no association was detected from the partial correlations, suggesting that AGV was not associated with internalizing problems. Buss et al. (
Our first hypothesis was that higher levels of maternal objective hardship and/or subjective distress, and/or a negative cognitive appraisal of the crisis would explain variance in the child's normalized AGV. The regression analyses revealed that when controlling for timing and handedness, greater objective hardship tended to predict larger right and left normalized AGV. Although there was no moderating effect of timing, a main effect of timing was detected in the hierarchical regression, such that later timing of exposure was associated with larger right and left normalized AGV. This timing effect is in contrast to Buss et al. (
Our second hypothesis was that normalized AGV would mediate the association between PNMS and behavioral problems. The partial correlations revealed that when controlling for timing, handedness and maternal life events, larger right and left AGV were associated with more externalizing problems. The mediation analysis revealed that objective hardship was associated with larger left and right AGV, which in turn was associated with more externalizing problems at 11½ years of age. The existing literature suggests that more externalizing problems are associated with smaller AGV in adults (Matthies et al.,
The lack of associations with internalizing problems is surprising. We have previously reported with this sample that objective hardship and subjective distress are associated with internalizing problems when controlling for sex (King et al.,
The various elements of the maternal stress experience (i.e., objective hardship, subjective distress, and cognitive appraisal) on the brain appear to be different for males and for females. We report that for girls, objective hardship seems to be the strongest predictor of AGV. For boys, subjective distress has some mild influence on AGV. Cognitive appraisal was not associated with AGV in neither sex. Our findings that only objective hardship, not subjective distress, was associated with externalizing behaviors via AGV in girls are surprising because they do not fit the conventional model of subjective distress and cortisol where more subjective distress during pregnancy causes an increase in maternal cortisol, which affects fetal development. This suggests that some distinct aspects of the objective hardship, perhaps unrelated to maternal cortisol levels, are influencing child prenatal development. Unfortunately, no other human studies have investigated the specific effect of objective maternal hardship on the offspring brain so it is difficult to discuss our findings in light of previous research. Interestingly, it was maternal subjective distress, and not objective hardship, in late pregnancy, that was associated with larger AGV and increased externalizing behaviors in boys. These sex-specific effects, and how they interact with timing of exposure
Both maternal subjective distress and objective hardship have been associated with cognitive and behavioral outcomes in the child (Watson et al.,
There are limitations to this work that need to be acknowledged. An important limitation regarding the sex specific effects is the sample size that limited our power to test the interaction with sex. However, based on the existing literature on sex-specific effects, we used statistical approaches that have been suggested as best practices in studies that were underpowered to test sex differences (i.e., sex disaggregation, Heidari et al.,
Another important limitation to this study is that we did not collect maternal cortisol at the time of the stressor. This is a logistical limitation as the ice storm was a sudden-onset natural disaster and, due to delays in obtaining ethics approval, data collection only began 5 months after the disaster. In addition, although we were able to control for maternal reported life events, we did not have a direct measure of child self-reported life stressors in childhood. We used an indirect measure of child life events by using maternal related postnatal life events. These appear to be an important factor related to normalized amygdala volume in boys, and warrants further investigation. Another limitation is that we did not collect information on pubertal stage during the assessment at 11½ years of age. However, age at menarche was collected in 23 female participants at the 13½ year assessment. The mean age at menarche was 12. Only 5 out of the 23 female participants had attained menarche by the time of the MRI scan at age 11½. When looking at a scatterplot of the PNMS-by-AGV correlation, none of those 5 participants were outliers. No association was found between AGV in girls at age 11½ and their age at menarche (
Our study's main strength was the use of a sudden-onset, quasi-randomly distributed natural disaster as a stressor, rather than studying maternal psychological state or potentially non-independent life events. This allowed us to test for dose-response effects of PNMS on child neural and behavioral outcomes while reducing the influence of genetic factors. Moreover, we were able to assess the various elements of stress soon after the event occurred, thus providing us with a highly personalized, deep-level measure of the various stress levels. Additional strengths of our present study include its longitudinal prospective design with multiple assessment points between birth and childhood. This allowed us to include a number of maternal, familial, and postnatal child factors that may also influence child development, such as perinatal and concurrent SES, obstetric complications, gestational age at birth, maternal smoking and drinking habits during pregnancy, and maternal life events which were included as control variables in this study. Finally, segmentation of all child brain MRI was done with gold-standard manual segmentation. This may be especially important for the assessment of AGV because it has been suggested that automatic segmentation protocols are less reliable for smaller subcortical structures (Tae et al.,
Our present findings suggest that in boys, a mother's distress from a natural disaster, when experienced in the second half of pregnancy, can influence the development of her child's amygdala, which in turn mediates the association between subjective PNMS and externalizing behaviors when measured 11½ years later. In girls, the objective hardship experienced from a natural disaster predicted larger right and left normalized AGV, which in turn was associated with more externalizing behavior. This is, to the best of our knowledge, the first report linking PNMS with subsequent AGV and behavioral problems in childhood. Findings from the present study provide support for the hypothesis that susceptibility to behavioral problems may, in part, be programmed
We haven't obtained consent (mothers) or assent (children) to make the data available in publically accessible repositories, but they are available on request to the corresponding author.
This study was carried out in accordance with the recommendations of The Code of Ethics of the World Medical Association, and approved by the Douglas Mental Health University Institute Research Ethics Board. All subjects gave written informed consent or written informed assent for participants under the age of 18 years, in accordance with the Declaration of Helsinki.
SK designed and implemented Project Ice Storm. RD, DL, JP, and SK conceived of the current experiment. RD and RP ran the automated segmentation pipelines, which was overseen by MC. GE and RD ran the statistical analyses. RD did manual corrections of the amygdala and total brain volume segmentations derived from the automated segmentations, overseen by JP. SK, DL, MC, and JP provided intellectual contributions to the interpretation of data. RD interpreted the data and drafted an early version of the manuscript. SJ, SK, and DL provided intellectual contributions for the rationale, interpretation of the data, and prepared the final manuscript for submission.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The scanning protocol was designed by Arnaud Charil. The children were escorted through the scanning protocol at age 11½ by Rowena Lung, Karine Ferron, Gila Foomani, Weirquan Zeng, Marjolaine Massé, and Maria Papastergiou. The scans were obtained by MRI technicians André Cyr and Carollyne Hurst at the
The Supplementary Material for this article can be found online at: