Hostname: page-component-78c5997874-m6dg7 Total loading time: 0 Render date: 2024-11-20T00:36:12.624Z Has data issue: false hasContentIssue false

Effects of maternal depression and prenatal SSRI exposure on executive functions and susceptibility to household chaos in 6-year-old children: prospective cohort study

Published online by Cambridge University Press:  07 September 2020

Gurpreet Dhaliwal
Affiliation:
Division of Developmental Pediatrics, University of British Columbia, Vancouver, Canada
Whitney M. Weikum
Affiliation:
Division of Developmental Pediatrics, University of British Columbia, Vancouver, Canada
Alexia Jolicoeur-Martineau
Affiliation:
Jewish General Hospital, Montreal, Canada
Ursula Brain
Affiliation:
BC Children's Hospital Research Institute, Vancouver, Canada
Ruth E. Grunau
Affiliation:
University of British Columbia, Vancouver; and BC Children's Hospital Research Institute, Vancouver, Canada
Tim F. Oberlander*
Affiliation:
University of British Columbia, Vancouver; and BC Children's Hospital Research Institute, Vancouver, Canada
*
Correspondence: Tim Oberlander. Email: [email protected]
Rights & Permissions [Opens in a new window]

Abstract

Background

Maternal depressed mood during pregnancy may shape a child's adaptation to their environment and engagement in goal-directed behaviour such as executive functions. Whether everyday household context also alters executive functions in children with prenatal selective serotonin reuptake inhibitor (SSRI) antidepressant exposure remains to be determined.

Aims

To examine the impact of prenatal depressed maternal mood and SSRI exposure on child executive functions and to determine whether these exposures shape a susceptibility to household chaos.

Method

A prospective cohort study of mothers and their children (118 mother–children dyads (47 SSRI-exposed, 71 non-exposed)) followed from the second trimester to 6 years. Regression models examined relationships between maternal depressed mood and household chaos on maternal report of child executive functions. Competitive-confirmatory regression models examined whether children were susceptible to household chaos or were positively influenced by less chaos.

Results

Prenatal SSRI exposure, third-trimester maternal depressed mood and household chaos in a three-way interaction were associated with executive functions within a model of differential susceptibility. When household chaos was low, children of non-prenatally depressed mothers had better executive function than children of prenatally depressed mothers, regardless of whether the mothers were SSRI-treated. However, when household chaos was high, SSRI-exposed children of mothers who were not depressed during pregnancy had poorer executive functions at 6 years of age compared with SSRI-exposed children whose mothers were symptomatic during pregnancy.

Conclusions

The impact of household chaos depended on whether mothers were prenatally depressed and whether mothers were SSRI-treated.

Type
Papers
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.
Copyright
Copyright © The Author(s), 2020. Published by Cambridge University Press on behalf of Royal College of Psychiatrists

Maternal mental health and its impact on childhood cognitive development and behaviourReference Hughes, Roman, Hart and Ensor1 have been widely reported, but studies continue to yield small to moderate effect sizes.Reference Stein, Pearson, Goodman, Rapa, Rahman and McCallum2 Developmental outcomes vary and it is often unclear why some, but not all, children are negatively affected, or even why some children benefit under positive environmental conditions.

Long before birth, maternal mental health can shape diverse biological and behavioural processes that regulate arousal, attention and thinking in ways that facilitate goal-directed behaviour.Reference Fox, Levitt and Nelson3 These higher-order cognitive tasks, termed executive functions,Reference Diamond and Lee4 include cognitive flexibility, inhibition and working memory. These constructs reflect ability to plan, organise and focus attentionReference Hughes, Roman, Hart and Ensor1 and affect childhood behaviour and academic achievement.Reference Bull, Espy and Wiebe5

Prenatal exposure to SSRIs, and household chaos

Although executive functions are heritable, they are susceptible to early-life environmental factors such as chronic exposure to depressed maternal mood during pregnancy. Inherent to maternal mental health during pregnancy are the selective serotonin reuptake inhibitor (SSRI) antidepressants widely used to manage prenatal maternal mood disturbances, adding another early-life exposure that potentially shapes brain development and emerging executive functions.

Executive functions are dependent on multiple neurotransmitters, including serotonin (5-hydroxytryptamine, 5-HT), the very target of SSRIs. SSRIs primarily act by blocking reuptake of serotonin transporter protein (5-HTT), increasing the extent to which intrasynaptic serotonin remains active and available for neurotransmission. SSRIs readily cross the blood–brain barrier and placenta,Reference Rotem-Kohavi and Oberlander6 therefore it is conceivable that early changes in 5-HT signalling have a developmental impact on systems that subsequently regulate self-control and executive functions.Reference Weikum, Brain, Chau, Grunau, Boyce and Diamond7 Changes in serotonin signalling in prefrontal cortical networks have been linked with cognition, although findings are inconsistent.Reference Homberg and Lesch8 This raises intriguing questions about what might underlie developmental variations.

Disorganised, noisy and unpredictable environments affect the development of executive functionsReference Matheny, Wachs, Ludwig and Phillips9 and may lead to increased conduct problems and lower IQ in kindergarten-age children.Reference Deater-Deckard, Mullineaux, Beekman, Petrill, Schatschneider and Thompson10 Further, household chaos is highly intertwined with parental depression and both have been associated with poorer child executive functions.Reference Hughes, Roman, Hart and Ensor1 Yet, even in disorganised and busy circumstances, some children remain resilient and understanding why some children are susceptible to adversity and others are not remains a pressing question.

‘Differential susceptibility’ and ‘vantage sensitivity’

To date research has focused on identifying adverse developmental outcomes in children with prenatal SSRI exposure, such as attention-deficit hyperactivity disorder (ADHD), autism spectrum disorder and anxiety.Reference Zubin and Spring11 This has led to a focus on ‘diathesis-stress’ models to explain why some children may be more negatively affected by adverse environmental conditions.Reference Zubin and Spring11 However, the developmental impact of prenatal SSRI exposure remains inconsistent and contradictory.Reference Rotem-Kohavi and Oberlander6 Beyond a direct effect related to prenatal exposure to SSRIs or depressed maternal mood, interactions with daily life may shape variations in child development. To further understand the impact of interactions between the environment (i.e. household chaos) and biological vulnerabilities (i.e. prenatal maternal depression, prenatal SSRI exposure), the concept of ‘differential susceptibility’ has been proposed to explain why some children may have better outcome than other children under positive environmental conditions and worse outcome under adverse environmental conditions.Reference Belsky12,Reference Belsky13 An alternative theory is ‘vantage sensitivity’, which suggests that some children may have better outcomes under positive environmental conditions than other childrenReference Belsky, Pluess and Widaman14,Reference Jolicoeur-Martineau, Belsky and Szekely15 (see Jolicoeur-Martineau et al Reference Jolicoeur-Martineau, Belsky and Szekely15 for a review of the concepts of ‘differential susceptibility’ and ‘vantage sensitivity’).

The hypothesis

To understand variations in behaviour in children with prenatal SSRI exposure, we investigated the impact of both prenatal exposure to SSRIs and depressed maternal mood on early executive functions at 6 years of age, and whether the impact of these exposures shapes a response to household chaos (Fig. 1). We expected that prenatal exposure to SSRIs and maternal depressed mood in the context of increased household chaos might be associated with poorer executive functions. However, given that prenatal SSRI exposure has been associated with improved or sustained executive functions in animal modelsReference Homberg and Lesch8 and humans,Reference Weikum, Brain, Chau, Grunau, Boyce and Diamond7 prenatal SSRI exposure might ‘buffer’ a child from the effects of prenatal maternal mood and thereby shapes a differential susceptibility to a stressful everyday household context. In this sense, using competitive-confirmatory analysesReference Jolicoeur-Martineau, Belsky and Szekely15 we expected that prenatal depressed maternal mood and SSRI exposure would interact to reflect differential susceptibility, or even a vantage sensitivity to household chaos.

Method

Approval was obtained from the University of British Columbia Clinical Research Ethics Board and the Children's and Women's Health Centre of British Columbia Clinical Research Ethics Board (H00-70500, H05-70629, H08-01712). Written informed consent was obtained from all mothers. All procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.

Mothers (n = 191) were recruited during their second trimester of pregnancy either through self-referrals from the community or through referrals from the departments of reproductive psychiatry and family medicine at BC Women's Hospital, Vancouver, Canada, between 2003 and 2009. Mothers and children were studied again when the children had reached 6 years of age, between 2009 and 2014. Inclusion criteria were singleton pregnancy, gestational age at birth (>37 weeks) and no fetal anomalies on antenatal ultrasounds. Exclusion criteria were presence of maternal bipolar disorder, concurrent illicit drug use and maternal medical conditions (e.g. diabetes, hypertension). Of the initial recruits, 8 withdrew before delivery, another 15 withdrew before the end of the child's first year and, by 6 years, 50 declined to participate, were unavailable for study (families had moved, etc.) or there were significant incomplete study data. This left 118 mother–child dyads (47 prenatally SSRI-exposed and 71 non-SSRI-exposed) available for this study. Comparison of the mothers originally recruited and their children who did not participate (n = 115 (60.21%) not SSRI-exposed, n = 76 (39.79%) SSRI-exposed) showed demographic similarities on birthweight, birth length, head circumference, gender, Apgar score and maternal education. However, in the non-participating group, maternal age was younger (32 v. 34 years, P = 0.03), prenatal depressed mood was slightly higher (mean Hamilton Rating Scale for Depression (HRSD) score 9.4 v. 7.3, P = 0.02) and gestational age was shorter (39.1 v. 39.7 weeks, P = 0.04).

Maternal mood

Maternal depressed mood was assessed at 26 weeks, 36 weeks and 6-year follow-up using the clinician-rated Hamilton Rating Scale for Depression (HRSD),Reference Hamilton16 yielding a range of symptoms reflecting no depression (0–7), mild depression (8–16), moderate depression (17–23) and severe depression (>23). Prenatal mood (HRSD score at 26 and 36 weeks) was tabulated as a mean of prenatal depressed mood. Initially, maternal mood was used as a continuous measure in the regression models. Then to investigate interactions between maternal mood and SSRI exposure, we used categorical approaches to examine how maternal mood was associated with offspring executive functions. An HRSD score >8 was used to denote the presence of mild to moderate depressive symptoms and to categorically separate depressed from non-depressed mothers.Reference Kearns, Cruickshank, McGuigan, Riley, Shaw and Snaith17 In addition, the self-report Beck Depression Index (BDI) was obtained at 3-year follow-up because data were obtained at 3 years using a mailing of questionnaires.

Home environment

Household chaos offers a measure of an environment that is high in noise and crowding and low in regularityReference Matheny, Wachs, Ludwig and Phillips9 and has been linked to poorer child behaviour.Reference Deater-Deckard, Mullineaux, Beekman, Petrill, Schatschneider and Thompson10 Home environment was assessed using the Confusion, Hubbub and Order Scale (CHAOS) at age 6. This is a self-report scale (completed here by the mothers) developed as a 15-item measure of home chaos and disorganisation.Reference Matheny, Wachs, Ludwig and Phillips9 Higher scores on this scale are indicative of more disorganised, confusing and noisy homes.

Executive functions

Children's behaviour was assessed using the Behavior Rating Inventory of Executive Function (BRIEF) – an 86-item parent-completed rating scale widely used in clinical practice.Reference Donders18 The BRIEF's Global Executive Composite Index reflects the behavioural manifestation of executive function (i.e. inhibition, shift, emotional control, initiation, planning). The BRIEF comprises eight scales (Working memory, Inhibit, Shift, Emotional control, Initiate, Plan/Organise, Organisation of Materials, Monitor), which yields two Indexes: (a) Metacognition, reflecting the ability to cognitively self-manage tasks and monitor performance; and (b) Behavioural Regulation, the ability to shift and modulate emotions and behaviour via appropriate inhibitory control. In this study, the BRIEF's Global Executive Composite score was used as the key dependent measure of overall executive functions. Higher scores indicate poorer executive functions.

Statistical analysis

Regression modelling was utilised to examine relationships between SSRI exposure, CHAOS score, pre- and postnatal maternal depression (HRSD) scores, and child BRIEF Global Executive Composite T-scores. Measures of postnatal maternal mood (when the infant was 6 months and 3 years) were also obtained but, owing to high collinearity between measures, these were not used in our analytic models. Postnatal maternal mood at 6 years postpartum was included in our models as a potential confounder, as maternal depressed mood may influence maternal perception of household chaos and child executive functions.

Models were constructed in a stepwise manner: step 1 was for main effects only, step 2 for main effects and two-way interaction and step 3 for main effects, two-way interaction and three-way interaction. Initially, models used prenatal maternal depression as a continuous variable, then, to examine additive interactions, a dichotomous measure of prenatal maternal depression was used (coded as 1 when HRSD ≥ 8 and 0 otherwise). We used the following reparametrised equations:

$${\rm BRIEF} = \beta _0 + \beta _{\rm dep}\;\lpar {{\rm CHAOS}\hbox{-}{C}} \rpar + \varepsilon \comma \;$$

for children exposed to perinatal depression (with or without SSRI exposure);

$${\rm BRIEF} = \beta _0 + \beta _{\rm noDep\_noSSRI}\;\lpar {{\rm CHAOS}\hbox{-}{C}} \rpar + \varepsilon \comma \;$$

for children non-exposed to both perinatal depression and SSRIs; and

$${\rm BRIEF} = \beta _0 + \beta _{\rm noDepSSRI}\;\lpar {{\rm CHAOS}\hbox{-}{C}} \rpar + \varepsilon \comma \;$$

for children non-exposed to perinatal depression but exposed to SSRIs.

The parameters were the intercept (β 0), the slope for children exposed to perinatal depression (β dep), the slope for children non-exposed to both perinatal depression and SSRIs (β noDep–noSSRI), the slope for children non-exposed to perinatal depression but exposed to SSRIs (β noDep–SSRI) and the cross-over point between the two slopes (C). The cross-over point (C) distinguishes a vantage sensitivity from differential susceptibility. Namely, if C ≠ max(CHAOS) (differential susceptibility), then the two lines cross over in the middle of the graph and the children non-exposed to prenatal depression may have a worse outcome than the exposed children when CHAOS > C. Alternatively, if C = max(CHAOS) (vantage sensitivity), then the two lines meet at the right of the graph (no cross-over) and children non-exposed to depression cannot have a worse outcome than children exposed to depression, at any level of household chaos. Both the differential susceptibility and vantage sensitivity models assume that children exposed to perinatal depression would not be influenced by the environment (i.e. that β dep = 0). However, as there remains a possibility that the environment exerts a slight effect even on exposed children, the vantage sensitivity and differential susceptibility models are further separated into two groups: a weak model (β dep ≠ 0, βdep < β noDep_noSSRI and β dep < β noDep_noSSRI) and a strong model (β dep = 0).

In strong models, some individuals are affected by the environmental exposure of interest, whereas others are not. In weak models, all are affected by the environmental exposure, but some more strongly than others.Reference Jolicoeur-Martineau, Belsky and Szekely15 We measured the effect size using the coefficient of determination (R 2).

Competitive-confirmatory analysis

To test whether the interaction between mother's perinatal depressed mood and household CHAOS reflected diathesis–stress, differential susceptibility or vantage sensitivity, we used competitive-confirmatory analyses.Reference Jolicoeur-Martineau, Belsky and Szekely15 The Akaike information criteria (AIC) and the 95% confidence interval of the cross-over point were used to determine which of the four models best fit the data. Only the vantage sensitivity and strong differential susceptibility model testing are reported; all models were tested using Statistical Package for the Social Sciences (SPSS) for Windows19 and SAS System 9.4 for Windows.20

Results

Maternal characteristics were generally similar for both SSRI-exposed and non-exposed groups (Table 1) and for subgroups of non-SSRI-treated mothers (depressed versus non-depressed (Table 2); however, mothers treated with an SSRI had higher depression symptom scores (HRSD) and fewer years of education (a proxy measure of socioeconomic status) than mothers not taking SSRIs. For years of education, in the SSRI-treated mothers the mean was 16.23 years (s.d. = 3.39) and in the non-SSRI-treated mothers it was 17.87 years (s.d. = 3.03), reflecting a slightly higher middle-class status. Mothers reported a mean 240 days (s.d. = 68) of SSRI treatment during their pregnancy. There were no significant differences between the cohorts in levels of home environment (CHAOS score) (Table 1). Maternal report of home environment was associated with depressed mood at 6 years (R 2 = 0.357; P < 0.01) but not prenatal depression symptoms (R 2 = 0.146; P = 0.113). We found no significant differences in BRIEF scores between SSRI-exposed and non-exposed groups. However, prenatally depressed mothers (HRSD ≥ 8) had children with significantly poorer BRIEF scores (50.76 v. 56.63; P < 0.001) and reported significantly higher CHAOS scores (20.63 v. 18.97; P = 0.004); neither gender, education nor mother's age at childbirth were significant in the model. None of the children were being treated with a psychotropic medication at the time of the 6-year follow-up.

Table 1 Mother–child cohort characteristics

SSRI, selective serotonin reuptake inhibitor; HRSD, Hamilton Rating Scale for Depression; BDI, Beck Depression Inventory; BRIEF, Behavior Rating Inventory of Executive Function; GEC, Global Executive Composite; CHAOS, Confusion, Hubbub and Order Scale.

a. Alcohol intake was measured by the number of drinks (units of alcohol) across pregnancy.

b. n = 69.

* P < 0.05.

Table 2 Maternal characteristics of subgroup of mothers not treated with an antidepressant during pregnancy

HRSD, Hamilton Rating Scale for Depression; BDI, Beck Depression Inventory.

a. Alcohol intake was measured by the number of drinks across pregnancy.

b. n = 69.

* P < 0.05.

Primary analysis: using continuous prenatal depression

Regression modelling examining the influence of prenatal maternal mood, SSRI exposure and household chaos on BRIEF scores, controlling for concurrent maternal mood (Table 3), demonstrated main effects for prenatal mood, SSRI exposure and CHAOS score. A two-way interaction emerged between SSRI exposure and prenatal maternal mood, suggesting that for mothers who took antidepressants prenatally, the more depressed they were prenatally (i.e. they did not benefit from pharmacotherapy), the less self-regulated (higher BRIEF scores) their child appeared to be at age 6. Additionally, when CHAOS scores were added, a three-way interaction emerged between prenatal mood, SSRI exposure and household chaos, suggesting that for mothers who took antidepressants prenatally, the more depressed they were prenatally, the less susceptible their children were to a chaotic household environment. Postnatal maternal depression was not significantly associated with BRIEF scores.

Table 3 Prediction of children's BRIEF scores at 6 years of age from the interaction of prenatal maternal depression (continuous or binary), SSRI exposure and CHAOS score (n = 118)a, b

BRIEF, Behavior Rating Inventory of Executive Function; SSRI, selective serotonin reuptake inhibitor; CHAOS, Confusion, Hubbub and Order Scale; AIC, Akaike information criteria.

a. Maternal prenatal and postnatal depression were assessed using the Hamilton Rating Scale for Depression (HRSD). Prenatal depression as binary is defined as 1 when HRSD ≥ 8, and 0 otherwise.

b. Main effects: prenatal mood, β = 0.20, 95% CI −0.23 to 0.62; SSRI exposure, β = −6.78, 95% CI −12.98 to −0.59; and CHAOS scores, β = 1.32, 95% CI 0.69 to 1.95. Two-way interaction: SSRI exposure and prenatal maternal mood, β = 2.70, 95% CI 1.02 to 4.38. Three-way interaction: prenatal mood, SSRI exposure and CHAOS scores, β = −0.11, 95% CI −0.18 to −0.03. Postnatal maternal depression was not associated with BRIEF scores (P > 0.05).

tP < 0.10, *P < 0.05, **P < 0.01, ***P < 0.001.

Secondary analysis: dichotomised prenatal depression

To further examine these interactions and investigate the role of prenatal maternal depressed mood in particular, a dichotomised prenatal maternal mood variable (coded 1 when HRSD ≥ 8, and 0 otherwise) was used to distinguish mild to moderate depressive symptoms from low non-impairing symptoms in a regression model. In this model the influence of prenatal maternal mood, prenatal SSRI exposure and household chaos on child BRIEF scores was examined, accounting for postnatal maternal mood (as a continuous variable) (Table 3). Main effects were observed for CHAOS score (β = 1.35, 95% CI 0.76 to 1.95), but were not observed for prenatal mood (β = 4.13, 95% CI −0.96 to 9.22) or prenatal SSRI exposure (β = −3.87, 95% CI −8.63 to 0.89). Maternal depressed mood at 6 years was not associated with BRIEF scores (β = −0.03, 95% CI −0.31 to 0.26).

A two-way interaction was also apparent between SSRI exposure and prenatal maternal mood, which suggests that mothers who took SSRI medication and were prenatally depressed had children with poorer executive functions (higher BRIEF scores) (β = 43.18, 95% CI 21 to 65.36).

In addition, a three-way interaction between prenatal mood, SSRI exposure and CHAOS scores also emerged, suggesting that children's prenatal exposure to antidepressants and maternal depression was associated with less susceptibility to household chaos (β = −2, 95% CI −3.01 to −0.98). This model had a considerably lower AIC (497.76) than the model with continuous prenatal maternal depression (503.14), suggesting that prenatal maternal depression may exert an effect when it is observed clinically but not as depression increased, thus acting in a binary fashion rather than continuously.

In this model, executive functions among children whose mothers were at least mildly symptomatically depressed during pregnancy (i.e. HRSD > 8) remained relatively stable even with increasing levels of household CHAOS scores, regardless of their prenatal SSRI exposure status: SSRI-exposed, β = −0.67, 95% CI −1.53 to 0.18, P = 0.12; non-exposed, β = 1.01, 95% CI −0.53 to 2.55, P = 0.20. However, for children of non-depressed mothers (i.e. HRSD < 8), increased household chaos was associated with poorer executive functions regardless of SSRI-exposed status (SSRI-exposed, β = 1.54, 95% CI 0.16 to 2.93, P = 0.03; non-exposed, β = 1.34, 95% CI 0.69 to 1.99, P = 0.0001).

Competitive-confirmatory analyses

Competitive-confirmatory analyses were used to determine whether the interactions reflected a differential susceptibility or a vantage sensitivity. For these analyses an n = 139 was used, given that postnatal depression measures did not contribute to self-regulation at 6 years. For children exposed to prenatal maternal depression (SSRI-exposed or not), CHAOS scores were not significantly associated with BRIEF scores, thus we considered these children to be non-susceptible (solid green line, Fig. 2). We also tested whether those exposed to both SSRI and maternal prenatal depression would constitute a non-susceptible group, given a negative slope that approached zero (β = −0.67). The model with this grouping had a poorer fit (based on the AIC) and was not considered further. The best model was the differential susceptibility strong model (AIC = 592.94), and in order from best to worst, the other models were vantage sensitivity strong (AIC = 593.79), differential susceptibility weak (AIC = 594.74) and vantage sensitivity weak (AIC = 595.11) (Table 4).

Relationships include the impact of the multiple effects of selective serotonin reuptake inhibitor (SSRI) treatment and exposure (i.e. improved prenatal and postnatal maternal mood, child brain development), maternal mood (pre- and postnatal) and the role of household chaos. We used this model to test the hypothesis that prenatal depressed maternal mood and SSRI exposure would interact to reflect a differential susceptibility, or even a vantage sensitivity, to household chaos.

Fig. 1 Graphic representations of the competitive-confirmatory model and relationships between the different elements affecting child executive functions at 6 years of age.

Table 4 Confirmatory models (n = 139)a, b

SSRI, selective serotonin reuptake inhibitor; AIC, Akaike information criteria.

a. n = 139 reflects that measures of postnatal depression were not used in this model since it had been shown in a previous model not to contribute to self-regulation at 6 years.

b. In the vantage sensitivity strong model, F(2,136) = 17.41***; in the differential susceptibility strong model, F(3,135) = 12.69***; in the vantage sensitivity weak model, F(3,135) = 11.80***; and in the differential susceptibility weak model, F(4,134) = 9.51***.

***P < 0.001.

For the differential susceptibility strong model, the intercept was significant (β = 56.94, s.e. = 1.12, P < 0.0001). In particular, the slope for children of non-depressed, non-SSRI-treated mothers was significant (β = 1.09, s.e. = 0.27, P < 0.0001), the slope for children of non-depressed SSRI-treated mothers was significant (β = 2.07, s.e. = 0.57, P = 0.0004) and the cross-over point was significant (β = 23.64, s.e. = 1.44, P < 0.0001, 95% CI 20.79–26.49), with a range of possible values (CHAOS score is between 15 and 27). R 2 for the model was 0.22. In contrast, the slope for children of depressed mothers was 0.

For the vantage sensitivity strong model, the intercept was significant (β = 57.58, s.e. = 1.05, P < 0.0001). The slope for children of depressed mothers was 0, whereas the slope for those of non-depressed and non-SSRI-exposed mothers was significant (β = 0.74, s.e. = 0.16, P < 0.0001), the slope for children of non-depressed and SSRI-exposed mothers was significant (β = 1.38, s.e. = 0.28, P < 0.0001) and the cross-over point was fixed at the maximum CHAOS score.Reference Hamilton16 The R 2 for the model was 0.20.

Discussion

At 6 years of age children of mothers who experienced symptoms of mild to severe depression during pregnancy had generally poorer executive functions than children of mothers who were not symptomatic during pregnancy (Fig. 1). Moreover, executive functions remained stable, even in the context of higher home chaos and regardless of whether mothers had been treated with an SSRI during pregnancy. In contrast, children of mothers who showed fewer depressive symptoms during the third trimester were affected by home chaos, but the degree to which they were affected depended on prenatal exposure to an SSRI. At lower levels of household chaos, SSRI-exposed children had better executive function skills compared with non-exposed children, but as chaos increased, executive functions suffered among SSRI-exposed children. Thus, both prenatal depressed maternal mood and SSRI exposure appeared to shape different responses to home environment. However, it is also conceivable that home environment may have been more chaotic because of child dysregulation and in this sense, chaos could have also been a predictor of child executive function, and the relationship between chaos and child development may in fact be bidirectional.

Susceptibility to household chaos

The magnitude of the cross-over point (i.e. the point where prenatal depressed mood was affected by level of household chaos) between patterns of executive functions and increasing levels of household chaos at 6 years in our competitive-confirmatory models suggested the possibility of differential susceptibilityReference Jolicoeur-Martineau, Belsky and Szekely15 (Fig. 2). Namely, to the left of the cross-over point, at lower levels of chaos, children of mothers who were not particularly symptomatic during their pregnancies had better executive functions than children whose mothers were symptomatic during pregnancy. To the right of the cross-over point, at higher levels of chaos, the converse was true: children of less depressed mothers had worse executive functions than children of depressed mothers. Additionally, children of mothers who were treated with an SSRI but were not depressed during pregnancy (as measured with a HRSD score < 8) and perhaps derived some therapeutic benefit from pharmacotherapy, were more susceptible to household chaos than children of mothers who were not prenatally depressed and not SSRI-treated.

Children of prenatally depressed mothers (regardless of SSRI exposure) were not susceptible (solid green line) to household chaos. However, children of non-depressed mothers living with a lower level of household chaos differed in their response to a calmer environment. Namely, children of non-depressed mothers showed better executive functions than children of prenatally depressed mothers, but only where Confusion, Hubbub and Order Scale (CHAOS) scores were <23.6 (left-side vertical bar on horizontal line). In contrast, children of non-depressed mothers where CHAOS scores were >26.5 had significantly higher Behavior Rating Inventory of Executive Function (BRIEF) scores (poorer executive functions compared with children of depressed mothers (right-side vertical bar on horizontal line)).

Fig. 2 Confirmatory model with the best fit (differential susceptibility strong model) (refer to Table 4).

Prenatal exposure to SSRI antidepressants to treat mood disturbances during pregnancy have added another developmental risk.Reference Rotem-Kohavi and Oberlander6 Given that serotonergic networks play a prominent role in the development of the prefrontal cortex, our findings might also reflect early changes in serotonin signalling. Converging evidence from animal models and human research links increased 5-HT signalling with improved cognition.Reference Munafo, Durrant, Lewis and Flint21 In animal models, increased intrasynaptic 5-HT, secondary to genetic variations, may improve cognitive functions and cognitive flexibility.Reference Brigman, Mathur, Harvey-White, Izquierdo, Saksida and Bussey22 In humans, similar genetic variations have been associated with improved attentional inhibition, cognitive flexibility, memory and reaction time.Reference Homberg and Lesch8

Maternal mood during early childhood

Our findings also highlight the importance of assessing maternal mood long after delivery and its impact on child executive functions.Reference Stein, Pearson, Goodman, Rapa, Rahman and McCallum2,Reference Comas, Valentino and Borkowski23 Previous studies report that greater maternal depressive symptoms when the child is aged 2 were associated with poorer performance on executive function tasks in children at age 3, and poorer executive functions performance at age 3 predicted increased externalising symptoms at age 6.Reference Roman, Ensor and Hughes24 Higher levels of maternal depression in early childhood are associated with poorer executive functions in offspring at age 18 months,Reference Brigman, Mathur, Harvey-White, Izquierdo, Saksida and Bussey22 which appear to extend to poorer attentional control at 8 years.Reference Pearson, Bornstein, Cordero, Scerif, Mahedy and Evans25

Parental scaffolding, or support of development of new skills, is viewed as critical to the development of children's executive functions,Reference Fay-Stammbach, Hawes and Meredith26 suggesting that goal-directed behaviour could be disrupted in chaotic home environments. Home environments that are more chaotic have been associated with fewer and lower-quality caregiver interactions.Reference Comas, Valentino and Borkowski23 Greater environmental confusion in the home environment was associated with lower executive function scores in 4-year-olds.Reference Berry, Blair and Willoby27 A number of conceptual frameworks have been proposed to explain how environment predicts development. Until recently, attention has focused on a diathesis–stress framework, whereby some individuals are more vulnerable to negative events because of an inherent ‘vulnerability’ (e.g. psychological trait or ‘risk gene’).Reference Pluess and Belsky28 The diathesis–stress model asserts that some individuals carry ‘risk’ factors (e.g. depression, genotype) that make them disproportionately susceptible to adverse environments.Reference Pluess and Belsky28 The model implies that, in the absence of adversity, there should be no differences between vulnerable and resilient individuals, whereby some individuals are disproportionately adversely affected by negative exposures, while others may conceivably benefit from supportive and enriching experiences.

The vantage sensitivity and differential susceptibility models

To date, studies have focused on searching for ‘main effects’ outcomes related to prenatal SSRI exposure. This has led to conflicting findings that offer limited insight into why some, but not all, children are at risk, or who might even benefit from maternal antidepressant treatment during gestation. To account for the negative effects of contextual adversity, the differential susceptibility model has been proposed to account for susceptible individuals who are not just vulnerable to adversity, but are more developmentally plastic or malleable.Reference Belsky and Pluess29 The differential susceptibility model asserts that some individuals are susceptible to both the ‘dark side’ as well as the ‘bright side' of environmental exposures.Reference Homberg and Lesch8,Reference Jolicoeur-Martineau, Belsky and Szekely15,Reference Belsky and Pluess29 However, for others a vantage sensitivity may make them disproportionately susceptible to positive environments and they may actually benefit from positive or supportive factors (i.e. genetic variations) or environments.Reference Pluess and Belsky28 In the current study, children of mothers who were not depressed during pregnancy were susceptible to both lower and higher levels of home chaos. Under less chaotic environments, they had better executive functions than children of depressed mothers in a similar setting, suggesting a ‘vantage sensitivity’. In a positive environment, susceptible children with such a vantage sensitivity could do very well, but as adversity increased, these children's behaviour suffered. For some children, maternal prenatal depression may be a ‘buffer’ against a subsequently chaotic home environment, whereas for others (i.e. with prenatal SSRI exposure), susceptibility to the home environment is increased. This may reflect a positive impact of maternal antidepressant treatment, which might confer a long-term benefit to the child under some, but not all, circumstances.Reference Jolicoeur-Martineau, Belsky and Szekely15

Limitations

Key limitations need to be mentioned. We could not control the sample size available for this follow-up component of our prospective longitudinal cohort study. The available sample may have constrained the power needed to test two- and three-way interactions, and larger studies will need to confirm our findings. Moreover, the use of maternal report of child executive functions raises the possibility that depressed mood may have biased behavioural observations; however, consistency between teacher and parent report has been previously reported.Reference Donders18 The study lacked information about the effect of fathers, maternal partners or significant others in shaping household chaos. Our use of dichotomised maternal mood measures could misclassify mood status and may not reflect the impact of patterns of mood trajectories across pre- and postnatal periods, which have been examined by Park et al.Reference Park, Brain, Grunau, Diamond and Oberlander30

Further, SSRI use in pregnancy is inherently related to maternal depression, but not all women in our cohort appeared to benefit from prenatal SSRI treatment. Some women in the SSRI group were relatively asymptomatic during pregnancy (i.e. HRSD scores <8) and this might reflect that SSRI treatment resulted in symptom remittance. However, variations in maternal illness severity still existed and unmeasured maternal characteristics may have made them inherently different from those who were treated with an SSRI and remained symptomatic. In particular, CHAOS scores for SSRI-treated women who had relatively low levels of depressive symptoms (prenatal HRSD < 8 score, n = 14), compared with SSRI-treated women who remained symptomatically depressed (HRSD > 8; n = 33), were not substantially different. Although an HRSD score >8 reflected the presence of mild to moderate depressive symptoms, we were not able to determine associations between her mood disturbance and everyday function. Moreover, this symptom range (i.e. >8) maybe too broad, as it included both mild and severe symptoms and could therefore be subject to many confounders, such as treatment adherence, drug dosing and comorbid maternal conditions. Further, in this study we examined the association of SSRI treatment (maternal) in pregnancy on child behaviour (executive functions) 6 years later, which likely reflects the impact of prenatal SSRI exposure on the developing fetal brain. Conceivably, this is a different context from exposure to an SSRI in a mature brain, suggesting that executive functions may be affected in a different way than when the exposure occurs during periods of early brain development. Finally, distinguishing the impact of maternal mood (pre- and postnatal) from SSRI exposure is challenging in cohort studies where exposures cannot be randomised nor can the impact of variations in the severity of maternal mood be controlled for introducing possible ‘confounding by indication’.

Implications

This study was undertaken to further examine variations in early child behaviour in the context of maternal mental illness, to understand how exposure to maternal depressed mood (pre- and postnatal) and the antidepressants used to treat these disorders may shape child executive functions and whether these exposures are associated with a susceptibility to household chaos in 6-year-olds. Maternal depressed mood and antidepressant treatment during pregnancy have been associated with a child's emerging ability to adapt to environmental demands and engage in goal-directed behaviour, known as executive functions. Why some, but not all, exposed children are affected remains a key question. Although depressed maternal mood prenatally was associated with lower executive functions in offspring at 6 years of age, the degree of susceptibility to household environment appeared to depend both on whether the mothers were prenatally depressed and whether they had been treated with an SSRI. Children exposed to SSRIs prenatally were more susceptible, but the impact depended on levels of home chaos. Although prenatal depressed maternal mood may ‘buffer’ a child against a chaotic home environment, prenatal SSRI exposure did not appear to confer a protection against a susceptibility to household confusion. Namely, children not exposed to prenatal maternal depressed mood had better executive functions at lower levels of chaos, but when chaos increased, demonstrated poorer executive functions. Interventions to test whether better treatment of maternal depression during pregnancy and in the early postnatal period, as well as modifying household chaos, improve developmental outcomes in this setting are warranted.

Data availability

The data that support the findings of this study are available on request from the corresponding author, T.F.O. The data are not publicly available due to privacy/ethical restrictions.

Acknowledgements

We are grateful to Marie-Noel Wharton and Dawna Brown for their editorial assistance.

Author contributions

T.F.O., G.D., W.M.W., U.B. formulated the research questions. T.F.O., G.D., W.M.W., U.B., R.E.G. designed the study. T.F.O., U.B. carried out the study. T.F.O., G.D., W.M.W., A.J.-M. analysed the data. T.F.O., G.D., W.M.W., U.B., A.J.-M., R.E.G. wrote the article.

Funding

T.F.O. is the R. Howard Webster Professor in Brain Imaging and Early Child Development at the University of British Columbia and his work is supported by the BC Children's Hospital Research Institute (BCCHR) and the Canadian Institutes for Health Research (MOP 54490, 57837, 86296). R.E.G. holds a salary award from the BCCHR. W.M.W. is supported by the Sunny Hill Foundation for Children.

Declaration of interest

None.

ICMJE forms are in the supplementary material, available online at https://doi.org/10.1192/bjo.2020.73.

Footnotes

*

Joint first authors.

References

Hughes, C, Roman, G, Hart, MJ, Ensor, R. Does maternal depression predict young children's executive function? A 4-year longitudinal study. J Child Psychol Psychiatry 2013; 54: 169–77.CrossRefGoogle ScholarPubMed
Stein, A, Pearson, RM, Goodman, SH, Rapa, E, Rahman, A, McCallum, M, et al. Effects of perinatal mental disorders on the fetus and child. Lancet 2014; 384: 1800–19.CrossRefGoogle Scholar
Fox, SE, Levitt, P, Nelson, CA III. How the timing and quality of early experiences influence the development of brain architecture. Child Development 2010; 81: 2840.CrossRefGoogle ScholarPubMed
Diamond, A, Lee, K. Interventions shown to aid executive function development in children 4 to 12 years old. Science 2011; 333: 959–64.CrossRefGoogle ScholarPubMed
Bull, R, Espy, KA, Wiebe, SA. Short-term memory, working memory, and executive functioning in preschoolers: longitudinal predictors of mathematical achievement at age 7 years. Dev Neuropsychol 2008; 33: 205–28.CrossRefGoogle ScholarPubMed
Rotem-Kohavi, N, Oberlander, TF. Variations in neurodevelopmental outcomes in children with prenatal SSRI antidepressant exposure. Birth Defects Res 2017; 109: 909–23.CrossRefGoogle ScholarPubMed
Weikum, WM, Brain, U, Chau, CM, Grunau, RE, Boyce, WT, Diamond, A, et al. Prenatal serotonin reuptake inhibitor (SRI) antidepressant exposure and serotonin transporter promoter genotype (SLC6A4) influence executive functions at 6 years of age. Front Cell Neurosci 2013; 7: 180.CrossRefGoogle ScholarPubMed
Homberg, JR, Lesch, KP. Looking on the bright side of serotonin transporter gene variation. Biol Psychiatry 2011; 69: 513–9.CrossRefGoogle ScholarPubMed
Matheny, AP, Wachs, TD, Ludwig, JL, Phillips, K. Bringing order out of chaos – psychometric characteristics of the Confusion, Hubbub, and Order Scale. J Appl Dev Psychol 1995; 16: 429–44.CrossRefGoogle Scholar
Deater-Deckard, K, Mullineaux, PY, Beekman, C, Petrill, SA, Schatschneider, C, Thompson, LA. Conduct problems, IQ, and household chaos: a longitudinal multi-informant study. J Child Psychol Psychiatry 2009; 50: 1301–8.CrossRefGoogle ScholarPubMed
Zubin, J, Spring, B. Vulnerability: a new view of schizophrenia. J Abnorm Psychol 1977; 86: 103–26.CrossRefGoogle ScholarPubMed
Belsky, J. Theory testing, effect-size evaluation, and differential susceptibility to rearing influence: the case of mothering and attachment. Child Dev 1997; 68: 598600.CrossRefGoogle ScholarPubMed
Belsky, J. Variation in susceptibility to environmental influence: an evolutionary argument. Psychol Inq 1997; 8: 182–6.CrossRefGoogle Scholar
Belsky, J, Pluess, M, Widaman, KF. Confirmatory and competitive evaluation of alternative gene-environment interaction hypotheses. J Child Psychol Psychiatry 2013; 54: 1135–43.CrossRefGoogle ScholarPubMed
Jolicoeur-Martineau, A, Belsky, J, Szekely, E, et al. Distinguishing differential susceptibility, diathesis-stress, and vantage sensitivity: beyond the single gene and environment model. Dev Psychopathol 2020; 32: 7383.CrossRefGoogle ScholarPubMed
Hamilton, M. A rating scale for depression. J Neurol Neurosurg Psychiatry 1960; 23: 5662.CrossRefGoogle ScholarPubMed
Kearns, NP, Cruickshank, CA, McGuigan, KJ, Riley, SA, Shaw, SP, Snaith, RP. A comparison of depression rating scales. Br J Psychiatry 1982; 141: 45–9.CrossRefGoogle ScholarPubMed
Donders, J. The Behavior Rating Inventory of Executive Function: introduction. Child Neuropsychol 2002; 8: 229–30.CrossRefGoogle Scholar
IBM. IBM SPSS Statistics for Windows. IBM Corp, 2017.Google Scholar
SAS Institute Inc. SAS for Windows. SAS Institute Inc, 2013.Google Scholar
Munafo, MR, Durrant, C, Lewis, G, Flint, J. Gene + environment interactions at the serotonin transporter locus. Biol Psychiatry 2009; 65: 211–9.CrossRefGoogle ScholarPubMed
Brigman, JL, Mathur, P, Harvey-White, J, Izquierdo, A, Saksida, LM, Bussey, TJ, et al. Pharmacological or genetic inactivation of the serotonin transporter improves reversal learning in mice. Cereb Cortex 2010; 20: 1955–63.CrossRefGoogle ScholarPubMed
Comas, M, Valentino, K, Borkowski, J. Maternal depressive symptoms and child temperament: longitudinal associations with executive functioning. J Appl Dev Psychol 2014; 35: 156–67.CrossRefGoogle Scholar
Roman, GD, Ensor, R, Hughes, C. Does executive function mediate the path from mothers’ depressive symptoms to young children's problem behaviors? J Exp Child Psychol 2016; 142: 158–70.CrossRefGoogle ScholarPubMed
Pearson, RM, Bornstein, MH, Cordero, M, Scerif, G, Mahedy, L, Evans, J, et al. Maternal perinatal mental health and offspring academic achievement at age 16: the mediating role of childhood executive function. J Child Psychol Psychiatry 2016; 57: 491501.CrossRefGoogle ScholarPubMed
Fay-Stammbach, T, Hawes, DJ, Meredith, P. Parenting influences on executive function in early childhood: a review. Child Dev Perspect 2014; 8: 258–64.CrossRefGoogle Scholar
Berry, D, Blair, C, Willoby, M, et al. Household chaos and children's cognitive and socio-emotional development in early childhood: does childcare play a buffering role? Child Res Q 2016; 34: 115–27.CrossRefGoogle ScholarPubMed
Pluess, M, Belsky, J. Vantage sensitivity: individual differences in response to positive experiences. Psychol Bull 2013; 139: 901–16.CrossRefGoogle ScholarPubMed
Belsky, J, Pluess, M. Beyond diathesis stress: differential susceptibility to environmental influences. Psychol Bull 2009; 135: 885908.CrossRefGoogle ScholarPubMed
Park, M, Brain, U, Grunau, RE, Diamond, A, Oberlander, TF. Maternal depression trajectories from pregnancy to 3 years postpartum are associated with children's behavior and executive functions at 3 and 6 years. Arch Womens Ment Health 2018; 21: 353–63.CrossRefGoogle ScholarPubMed
Figure 0

Table 1 Mother–child cohort characteristics

Figure 1

Table 2 Maternal characteristics of subgroup of mothers not treated with an antidepressant during pregnancy

Figure 2

Table 3 Prediction of children's BRIEF scores at 6 years of age from the interaction of prenatal maternal depression (continuous or binary), SSRI exposure and CHAOS score (n = 118)a,b

Figure 3

Fig. 1 Graphic representations of the competitive-confirmatory model and relationships between the different elements affecting child executive functions at 6 years of age.

Relationships include the impact of the multiple effects of selective serotonin reuptake inhibitor (SSRI) treatment and exposure (i.e. improved prenatal and postnatal maternal mood, child brain development), maternal mood (pre- and postnatal) and the role of household chaos. We used this model to test the hypothesis that prenatal depressed maternal mood and SSRI exposure would interact to reflect a differential susceptibility, or even a vantage sensitivity, to household chaos.
Figure 4

Table 4 Confirmatory models (n = 139)a,b

Figure 5

Fig. 2 Confirmatory model with the best fit (differential susceptibility strong model) (refer to Table 4).

Children of prenatally depressed mothers (regardless of SSRI exposure) were not susceptible (solid green line) to household chaos. However, children of non-depressed mothers living with a lower level of household chaos differed in their response to a calmer environment. Namely, children of non-depressed mothers showed better executive functions than children of prenatally depressed mothers, but only where Confusion, Hubbub and Order Scale (CHAOS) scores were 26.5 had significantly higher Behavior Rating Inventory of Executive Function (BRIEF) scores (poorer executive functions compared with children of depressed mothers (right-side vertical bar on horizontal line)).
Supplementary material: File

Dhaliwal et al. supplementary material

Dhaliwal et al. supplementary material

Download Dhaliwal et al. supplementary material(File)
File 7 MB
Submit a response

eLetters

No eLetters have been published for this article.