Concerted changes in social and neuroanatomical factors in adolescence have implications for vulnerability and resilience to anxiety and depression. However, the directionality of associations between specific social, emotional, and neuroanatomical factors has yet to be untangled in adolescence, such as whether aspects of the peer and family environment are longitudinally associated with changes in emotional symptoms and brain structure. To add, differences between sexes in these associations have seldom been considered. Previous studies in the field are limited in (1) the failure to probe specific directional and associative relationships between variables over time in the same model, and (2) the lack of consideration of measurement error (Kline 2016
; Könen & Karbach 2021
). The current study addresses those problems by using latent change score modelling on a large adolescent dataset, to investigate the extent to which peer problems, family support, and socioeconomic stress predict emotional symptoms and the structural development of key brain regions of interest over time.
The social network widens from middle childhood (around 6 to 8 years old), with the focus shifting away from family relationships to peer relationships (Rueger et al. 2016
). Peer relationships are particularly salient at this time, with peer problems and threats to peer group membership predictive of anxiety and depression symptoms (Parr et al. 2020
; Rueger et al. 2016
; Shin et al. 2016
). At the same time, family support remains important during adolescence, and it is reported as the strongest predictor of mental health outcomes (Rothon et al. 2012
; Rueger et al. 2008
). Lack of parental warmth has been associated with greater psychological symptoms in adolescence, again underscoring the importance of positive parental behaviors in adolescent wellbeing (Muris et al. 2003
). Overarching these interpersonal relationships is family socioeconomic status (SES; Bai et al. 2021
), where low SES has been associated with lack of family support (Devenish et al. 2017
). The family stress model explains these associations in terms of increased stress, resulting in lower parental support and reduced involvement with the child, with subsequent increases in child emotional problems (Conger & Conger, 2002
; Conger & Donnellan, 2007
). One remaining question is whether stress from SES affects the child’s relationships outside of the family, for which there is some evidence (Devenish et al., 2017
). In one study, the child’s perceived stress found to mediate the relationship between SES and peer relationships, which again suggests that increased stress extends to other adolescent relationships (Bai et al., 2021
). Thus, consideration of socioeconomic stress, family support and peer relationships together are needed to determine how these aspects of the social environment affect each other, and to establish which factors are the strongest predictors of emotional problems in adolescence.
Together with social factors, the brain undergoes rapid development in adolescence. The dual-systems model of brain development states that affective subcortical regions mature earlier than higher-level frontal regions in adolescence, which has implications for control over socioemotional processes and vulnerability to mental health problems (Blakemore 2008
; Nelson et al., 2016
). Subcortical regions such as the amygdala increase in volume from late childhood to late adolescence (age 16 years) before stabilising in the early 20s (Mills et al., 2014
; Wierenga et al., 2014
). On the other hand, the prefrontal cortex (PFC) decreases in grey matter volume (GMV) from early adolescence into the early 20s (Mills et al., 2014
). This reduction in GMV is attributed to synaptic pruning to increase neural efficiency and refine cognitive control functions (Blakemore, 2008
). This has been demonstrated in a specific region of the PFC known as the ventromedial prefrontal cortex (vmPFC), which has previously been defined as the anterior PFC, including the medial and orbital frontal cortex (mOFC; Hiser & Koenigs, 2018
). Accelerated thinning in the vmPFC has been associated with fewer anxiety and depression symptoms in adolescence (Ducharme et al., 2014
). In contrast, slower growth of the left amygdala was associated with resilience to psychopathology between early and mid-adolescence (Whittle et al., 2013
). When investigated together, maturational coupling of less growth in amygdala volume and greater thinning in the anterior PFC (including vmPFC) was associated with fewer depressive symptoms across adolescence (Vijayakumar et al., 2017
). This shows that the pattern of development within these regions is predictive of mental health problems, and adolescence could reflect a sensitive period that lays the foundation for a person’s social and emotional trajectory throughout their life course (Lamblin et al., 2017
The connection between brain structure and socioemotional experiences is not deterministic; social experiences also shape the structure of the brain. Social network size has been positively associated with GMV in regions involved in emotional and social processing, including the amygdala and vmPFC/mOFC (Noonan et al., 2018
). Social experiences have also been found to predict the developmental trajectory of socioemotional brain regions in adolescence. Increased adolescent social stress was associated with smaller decreases in GMV in prefrontal regions including the vmPFC/mOFC (Tyborowska et al., 2018
). To add, positive parenting—defined as happy, validating and affectionate behavior during a family interaction assessment—has been associated with attenuated growth of the amygdala for boys and accelerated thinning of the mOFC for both sexes (Whittle et al., 2014
). Left mOFC GMV has also been negatively associated with peer problems for both sexes (Kelly et al., 2015
), although this study was cross-sectional and focused on childhood maltreatment. Altogether, it is unclear the degree to which socioeconomic stress, family support, and peer problems predict the structural development of the amygdala and vmPFC/OFC when considered together, and whether there are specific effects due to sex.
As alluded to in the previous point, sex is another factor that influences both neural development and social experiences. In terms of structural brain development, there is evidence that female brains mature faster than males’, with GMV peaking earlier and increasing more rapidly in females compared to males in regions including the amygdala (Goddings et al., 2014
). This may explain neuroimaging findings mentioned previously, such as the relationship between positive parenting and slower growth of the amygdala for males in early adolescence (Whittle et al., 2014
). Further, there are differences in social experiences between sex, with males more likely to interact with peers in larger groups (Rose & Rudolph, 2006
), experience a range of peer victimization (Wang et al., 2010
), and have less friend support compared to females (Rueger et al., 2008
). These differences in social exposure/intimacy between the sexes could present unique opportunities where social experiences affect sensitive periods of brain development.
Additional factors other than peer and family relationships have been associated with both internalizing symptoms and structural brain development. Early stressful life events have been associated with changes in brain volume, emotional symptoms, and social functioning (Gorka et al., 2014
; Hanson et al., 2010
). To add, pubertal status varies between person in adolescence, and has been implicated in brain development (Giedd et al. 2006
) and symptoms of anxiety and depression (Huerta and Brizuela-Gamiño, 2002
). Whole brain volume has also been shown to differ between sexes (Kaczkurkin et al., 2019
), and thus must be included to compare sex differences in regional brain volume. Psychiatric diagnosis also affects social and emotional functioning directly and through stigma (Kaushik et al., 2016
), and recurrent emotional problems has been associated with regional changes in GMV, including the amygdala and frontal lobe (Bora et al., 2012
). To add, the effects of location and recruitment center must be considered in multi-center studies, particularly due to potential variability between MRI scanners (Schumann et al., 2010
). Thus, these variables these must be included in a model to account for confounding effects when assessing the link between social, emotional and neuroanatomical factors.
Untangling the synergistic changes in social, emotional, and neuroanatomical factors in adolescence has implications for vulnerability and resilience to mental health problems such as anxiety and depression. To this end, the longitudinal interplay between peer problems, emotional symptoms, amygdala volume, and vmPFC GMV was examined across adolescence. Family support and socioeconomic stress in early adolescence were also investigated as predictors of change in the above variables of interest. Peer problems and emotional symptoms changed together for both sexes, but it was not the case that one affected the other. There were sex-specific findings: for males only, the vmPFC GMV changed together with peer problems and emotional symptoms, and family support predicted change in amygdala volume. Socioeconomic stress was not a predictor of change in peer problems, emotional symptoms, or regional brain volume for either sex. Exploration of the findings showed that greater total negative life events and higher levels of emotional symptoms predicted change in vmPFC GMV for females. This shows that there may be sex-specific interventions to promote brain development that supports socioemotional functioning.
For the first hypothesis, a directional relationship was not observed between peer problems and emotional symptoms. Rather, there was baseline covariance between values at age 14 years and correlated change between variables between ages 14 and 19 years. The magnitude of the relationship was similar between the sexes, and the relationship persisted even with the addition of covariates in the multivariate model. Previous research has found longitudinal links between peer relationships and internalizing problems across childhood and adolescence (Shin et al. 2016
; Siegel et al., 2009
; van Harmelen et al., 2016
). The current study differed in the use of LCSMs, which simultaneously modelled different types of directional and associative relationships and allowed for a more robust detection of change that accounted for measurement error. Thus, previous research may have found correlated changes between peer problems and emotional symptoms, rather than a direction of the relationship. Indeed, more recent research using LCSMs supports this view; friendship quality did not predict subsequent resilient functioning across adolescence, rather these concepts changed together over time (van Harmelen et al., 2021
). Modelling peer problems and emotional symptoms as latent variables included a range of indicators related to peer integration and victimization, such as having at least one good friend or being bullied. Other studies independently investigated different aspects of peer relationship problems that may have contributed to the differing results (e.g. peer victimization only, Siegel et al., 2009
). Latent variable modelling and measurement invariance tests accounted for measurement error and ensured that the same concept was measured between sex and over time. There was partial measurement invariance for both peer problems and emotional symptoms that resulted in adjustments to the model, so that the latent means could be meaningfully compared between groups. For example, the item related to being bullied was less likely to be affirmed by males at age 19 years compared to age 19 females and both sexes at age 14, but this was not related to the latent variable of “peer problems”. Therefore, previous associations in specific components of peer problems may have been due to sex or age differences. The current study suggests that the core concepts of peer problems and emotional symptoms change together across adolescence.
In terms of the second hypothesis, there were originally some sex-specific findings: peer problems predicted change in vmPFC GMV for females only, and age 14 amygdala volume predicted change in peer problems for males only. However, these effects were not statistically significant when other predictors and covariates were entered into the model. Instead, there were unexpected findings which were found from the bidirectional investigation of relationships inherent to latent change score models. Emotional symptoms at age 14 years predicted change in vmPFC GMV for females, which was driven by the addition of negative life events before and after 14 years into the model, which also both predicted changes in vmPFC GMV. However, the effect of emotional symptoms on changes in vmPFC GMV was not found to be statistically significant when mood or anxiety disorders were controlled for. This may have been because the range of emotional symptoms were restricted; those with a mood or anxiety disorder had higher levels of emotional symptoms, which attenuated the association between emotional symptoms and change in vmPFC GMV. Together, this shows that both the presence of subjective negative life events and emotional distress in early adolescence affect the developmental trajectory of the vmPFC, which appears to trump the impact of peer problems. Associations between early life stress and regional prefrontal volume, including the vmPFC, have been found previously, with some studies suggesting that brain volume mediates the relationship between early life stress and emotional distress (Gorka et al., 2014
; Hanson et al., 2010
). Indeed, there is evidence to suggest that sex differences in the role of stress on the developing brain may be due to the interaction between sex-specific hormones and stress hormones, particularly in the PFC due to its protracted development into adolescence and adulthood (Shaw et al., 2020
). This study shows that targeting the experience of negative life events and emotional distress may be necessary to buffer against the deleterious impact on vmPFC structural development in females.
For the third hypothesis, amygdala and vmPFC GMV did not mediate the association between peer problems and emotional symptoms; prior discussed results did not warrant mediation analyses. Instead, in the covariate-corrected multivariate model, there was correlated change between peer problems, emotional symptoms and vmPFC GMV for males. As with the previous finding of correlated change between peer problems and emotional symptoms, this was significant after accounting for baseline levels, changes within-variable, and changes between variables. Results from the univariate analysis in the current study found that males had a greater variance in change in vmPFC GMV compared to females during this age range. GMV in frontal regions have also been found to peak later and increase less rapidly in males compared to females, and male brain structure has been found to change more during childhood and early adolescence compared to females (Kaczkurkin et al., 2019
; Lenroot et al., 2007
). Taken together, this correlated change between peer problems, emotional symptoms, and vmPFC GMV may reflect a period of concerted change between social relationships, mental health, and frontal brain regions for males during this period rather than a specific direction of effect.
The fourth hypothesis was partially supported; greater parent-reported family support at age 14 years predicted less amygdala volume increase in males only. This remained statistically significant after correcting for WBV and mean PDS score, and when family support was modelled separately as a latent variable. This is in line with previous research which found that higher frequency of positive maternal behaviour predicted attenuated growth in the right amygdala for males between the ages of 12 and 16 years (Whittle et al., 2014
). That previous research only found a significant effect in the right amygdala, however, this study modelled amygdala volume as a latent variable using both left and right amygdala volume as indicators. This suggests that the findings were able to uncover the association between family support and whole amygdala volume when these variables were not obscured by measurement error. There is also previous evidence that the rate of growth of the amygdala is associated with psychopathology; experience of Axis-I DSM-IV psychopathology between early and mid-adolescence has been associated with faster growth of the left amygdala (Whittle et al., 2013
). Thus, the attenuated growth of the amygdala through family support may show a protective effect against psychopathology. At age 14 years, boys may still be reorienting their social focus from family to peers, therefore it may be a time that is sensitive to the secure base of family support (Jenkins et al., 2002
). Girls, on the other hand, may have already completed this social transition in early adolescence, and are further along in neural development, so the same associations are not found as for boys (Kaczkurkin et al., 2019
; Rueger et al., 2008
). Future research should look at whether the patterns observed for males in this study are revealed earlier for females, such as in late childhood.
For the fifth hypothesis, socioeconomic stress did not predict changes in peer problems, emotional symptoms, or amygdala and vmPFC GMV. This was surprising given that previous research has found that low socioeconomic status is associated with less family support (Devenish et al., 2017
) and that greater adolescent social stress was associated with smaller decreases in vmPFC GMV (Tyborowska et al., 2018
). The reason for these results may be that socioeconomic stress was parent-reported, and thus the level of stress that the adolescent feels from the socioeconomic circumstances is unknown. Future research should aim to discern whether the adolescent’s perspective predicts socioemotional and neuroanatomical outcomes.
Strengths and Limitations
The current study used the IMAGEN dataset, which is a rich, longitudinal dataset that contains social, psychological and neurobiological measures in adolescence (Schumann et al., 2010
). The large sample size strengthened the ability to detect robust findings (Kievit et al., 2018
). Additionally, IMAGEN participants were recruited from multiple European countries, which increases the generalizability of the findings. However, it must be highlighted that only participants of European descent were recruited in IMAGEN, which brings into question whether the interplay between peer and family relationships, mental health and neuroanatomy is similar in people of different ethnic backgrounds. Future research should explore this in a more diverse sample.
This analysis employed techniques such as latent change score modelling, which allowed interrogation different parameters of change and the potential direction of effects (Kievit et al., 2018
). Measurement invariance tests were conducted to assess whether the concepts of peer problems and emotional symptoms (both measured by the SDQ) are similar between sex and over time. This has implications for the interpretation of change over time such as whether there is true change or the result of measurement error. This was particularly important because, in the IMAGEN dataset, there were slight differences in the wording of the items for the SDQ versions used at age 14 years (11–17-year-old version) and at age 19 years (17+ years version), for example, 11–17 ‘Other people my age generally like me’
and 17+‘Other people generally like me’
. Measurement invariance tests revealed that the 17+ version was more likely to be affirmed compared to the 11–17 version for both sexes. This brings into question whether this was due to a difference with this age group or whether it was due to the wording of the item, which is broader. In this analysis, the item intercept was adjusted due to noninvariance; previous research has found that not adjusting for non-invariance between groups produced significant bias in regression parameter estimates in SEM (Guenole & Brown, 2014
). This would be a problem for other studies that would want to compare longitudinal scores from adolescent to adult sample. In addition, low coefficient omega values suggested sub-optimal reliability which was different between the different age versions of the SDQ used. This brings into question the suitability of the SDQ items for investigating the concepts of interest and highlights the need to use other measures to reinforce the findings. Previous research has indicated that the SDQ has good psychometric properties from a cross-sectional sample of child and adolescent versions (Goodman, 2001
) and the psychometric properties between adolescent and adult versions have been found to be similar in a cross-sectional sample (Brann et al., 2018
). However, there is limited research into the psychometric performance of the SDQ when measuring change from the adolescent and adult versions. This is important with the proliferation of multi-year population cohort studies that assess changes throughout adolescence and extending into adulthood.
The correlated change observed in the current study can occur due to methodological issues or it may signify the presence of a third variable (Kievit et al. 2018
; Könen & Karbach, 2021
). This study attempted to address this by controlling for baseline scores and by including covariates that may be a common source of variance in the variables (Könen & Karbach, 2021
). The covariate-corrected multivariate model included covariates such as stressful life events and psychiatric diagnosis, which had a minimal impact on the magnitude of the correlated change relationship between peer problems and emotional symptoms. It is possible, however, that variables not included in the study may have contributed to the change in both peer problems and emotional symptoms. Future research could investigate whether other aspects related to peer relationships, such as peer social skills (Nilsen et al., 2013
; Segrin, 2000
), explain the correlated change between peer problems and emotional symptoms for both sexes, and vmPFC GMV for males, during this age range.
The findings suggest that there is concerted change between peer problems and emotional symptoms for both sexes, and this extends to vmPFC GMV for males. Furthermore, greater negative life events and higher levels of emotional symptoms predicted change in vmPFC GMV for females; for males, family support predict change in amygdala volume. This shows that, whilst there are commonalities between sexes, there are differences that may inform on the timing and targets for intervention—enhancing family support for males and protecting against negative life events and emotional symptoms for females. For the findings related to correlated change, it suggests that there is not a simple directional relationship between these variables, rather they are changing in concert. Further research is needed to elucidate whether these changes are being driven by a third variable not included in this study. This would have implications for where to direct potential intervention studies. For example, a social intervention targeted at reducing peer problems and increasing peer integration may show an association with change in emotional symptoms and vmPFC development, but that may have been driven by other factors, such as improving social skills rather than peer integration specifically.
FN is part of the IMAGEN Consortium.
The full IMAGEN Consortium author list is as follows:
Tobias Banaschewski4, Arun L. W. Bokde7, Sylvane Desrivières8, Herta Flor5,9, Antoine Grigis10, Hugh Garavan11, Penny Gowland12, Andreas Heinz13, Rüdiger Brühl14, Jean-Luc Martinot15, Marie-Laure Paillère Martinot16,17, Eric Artiges18, Dimitri Papadopoulos Orfanos10, Tomáš Paus19, 20, Luise Poustka21, Sarah Hohmann4, Sabina Millenet4, Juliane H. Fröhner22, Michael N. Smolka22, Nilakshi Vaidya23, Henrik Walter13, Robert Whelan23, Gunter Schumann23,24
JS conceived of the study, participated in its design and coordination, performed the statistical analysis, interpreted the data, and drafted the manuscript; LM participated in the design of the study and interpretation of the data, and helped to draft the manuscript; PQ participated in the design of the study and interpretation of the data, and helped to draft the manuscript; RE participated in the design of the study and interpretation of the data, and helped to draft the manuscript; FN participated in its design of the study and helped to draft the manuscript. All authors read and approved the final manuscript.
JS is funded by the ESRC-BBSRC Soc-B Centre for Doctoral Training Programme (ES/P000347/1). LM is supported by the National Institute for Health and Care Research Applied Research Collaboration Greater Manchester (NIHR ARC-GM; funding reference: NIHR200174). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. The IMAGEN Consortium is funded by the following sources: the European Union-funded FP6 Integrated Project IMAGEN (Reinforcement-related behaviour in normal brain function and psychopathology) (LSHM-CT- 2007-037286), the Horizon 2020 funded ERC Advanced Grant ‘STRATIFY’ (Brain network based stratification of reinforcement-related disorders) (695313), Human Brain Project (HBP SGA 2, 785907, and HBP SGA 3, 945539), the Medical Research Council Grant ‘c-VEDA’ (Consortium on Vulnerability to Externalizing Disorders and Addictions) (MR/N000390/1), the National Institute of Health (NIH) (R01DA049238, A decentralized macro and micro gene-by-environment interaction analysis of substance use behavior and its brain biomarkers), the National Institute for Health Research (NIHR) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London, the Bundesministeriumfür Bildung und Forschung (BMBF grants 01GS08152; 01EV0711; Forschungsnetz AERIAL 01EE1406A, 01EE1406B; Forschungsnetz IMAC-Mind 01GL1745B), the Deutsche Forschungsgemeinschaft (DFG grants SM 80/7-2, SFB 940, TRR 265, NE 1383/14-1), the Medical Research Foundation and Medical Research Council (grants MR/R00465X/1 and MR/S020306/1), the National Institutes of Health (NIH) funded ENIGMA (grants 5U54EB020403-05 and 1R56AG058854-01), NSFC grant 82150710554 and European Union funded project ‘environMENTAL’, grant no: 101057429. Further support was provided by grants from: - the ANR (ANR-12-SAMA-0004, AAPG2019 - GeBra), the Eranet Neuron (AF12-NEUR0008-01 -WM2NA; and ANR-18-NEUR00002-01 - ADORe), the Fondation de France (00081242), the Fondation pour la Recherche Médicale (DPA20140629802), the Mission Interministérielle de Lutte-contre-les-Drogues-et-les-Conduites-Addictives (MILDECA), the Assistance-Publique-Hôpitaux-de-Paris and INSERM (interface grant), Paris Sud University IDEX 2012, the Fondation de l’Avenir (grant AP-RM-17-013), the Fédération pour la Recherche sur le Cerveau; the National Institutes of Health, Science Foundation Ireland (16/ERCD/3797), U.S.A. (Axon, Testosterone and Mental Health during Adolescence; RO1 MH085772-01A1) and by NIH Consortium grant U54 EB020403, supported by a cross-NIH alliance that funds Big Data to Knowledge Centres of Excellence.
Data Sharing and Declaration
The data that support the findings of this study are available from the IMAGEN project. Restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. To access the IMAGEN dataset, researchers must submit a study proposal form to the IMAGEN Executive Committee (see https://imagen-project.org/the-imagen-dataset/
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.