The interaction between individuals and their environment is a dynamic process, that occurs at multiple levels, including macro and micro-environment. While macroenvironment encompasses broad factors at neighbourhood level, the microenvironment refers to immediate surroundings and contexts in which individuals live their lives. Throughout life, individuals are exposed to multiple adverse events, within their microenvironment, that may create a cumulative burden of adversity, known as allostatic load [1, 2]. When faced with adversity physiological regulatory systems, such as the hypothalamic-pituitary-adrenal (HPA) axis, the autonomic nervous system (ANS), the metabolic system as well as the immune system, produce physiological response. While these responses are adaptive in the short term, repeated exposure to adverse events or chronic stress can lead to long-lasting alterations in these systems. This results in a so-called wear-and-tear or allostatic load which is known to significantly contribute to the emergence and maintenance of mental and physical illnesses [3]. According to its systemic cascades, the brain represents one of the critical targets for allostatic load. For example: chronic release of glucocorticoids or inflammatory cytokines can lead to changes in the brain [1, 2].

The human brain is well known to exhibit plasticity, an intrinsic ability to reorganise its structure and function throughout the lifespan [4]. This allows the brain to adapt to changes in the external environment or internal milieu. With respect to external environmental influences, like the experience of adverse events, the brain plays a central role in the processes of allostasis. Allostatic changes can lead to both successful adaptation and the development of resilience as well as dysfunctional behaviours and the emergence or maintenance of disorders [1,2,3]. In the latter case, the brain becomes vulnerable to dysregulation, leading to alterations in response to prolonged or severe adverse events.

Adversity can manifest in various forms including psychosocial factors such as childhood maltreatment, peer victimisation, social isolation, affective loss, domestic conflict, or poverty, as well as exposure to environmental toxins. Despite the shared aspect of stress biology underlying these adversities, they may still engage both common and distinct mechanistic pathways. The current understanding of these adverse events, however, is limited to a largely unidimensional perspective, with individual studies often focusing on isolated events [5,6,7,8,9]. While the nature of the impact is interconnected, the empirical evidence regarding the neurobehavioral effects from combination of different adversity types remains limited. It is crucial to move beyond the study of isolated events and investigate the cumulative/interactive effect or allostatic load, resulting from exposure to multiple adverse events, both simultaneously and successively, i.e., mirroring the complex and interconnected nature of real-life situations.

To gain insights into the effects of multiple adverse events, we can leverage data from existing large cohorts. In recent years, several large, well-phenotyped cohorts have emerged like IMAGEN [10], ABCD [11], cVEDA [12, 13], CHIMGEN [14], Generation R [15], ALSPAC [16, 17], UK Biobank [18], etc. Although they have been initiated under partly different research foci, they all encompass comprehensive information on the experiences of various adverse life events. These cohorts offer an opportunity to address the power limitations often encountered in smaller-scale studies and together with their multi-modal assessment batteries facilitate the examination of the combined effects of multiple exposures. By employing data-driven analyses in these large datasets, we can gain a deeper understanding of the intricate neurobiological effects of allostatic load. Finally, by encompassing diverse populations across cohorts, population neuroscience research can inform global precision psychiatry.

The objective of this review is to provide a comprehensive understanding of changes in brain structure and function following experiences of adverse life events. We have included those adversities that have been extensively investigated in the field of neuroscience. The choice was driven by both the prevalence of these experiences and their documented impact on brain across diverse populations. In the future perspective, we advocate for data-driven approaches to understand the cumulative impact of adversities, that can extend beyond the confines of selected categories. The exposures we targeted included childhood maltreatment, peer victimisation, loneliness, affective loss, domestic conflict, poverty, and toxins. Using title/abstract text words we combined these exposures with neuroimaging (MeSH) and human filter. The search was carried out in PubMed, covering the period from January 1, 2010, to April 8, 2023. The reference list of relevant systematic reviews identified in our structured search were hand-searched for relevant literature. In case of recent reviews/meta-analysis, direct citations were included in the manuscript. We exclusively reviewed cohort studies with participants recruited from the general population, deliberately excluding clinical samples, giving examples from cohorts with larger sample sizes (n > 300) and/or longitudinal designs, where possible. We highlight the extensive research conducted on various stressors at the individual level, summarising all identified papers in tables and representative examples in text; and emphasise the necessity of data-driven, multimodal approaches to better understand the complex relationships between multiple life events and brain. Lastly, we provide an outlook presenting an overview of various statistical approaches that can be employed in future studies.

Childhood maltreatment

Current evidence suggests that negative experiences in the form of abuse and neglect during sensitive developmental periods can result in neuroplastic processes, disrupting normal brain functioning [5]. Studies among children, adolescents and adults with adverse childhood experiences all have reported detrimental effect of such experiences in multiple brain regions (see Table 1 for details). Specifically, the amygdala, the hippocampus, and the medial pre-frontal cortex (mPFC) are implicated, possibly due to the presence of dense glucocorticoid receptors and the timing for pathway development during childhood [1, 19, 20]. Decreased amygdala and mPFC volumes were linked to childhood abuse and decreased dorsolateral prefrontal (dlPFC) volume to childhood neglect [21,22,23,24]. Decreased hippocampal volume was linked to both childhood abuse as well as neglect [21, 25, 26]. In a study investigating the impact of childhood sexual abuse on brain development [27], sexual abuse was associated with lower hippocampal volume during childhood (3–5 and 11–13 years), but with lower frontal cortex volume during adolescence (14–16 years). The impact of childhood maltreatment may therefore depend on the type of adversity, in combination with a variation across different stages of development, potentially influencing different cognitive and emotional processes.

Table 1 Neuroimaging studies of childhood maltreatment.

Neuroimaging studies at the functional level (fMRI) have mainly reported alterations in the amygdala, but findings are inconsistent [21]. One study on self-reported exposure to adversity suggests that the distinct pattern of amygdala activation depends on the timing of exposure during sensitive periods of development [28]. In this study, adversity reported during early childhood (3–6 years) was associated with blunted amygdala response, in contrast, exposure to adversity during early adolescence (13–15 years) was associated with an augmented amygdala response. In another study, differences within specific areas of amygdala, were attributed to the type of adversity [22]. Childhood abuse was linked with heightened reactivity in the ventral region of the amygdala, whereas experiences of neglect to heightened reactivity in the dorsal region of the amygdala.

Further, studies focussing on the frontal-limbic pathways also show converging effects on the amygdala [19, 29], reporting accelerated amygdala and mPFC connectivity in response to early adversity [30, 31]. These changes support behavioural research for increased threat processing i.e., inability to differentiate between safe and threatful stimuli, as form of adaptation [32] and provide backing to the stress acceleration hypothesis [33], i.e., early adversity expediting the development of the emotional regulation neural pathways. fMRI studies on emotional regulation following childhood maltreatment, also report alterations in the connectivity and activity of neural circuits in the frontal-limbic regions, more specifically in the amygdala and ventral anterior cingulate cortex (ACC) [34]. Differences in impact due to adversity type was again reported, in a longitudinal study, where childhood abuse was associated with increased amygdala activity while childhood neglect with decreased ventral striatum response to happy faces, in adolescents [35]. Other frontal-limbic regions beyond the amygdala that play a crucial role in the cognitive modulation of emotions like the dlPFC [36], and, in the automatic regulation of stress hormones like the hippocampus [37], are also implicated. However, the exact nature and direction of these functional changes have varied across studies [38].

In fMRI studies on reward processing, higher activation in the striatum among adolescents who have experienced childhood adversity is consistently reported [38, 39]. Additionally, adults who have experienced childhood emotional abuse exhibit reduced grey matter volume (GMV) in the ventral striatum [40]. Overall, a reduced anticipatory response to rewards is observed [29], which may represent an adaptive regulation towards avoidant responses during approach-avoidance conflict situations. This adaptive regulation is believed to increase the likelihood of survival in adverse environments. However, it can also hinder exploratory behaviour making it difficult to identify sources of reward in new environments [38, 41].

Other consistent findings include the effects on the pituitary gland, with longitudinal studies reporting accelerated development in response to childhood neglect [42, 43]. Lower superior parietal volume [21] and hyperactive superior temporal gyrus [44], linked to social perception and social cognition respectively, were also consistently reported. According to Nelson et al. [45], social information travels from the detection node to the affective node and then to the cognitive-regulation node. The impairments observed in all these networks, together reflect the complex interplay between early adversity, brain development, and resulting neurobiological changes.

Institutionalised care

Children fostered from institutions experience lack of individual attention and have minimal opportunities for cognitive and social stimulation during crucial early years of development, when neural systems are highly plastic. Studies examining the effects of institutionalised care have found that these children exhibit smaller total GMV, reduced cortical thickness, impaired executive function skills, and atypical reward processing [46]. The availability of information on the known duration of institutionalised care provides a unique opportunity for investigating the precise relationship between the duration of adversity and its impact on neural outcomes. Here, we specifically review studies investigating the duration of institutionalised care (details in Table 1) to gain insights into the effects of temporally discrete adversity.

The duration of institutionalised care, as measured by the age at adoption, has been linked to both structural and functional brain alterations. Structural analyses have exhibited a dose-related relationship, where longer periods of time spent in foster care was associated with reduced hippocampus [47] and larger amygdala volumes [48]. Functional amygdala findings also showed an age-at-adoption relationship. Specifically, younger age-at-adoption was linked to more typical differentiation between mother and stranger stimuli, while older age-at-adoption was associated with reduced discrimination [49]. This struggle to differentiate between safe and dangerous stimuli, is reflective of an increased threat processing or fear generalisation. Further, institutionalised children exhibit altered risk-taking behaviour, with later adoptees making fewer risky decisions [50]. Prolonged institutionalisation is also associated with the potential for learning and adaptation in enriched environments post adaptation. In a notable study, children adopted after 5 years of institutionalisation failed to improve their performance in response to reward in a modified monetary incentive delay task at age 12, while those adapted by 2 years on average did improve [51].

These findings underscore the importance of capturing the duration of adversity. Researchers studying the impact of childhood adversity have predominantly relied on questionnaires [such as the Childhood Trauma Questionnaire [52]] to assess various forms of adversity experienced during the first 18 years of life. While these questionnaires provide valuable information, there is a growing recognition of the need to incorporate more specific details regarding developmental timing. By understanding the timing of exposure, researchers can explore how specific developmental stages may shape the impact of adversities on various outcomes, such as cognitive, emotional, and social functioning. Additionally, it is also important to consider the duration of exposure. Adversities that persist over a prolonged period can have distinct effects on individuals' development and well-being compared to isolated experiences. Including measures of duration in data collection will allow for a more nuanced understanding of allostatic load.

Peer victimisation

Peer victimisation associated with bullying refers to persistent and repeated instances of aggression or intimidation, which can include verbal or physical assaults, social exclusion or peer rejection, name-calling, and threats [53] and is linked with alterations in several brain regions (see Table 2 for details). Structural findings in large cohort of children who were frequently bullied, calculated based on separate reports by parents and teachers, exhibit greater cortical thickness of the fusiform gyrus [54], a brain region known for its involvement in facial processing. Structural difference in the striatum has also been consistently reported [7]. In a longitudinal study, adolescents who experienced chronic victimisation and had larger putamen (dorsal striatum) volume at age 14 and showed a more rapid decrease in putamen volume compared to their counterparts who experienced less victimisation [55]. Putamen is associated with regulation of risky behaviour and processing of rewards.

Table 2 Neuroimaging studies of social pain.

fMRI studies on these behaviours (risk-taking and reward processing), have had mixed results, as reported in a recent systematic review [7]. However, an interesting pattern was observed in relation with rejection sensitivity or need for belonging. A higher likelihood of engaging in risky behaviour is commonly observed during adolescence, especially in social situations involving peers. For example: in a driving stimulation task, an increased activation in the ventral striatum and OFC associated with risk-taking was observed in the presence of peers [56]. Thus, when peer evaluation or the need for belonging is present, there is a heightened sensitivity to the potential reward value of risky decision. In another study, female adolescents with a history of higher exposure to peer victimisation showed greater activation in the amygdala, ventral striatum, fusiform gyrus, and temporoparietal junction in response to in-group rather than out-group peers, indicating greater anticipation of reward and outcome value towards in-group peers [57]. These findings suggest that higher risk-taking behaviour could be a response to avoid peer rejection and thus highlight the need to study overt (physical) and covert (relational) victimisation separately.

Covert victimisation is linked to social pain, i.e., emotional distress or discomfort experienced due to negative social experiences, including rejection, exclusion, or betrayal. Social pain reported during exclusion task paradigms, is consistently associated with increased activation in ACC and insula [58]. Further, studies investigating neural correlates of social exclusion have consistently observed increased neural activity in regions associated with the processing of emotions, such as the amygdala, dorsolateral ACC, and inferior fusiform gyrus, in individuals who have experienced peer victimisation compared to those who have not [7]. Thus, peer victimisation might exacerbate the emotional pain experienced during social exclusion.

Social isolation/loneliness

Loneliness, a complex socio-emotional trait is a strong predictor of mental illness. Loneliness has been associated with larger GMV of the dlPFC and smaller volume of amygdala, anterior hippocampus, posterior para-hippocampus, and cerebellum [59,60,61], interestingly in the left hemisphere for all aforementioned regions (details in Table 2). Further, loneliness is strongly correlated with perceived social support. A longitudinal investigation revealed that individuals who reported higher levels of perceived social support experienced less decline in total brain volume as opposed to those with suboptimal perceived social support [62]. Of note, in a large cohort study of middle-aged adults, enlarged ventricles were also associated with loneliness [61]. Thus, loneliness could result in steeper cognitive decline.

In resting-state data-driven analysis, loneliness was associated with increased functional connectivity (FC) of the default mode network, frontoparietal network and attention and perceptual networks [63, 64]. Hypervigilance and stress reactivity, which are believed to be associated with loneliness, may be connected to these networks. According to the loneliness model, social isolation leads to unconscious surveillance for social threats, i.e., implicit hypervigilance [65, 66] and is paired with attentional bias and confirmatory behaviour. Attentional bias is characterised by a heightened focus on negative social cues, which can contribute to feelings of rejection [67]. Confirmatory behaviour involves engaging in inappropriate social and withdrawal behaviours, which can elicit negative reactions and reinforce the initial negative beliefs about interpersonal interactions [68]. Together, this causes lonely individuals to perceive the social world as threatening and display negative social behaviour and affect [69]. This is further supported by a large cohort study [70] which showed increased FC between inferior frontal gyrus (IFG) with superior parietal lobule, precentral gyrus and supplementary motor area. Heightened inferior frontal connectivity is associated with selective attention and social cognition [71, 72], superior parietal lobule is associated with working memory, attention and visuospatial perception [73, 74], and precentral gyrus is associated with pain appraisal [75]. Thus, lonely individuals actively observe their environment, are cautious about negative cues with heightened threat appraisal.

Affective loss

Separation, ending a significant relationship or death of loved one is categorised as affective loss. While transient subclinical symptoms of anxiety and depression might be present initially, many studies have demonstrated long-term clinical effects as well, referred to as prolonged grief disorder (PGD) [6]. We did not find any large cohort studies investigating FC associated with affective loss and have reviewed smaller studies to gather information on this topic. fMRI studies on bereavement of a first-degree relative [76, 77] or spouse [78], the loss of an unborn child [79] or only child [80], the breakup of a relationship [81] and the loss of a pet [82] have reported altered neural activations in the networks of pain and emotions, including the cingulate, amygdala, hippocampus, and OFC (details in Table 2). Functional studies of amygdala have reported heightened activation with dlPFC linked to attentional bias and with rostral ACC linked to emotional dysregulation [82]. The posterior cingulate cortex involved in autobiographic memory also shows heightened activation during grief [6]. Finally, repetitive thinking involving OFC was also observed [76].

However, the statistical power of reported studies has been limited due to small sample sizes which could potentially cause problems with replicability [83] and type I errors, i.e., risk of obtaining false positive results [84]. For example: while a small study (n = 32) reported increased GMV in cerebellum [85], results were not replicated in a larger (n = 192) sample [86]. Studies on structural MRI though few, on the other hand have recruited larger sample (n ~ 190) [86, 87] or are nested in cohort: Rotterdam cohort (5501 participants) [88]. These studies associate affective loss with smaller GMV (246), specifically in the amygdala [87].

Domestic conflicts

Exposure to domestic conflicts, whether as a witness or a victim of abuse, is highly distressing and potentially traumatic. Maternal reports of higher interparental conflict have been linked with increased neural responses in infants to highly angry speech compared to neutral speech. This hyperactivation was observed in several brain regions involved in emotional processing and stress reactivity, such as the rostral ACC, caudate, thalamus, and hypothalamus [89]. Interparental conflict may have an impact on early emotional development due to poor caregiving and via direct exposure to aggressive interactions between caregivers, leading to challenges in the emotional regulation within this dyadic relationship.

Victims of intimate partner violence (IPV) are known to adapt their emotional regulation strategies to manage recurring stressful events, thus affecting their underlying brain connectivity. To get an overview of the brain correlates of IPV, we again looked at smaller population-based studies (Table 2 for details). Structural network connectivity study shows altered connectivity in the victim group, in regions involved in cognitive-emotional control. Specifically, the caudal ACC, middle temporal gyrus, left amygdala, and ventral diencephalon (including the thalamus) were implicated in these alterations [90]. A study on laboratory-based and real-world intimate partner aggression (IPA) showed distinct patterns in males and females. Specifically, men’s IPA was associated with reduced reactivity in the posterior cingulate during provocation, while women’s IPA was associated with decreased activity in the ventromedial prefrontal cortex during aggressive event [91]. However, due to the cross-sectional nature of these studies, it is not possible to establish a causal pathway.


Socioeconomic status (SES) as a comprehensive measure includes various aspects of sustained experience, incorporating both objective factors such as education and income, as well as subjective factors such as social standing. SES influences an individual’s environment throughout their lifespan and is associated with experience-based neural plasticity [4]. The environmental stimulation, or lack thereof in impoverished conditions, has a potential effect on the brain. Brain structure and functional connectivity are associated with SES. However, these associations do not converge on specific regions or networks and are more widespread (which could also be a result of varied assessments). Whole brain structural studies reporting positive association between SES and global brain measures including cortical thickness, cortical surface area and GMV [92,93,94,95] reflect the global effects of SES. Lower SES is further associated with lower GMV in specific regions such as the amygdala, hippocampus, striatum, thalamus, cingulate cortex, occipital cortex, and frontal and temporal lobes [9, 96, 97]. These findings are replicated in large [92, 98], longitudinal [93, 95, 99] cohorts and are independent of genetic architecture [94] (details in Table 3). Taken together, these regions are responsible for language processing, reading skills, visuo-spatial abilities, decision-making and executive functioning.

Table 3 Neuroimaging studies of poverty.

From studies on children and adolescents, one can reliably state that socio-economic disadvantage is linked to changes in overall cognitive development. Children living in poverty have a higher likelihood of encountering developmental delays, lower performance on cognitive and academic assessments, and an increased occurrence of behavioural and emotional issues compared to their more privileged counterparts [96, 97]. In one study, children living 1.5 times below the federal poverty line in the US had regional GMV that were 3–4 points lower than the developmental norm, defined by index of structural brain development based on full longitudinal study sample. Further, these lower volumes mediated the association between low-income status and scoring 4–7 points lower on standardised tests - the Wechsler Abbreviated Scale of Intelligence and Woodcock-Johnson III Tests of Achievement [92]. Another study reported that brain variations can be detected as early as 2.5–6.5 years [99]. In the temporal, parietal, and occipital lobes grey matter differences between children from low SES and those from higher SES were observed as early as 2.5 years of age, and these disparities persisted throughout the studied age range up to 20 years. While differences in GMV in the frontal lobe emerged later in development, around 6.5 years of age, and continued to intensify until 20 years.

Among adults, associations have been reported with childhood SES [100, 101] as well as current SES [102, 103]. Lower family income at age 9 was associated with reduced activity in the vlPFC and dlPFC and failure to suppress amygdala activation while regulating negative emotions in 24-year-olds [101]. Both the dlPFC and vlPFC play crucial roles in cognitive control and executive functioning, supporting the regulation of goal-directed behaviours. Decreased activity in these regions along with amygdala has been linked to disruptions in down-regulation of negative emotions [101]. Current SES in middle aged adults was associated with reduced hippocampus and amygdala volumes [102] and thinner average cortical grey matter [103] after controlling for childhood SES. Further, current SES moderated the association between age and brain system segregation [103]. Middle aged adults with lower SES showed decreased system segregation compared to those with higher SES. Taken together, these findings underscore the importance of SES as an important factor that influences individuals across the entire lifespan.

In fMRI studies lower SES has been linked to reduced activation of hippocampus and amygdala in resting-state fMRI [9]. In task-based fMRI, with working memory paradigm, reduced activation of frontal and temporal regions linked with cognitive functioning are observed for lower SES [9, 104]. Further, lower SES is associated with higher threat reactivity and risk aversive decision making linked to increased amygdala and mPFC, respectively [105]. In a longitudinal study on response inhibition, lower SES was linked to higher activation of ACC [106]. Finally, lower SES is associated with hypoactivation of the executive network and hyperactivation of the reward network [104].

fMRI studies on reward processing have also associated low SES with widespread neural correlates. Alterations in reward processing is linked to caudate/striatum and OFC and parietal cortex in MID task; the dorsomedial frontal, subgenual ACC, dlPFC, and parietal cortices in gambling task and the dorsomedial frontal cortex in guessing task [104]. Individuals who experience material deprivation may face challenges in optimising their rewards, as the costs associated with delayed rewards and missed opportunities are disproportionately higher for those with limited available resources [107].

Overall, a multitude of neural associations are found across various brain regions. A recent meta-analysis of structural and functional studies also supported this widespread associations [104]. SES is a multi-dimensional construct which is nested in an ecological system. Therefore, its effects must be comprehensively understood within the psychosocial context of the population being studied. Individuals from low SES experience a range of unfavourable psychosocial and physical conditions that occur together and are interrelated [108]. These conditions are often suboptimal and therefore underscore the impact of poverty. For example: A person’s exposure to familial violence and crime incidence in their neighbourhood is negatively correlated with household income. Similarly, social class is correlated with contact with aggressive peers [108, 109]. Further, low-income families may live in communities that have higher physical adversities. In the second article of this series (Polemiti et al. [110]), the review focuses on the physical environment at the community level, while the impact of neurotoxins at the individual level is discussed below. Overall, exposure to violence, inadequate cognitive stimulation or social support, and a range of other hindrances and discomforts make it challenging to establish a single and straightforward explanation for the associations with poverty. Therefore, it is possible that the widespread effects observed across various brain regions in different studies may be capturing complex underlying interactions among these factors and could be disentangled in future studies.


While neurotoxins are not strictly part of the microenvironment, they are environmental factors that individuals may be exposed to within their immediate surroundings. Humans are continuously exposed to a wide range of neurotoxins [111], including heavy metals or metalloids (e.g., arsenic, lead) [112,113,114] as well as man-made chemicals (e.g., polychlorinated biphenyls (PCBs), phthalates) [111, 115]. The timing and dosage of such exposure plays a crucial role and is associated with more pronounced effects [111]. Among all studies (details in Table 4), changes in GMV have been reported. This included changes as a consequence of arsenic exposure mostly through food or contaminated water [116], which is high in particular regions of the world, including South America, the United States and central Asia, but also due to exposure to lead, which is often used in various products, including pipes or gasoline [117], and finally due to pesticides, such as organophosphates, which are used to protect harvest or preserve food throughout the world and are highly toxic, especially to infants and children [111].

Table 4 Neuroimaging studies of toxins.

Higher arsenic exposure was associated with higher GMV in the IFG and lower GMV in the right inferior temporal cortex, right rostral ACC, and left insula [118]. Higher lead exposure in children was associated with smaller overall cortical volume and surface [119], particularly in the corpus callosum [120]. This is in line with studies on adults with high lead exposure in childhood, showing smaller cortical surface area and smaller hippocampal volume [121], and current exposure associated with smaller GMV in the cingulate gyri, insula, and corpus callosum [122].

Finally, alterations in functional activity and connectivity, including the insula, ACC, and hippocampus, could be observed for arsenic exposure [118]. For pesticides exposure lower FC in the left IFG and bilateral superior frontal gyrus was observed, during a behavioural inhibition task [123]. Aside the fact that more studies are needed, the existing ones provide similar evidence as also found for the effect of psychosocial adverse events, namely a complex interaction of exposure type, and age, on brain and behaviour.


Each form of adversity leaves its mark on the brain, affecting multiple regions. Across various forms of adversity, some common areas emerge (Fig. 1). This suggests that repeated and/or simultaneous occurrence of adverse life events may exacerbate allostasis, resulting in a cumulative impact on the neurobiology of specific brain regions. The exact nature of this cumulative impact however remains elusive.

Fig. 1: Allostatic load.
figure 1

This schematic illustration depicts the interconnectedness between the impact of various adversities on selected brain regions. Although each adversity may have distinct manifestations, they converge on common brain regions. Understanding the cumulative effects of these adversities on the brain can provide valuable insights into allostatic load.

The amygdala, hippocampus, prefrontal cortex and ACC are among the key regions that consistently show alterations in response to different adversities. These regions play crucial roles in emotional regulation, memory, and decision making, and their dysregulation has been linked to poor mental health outcomes [124, 125]. For example, the mechanisms of emotions, memory, and cognitive appraisal are interconnected in the amygdala and hippocampus, spanning from perception to reasoning. The amygdala-hippocampus is associated with two distinct memory systems, which interact with each other in emotional contexts. More specifically, the amygdala influences hippocampal-dependent memories, particularly episodic memory related to emotional stimuli. Conversely, the hippocampus can impact the response of the amygdala when encountering emotional stimuli. Thus, the amygdala and hippocampus modulate emotional memory processes, demonstrating their intertwined role in cognitive and emotional functioning [126]. Individuals experiencing alterations in these systems due to adversity may struggle with suppressing irrelevant aversive information [32, 65], impacting their emotional reactivity. Consequently, when faced with another adverse event, their response to the effects of such experiences may be heightened.

Cumulative effects

The effects of various adversities are mostly studied in isolation, despite the interconnected nature of their impact. Investigations have examined either a single or limited number of exposures [127, 128], often using simple sum scores to access cumulative effects [129,130,131]. To unravel the biological underpinning of the combined effects of multiple adversities and understand if these effects are cumulative and/or synergistic, data-driven approaches can be explored. Recently, there have been some emerging but fragmented attempts of using data-driven approaches including machine learning [132], factor analysis [133] and clustering [131]. While on the one hand, these studies provide proof-of-concept for the benefits of employing a comprehensive and entirely data-driven approach to unravel the complex associations between diverse adversities and neurobiology. On the other hand, in 2 of these 3 studies, adversity sum score was taken.

Adversity itself exhibits considerable heterogeneity, further complicated by co-occurrence and chronicity. It is also important to acknowledge the heterogeneity underlying neuronal profiles, as different adversities will have common as well as varying effects on different individuals. Data-driven methodologies can aid in the development of conceptual models to achieve a comprehensive understanding of these intricacies. To achieve this, we can integrate profiles from multiple markers and across different modalities to characterise multivariate profiles of adversity. In practice, there may exist multiple partially overlapping risk profiles that operate differently in different individuals.

Risk and resilience

Regions that are susceptible to the effects of adversity might experience exacerbated impairment with each hit, occurring simultaneously or across lifespan. This suggests a compounding effect, wherein the negative impact on neural structure and functioning is amplified with each subsequent exposure. For example, the presence of two or more early-life events worsened the age-related decline in hippocampus and amygdala volume [130]. In parallel, there is a growing body of literature on adaption-based approach to resilience (or hidden talents) that highlights the presence of intact or even enhanced social, cognitive, and affective skills among individuals who have experienced high levels of adversity [134]. These alternative perspectives emphasise the adaptive nature of certain phenotypes like attention, perception, learning, memory, and problem solving that emerge following adversity. They acknowledge that individuals who have experienced adversity may develop unique traits, skills, or coping mechanisms that can be beneficial in navigating challenging situations [135, 136]. For example, enhanced amygdala reactivity resulting from early adversity has been associated with improved goal-directed behaviour in situations where the goal aligns with threat-detection [137].

The impact of adversity can thus lead to distinct outcomes, with some individuals being categorised as at-risk and others as stress-adapted [136]. These observations underscore the heterogeneity in the impact of adversity, emphasising the need to shift from group-level inferences to individual-level predictions. By recognising the diverse and individualised responses to adversity, we can better understand the complex interplay between environmental exposures, neural mechanisms, and mental health outcomes.


Our review focused on investigating the effects of adversity on the brain, through an examination of population-based studies. Through this approach, we have summarised impairments in various brain regions and networks that were consistently associated with adversity. These neurobiological changes may have significant implications for psychiatry, potentially increasing an individual’s vulnerability to developing mental health disorders [138, 139]. For example: childhood adversity is associated with the onset of over 40% of childhood psychiatric disorders and more than 25% of adult psychiatric disorders [138]. Additionally, the observed variability and severity of symptoms or progression of the disease and comorbidity could potentially be attributed to brain changes from prior experiences of adversity. This has been reported for bipolar disorder [140], depression [141], conduct disorder [142], obsessive-compulsive disorder [143] and substance use [144]. Longitudinal studies also support these causal effects of adversity on psychopathology via brain changes. Blunted activation in the right amygdala associated with childhood adversity mediated its link with later externalising symptomatology [145]. In another study, changes in resting-state functional connectivity associated with childhood maltreatment [146] mediated the relationship with depression. And poverty was found to be associated with changes in hippocampal-amygdala connectivity, which also led to negative mood symptoms [147]. These findings highlight the implications of neurobiological changes resulting from adversity in psychopathology. By utilising multivariate predictive machine learning techniques, researchers can extend their investigations to make predictions regarding the initiation, progression, and outcomes of various illnesses. This approach holds promise for advancing precision medicine and offering valuable insights into tailored prevention and intervention approaches at the individual level.

Future directions

Most studies we reviewed have principally identified group level associations with biological markers, generally in isolation from one another. While this provides an important first step in understanding the effects of adverse life events on the brain, we identify several key goals that we consider are necessary for the field to move towards a comprehensive understanding of real-life environmental impact, which can all leverage advances made possible by the increasing availability of big data cohorts.

Unveiling multifaceted insights

To adequately capture and analyse higher-order interactions of highly collinear factors, it is essential to have large sample sizes to achieve sufficient statistical power [148, 149]. Large datasets enable the detection of subtle effects that may not reach statistical significance in smaller samples. Further, they ensure that the effects being investigated are robust and reproducible, allowing for more accurate and meaningful conclusions to be drawn from the data [148, 149]. We have specifically reviewed large samples, where possible, however most of these studies have used a traditional research approach, primarily providing descriptive findings. These studies have either examined associations between specific variables or have compared groups exposed to adversity with groups not exposed. While these approaches have provided foundational knowledge, there is a need to move beyond group-level effects towards understanding individual-level differences.

Machine learning methodologies offer a pathway for this transition from univariate associations to multivariate predictions. The utilisation of machine learning approaches to make predictions and separate groups is widespread in biological psychiatry [150] and such multivariate approaches also hold promise for integrating and understanding the cumulative effects of multiple adversities, more effectively than simple sum score methods. These methods are specifically designed to analyse multidimensional data, allowing researchers to uncover multivariate patterns that may not be readily apparent using traditional sum-score approaches. Data-driven approaches including supervised machine learning approaches [151, 152] and multivariate regression methods [153] such as partial least squares [118, 154] are emerging, and their potential is only beginning to be explored [155]. These techniques offer valuable tools for identifying latent variables within complex datasets, categorising individuals based on their unique profiles, integrating diverse sources of data, and finally facilitating prediction.

Another key goal for the field to move forward is developing optimal approaches for parsing heterogeneity, for which many approaches have been proposed in the field of psychiatry [156]. Parsing heterogeneity refers to the process of systematically analysing distinct sources of variability within a given dataset. By dissecting and categorising different sources of variability, researchers can gain a more intricate understanding of how various factors contribute to the overall outcomes observed. To do so, heterogeneity can be parsed at the variable level (i.e., to yield symptom groups or latent profiles) or at the individual level (to yield subtypes), which have been classically approached using variants of factor analysis [157] and clustering [158] respectively. To further parse heterogeneity approaches such as canonical correlation analysis [159], normative modelling [160] and anomaly detection methods [161, 162] can be explored. For example: In a recently published paper from our group, Holz et al. [163] employed a voxel-wise normative modelling approach to quantitatively assess heterogeneity in adversity effects. To estimate a pattern of regional deviations from typical brain structure for each participant, normative probability maps (NPM) were derived. Further, dice coefficients were calculated to assess the contribution of each adversity. This approach considered the correlated nature of adversities and helped explore both independent and combined long-term effects. Consequently, we found distinct neuroanatomical trajectories associated with specific adversities, indicating accelerated or delayed development in specific brain regions. Table 5 provides a brief overview of various computational approaches. For more details on statistical methodologies, we refer interested readers to Alpaydin [164] and Marquand et al. [154].

Table 5 Computational methods to parse heterogeneity.

Cohort synergy

To study individual level inferences, longitudinal designs that enable the investigation of within-individual change over time, are considered gold standard [165, 166]. The utilisation of longitudinal models to delineate trajectories would help elucidate the specific nature of deviations caused by adversity, including potential delays or accelerations in development. Further, adversity’s effects may differ depending on the timing of exposure. Despite evidence for the same [27, 28, 111], few studies have examined sensitive periods. Understanding the underlying mechanisms and what is biologically embedded during sensitive periods is crucial for comprehending how experiences shape neurobehavioral outcomes. Incorporating insights from formal modelling can help bridge these gaps [167]. Finally, questionnaires could be modified to capture not only the timings of adversity but also their duration, for a more intricate understanding of exposure. This information could serve as the foundation for conducting comparative analyses across different age groups, enabling the discernment of age-, duration- related variations in the impact of adverse experiences on brain.

While longitudinal tracking is a viable alternative, it is necessary to have a sizable group of children across a wide age range who have been exposed to adversity, as well as those who have not. To recruit and track a large sample is a challenge and consequently, the next crucial step is to develop methods for data pooling across different labs. By pooling cohorts and data, the size of datasets can be significantly increased. Machine learning approaches can then be applied to understand within population differences (e.g.: SES may vary across cohorts) and between population similarities (e.g.: loneliness). Further, data can be pooled across cohorts from different timepoints in an accelerated longitudinal design [168] with structural equation models estimate to converge multiple pieces into a single latent growth or specific latent classes by age [169]. Additionally, adopting age as a proxy for time, as in pseudo-longitudinal designs [170], offers a valuable strategy. Pseudo-longitudinal designs enable the exploration of temporal trends without the extended time commitment of traditional longitudinal research.

To address the challenge of binding data from different cohorts, neuroimaging studies can utilise statistical harmonisation techniques such as ComBat [171]. ComBat is a widely used method in neuroimaging research that aims to reduce batch (site/cohort) effects across different datasets but can also introduce bias [172]. Alternatively normative modelling can be used to accommodate site effects by providing a common reference to bind different samples together, thereby facilitating pooling of data for analysis [173]. Harmonisation of other kinds of data bring additional challenges such as aligning different measurements of the same construct. However, efforts have been initiated in these directions [174] and will be a major focus within the environMENTAL consortium (Schumann et al. [175]). This project aims to leverage existing cohorts from the past two decades, to enhance the efficiency of population neuroscience research, by harmonising the data that has already been collected. Also, cohorts (where participants can be re-contacted) will be enriched to address measurement issues and achieve the necessary depth of phenotyping for data-driven models.

Unravelling specific mechanisms

Big data studies have the advantage of large sample sizes, however, they may not capture the prevalence of some adversities (e.g.: institutional rearing). Thus, meta-analysis or multi-site collaborative studies are also important. The integration of information across centres and modalities might contribute to systematic mapping of neuronal plasticity. Further, certain adversities may exhibit small effect sizes individually but converge on similar outcomes (e.g.: toxins). Although these effects may be small, it is crucial to consider them for a nuanced understanding of the overall cumulative impact. A useful analogy can be drawn from genetics, where the aggregation of small genetic polymorphisms is compared to a watershed phenomenon [176]. As these genetic variations accumulate downstream, they eventually manifest in the syndromic expression of a disorder. Similarly, studying the effects of multiple diverse adversities can provide a comprehensive understanding of their combined influence.

Lastly, while large-scale studies provide generalisability, they may lack the detailed investigation of specific mechanisms. On the other hand, smaller experimental studies have the advantage of greater control, to delve into intricate processes. Thus, a combination of both large-scale and smaller experimental studies can enrich our understanding, with the former providing broader insights and the latter being more focussed and detailed.


The complexities of cumulative adversity and its effects on the brain pose significant challenges in terms of unravelling the underlying mechanisms and establishing comprehensive models. The heterogeneity in individuals’ experiences of adversity, the timing and duration of exposures, and the potential moderating factors all contribute to the complexity of the phenomenon. An examination of the current state of the field underscores the significance of conducting biological investigations on large-scale samples. Linking adversity to underlying biological mechanisms can help gain insights into the unique profiles and needs of individuals. Moreover, the utilisation of advanced computational approaches to disentangle heterogeneity and the combined effects of diverse adversities hold promise in this area. Continuous improvement of conceptual models, integrating insights gained from such research endeavours, will be critical for advancing knowledge and facilitating the translation to understanding data-driven sources of individual variance.