A stable and replicable neural signature of lifespan adversity in the adult brain

The study investigates whether exposure to environmental adversities across the lifespan leaves a persistent, developmentally stable signature in adult brain structure, and whether individual neurobiological deviations from this signature relate to psychopathology. Prior work shows adversities increase risk for psychiatric disorders and suggests neural adaptations that may be region- and adversity-specific. However, progress has been limited by (1) ROI-focused approaches emphasizing limbic circuitry that may overlook whole-brain alterations and interindividual variability; (2) reliance on group-mean comparisons that obscure heterogeneous individual effects and can yield inconsistent or null findings; (3) a lack of developmental context, as adversity effects are often interpreted without reference to normative age- and sex-specific brain trajectories; (4) correlated adversities complicating attribution of effects to single exposures; and (5) scarcity of longitudinal data to establish stability of adversity-related brain alterations into adulthood. To address these gaps, the authors employ voxel-wise normative modeling that quantifies individual centiles of variation relative to a population reference, using a rich, prospectively assessed profile of prenatal, perinatal, and psychosocial adversities in the Mannheim Study of Children at Risk (MARS) cohort. They also construct a large lifespan normative model of age-related morphometric change to interpret adversity effects as accelerated versus delayed maturation and test whether individual deviations from the adversity-based model predict psychopathology, with a focus on anxiety, depression, aggression, and attention problems.

Meta-analyses and prior studies report adversity-linked structural and functional alterations in regions subserving affective and cognitive regulation, including but not limited to limbic structures and prefrontal control regions. Nonetheless, findings are heterogeneous, sometimes showing opposite effects (e.g., volume increases vs decreases) for the same adversity-outcome associations. Contributing factors include ROI-centric designs, between-study differences in assessments and analyses, and group-mean methodologies that mask individual variability. Developmentally, interpreting adversity effects without normative growth curves is problematic due to nonlinear trajectories (e.g., inverted-U volume changes), potentially leading to misinterpretation as accelerated versus delayed maturation depending on age. Adversities co-occur (e.g., poverty with family stress and life events), complicating single-adversity analyses, especially in adulthood. Limited longitudinal work suggests some stability of adversity-related neural alterations across development, but methodological differences and lack of replication reduce confidence. Dimensional models of adversity (e.g., threat vs deprivation) have been proposed to explain divergent effects on maturation in youth; however, adult brain correlates may reflect accumulated and correlated exposures, blurring such distinctions.

Design and cohorts: A longitudinal, prospectively characterized at-risk birth cohort (MARS) was used to build voxel-wise normative models relating lifespan adversities to adult brain morphometry. MRI was acquired at ages ~25 (T1; n=169 after exclusions; 58% female) and ~33–34 (T2; n=114). An out-of-sample replication was performed in a sociodemographically similar IMAGEN subsample at age ~22 (n=115) with comparable adversity measures (life events and childhood trauma). A large multisite lifespan dataset (n=19,759, ages 8–97; nine sites) combined public repositories with MARS and IMAGEN to build age-related brain development normative models. Adversity measures: Seven adversities spanning prenatal to adulthood were included—maternal smoking during pregnancy; prenatal maternal stress; maternal sensitivity (3-month dyadic interaction); obstetric adversity (e.g., preterm labor, asphyxia); psychosocial family adversity (assessed repeatedly from 3 months to 11 years: parental, partnership, and family environment risks); childhood trauma (CTQ at 23 years); and negative life events (assessed repeatedly from infancy to 25 years). Scores were primarily prospective; adversities were binned into up to four categories to balance severity and sample sizes (with sensitivity analyses using raw scores). Imaging and features: High-resolution T1-weighted MRI (3T Siemens TRIO at T1; 3T Siemens PrismaFit at T2). Anatomical preprocessing in FSL. Morphometry quantified as Jacobian Determinants (JDs) of deformation fields from nonlinear registration (log-transformed), indexing regional volumetric expansion/contraction. Total intracranial volume (TIV) estimated; sex included as a covariate. Normative modeling: Bayesian linear regression (PCNtoolkit) with tenfold cross-validation to predict voxel-wise JDs from adversity predictors plus covariates (sex, TIV) in MARS T1; replication at MARS T2 and IMAGEN. Model accuracy assessed via voxel-wise correlation between observed and predicted JDs and standardized mean squared error. Structure coefficients (correlations between individual predictors and expected outcomes) quantified each adversity’s contribution to the predicted morphometric pattern. PCA summarized correlated adversities into three principal components (PCs) explaining ~63% variance; normative models were re-estimated using PCs to map multivariate adversity effects. Lifespan normative model of age-related development: Warped Bayesian linear regression with B-spline basis over age, covarying sex and site, trained/tested via split-half holdout, explained up to ~60% variance in JD trajectories. Age-related expansions and contractions mapped across regions. Individual deviations and psychopathology: For each subject, voxel-wise z-scores from adversity-based normative models defined deviations (negative: z < −2.6; positive: z > 2.6). Stability across T1–T2 assessed. Linear mixed models (random intercepts) tested whether individual deviations at T1 predicted concurrent and later psychopathology (anxiety, depression, aggression, attention) at T1 and T2, including models with T2 deviations; corrections applied for multiple outcomes and deviation types. Sensitivity analyses varied predictors (raw vs binned adversities; with/without TIV; adversity-only models; exclusion of obstetric adversity; alternative outcome using modulated gray matter volume) and compared prospective life events vs retrospective trauma models.

- Widespread, developmentally stable neuroanatomical signature of lifespan adversity: Adversity-based normative models at age 25 predicted voxel-wise morphometric variation across distributed regions, including vmPFC/vmOFC, ACC, thalamus, hippocampus, amygdala, basal ganglia, middle/superior frontal gyri, occipital and precentral gyri. This pattern replicated longitudinally at age 33–34 and in the independent IMAGEN sample at age 22. Model-observed correlations (two-sided) for prediction maps: MARS T1 r=0.62 (P<0.001), MARS T1 (subset) r=0.55 (P<0.001), MARS T2 r=0.67 (P<0.001), IMAGEN r=0.49 (P<0.001). - Specific adversity effects (structure coefficients): Low-to-moderate overlap between adversity-specific brain patterns (Dice coefficients), highest among psychosocial risks (family adversity, trauma, life events; Dice up to 0.54). Prenatal smoke exposure (mean Dice 0.23) and obstetric adversity (0.24) showed the least overlap with other adversities, indicating distinct patterns: obstetric adversity linked to expansions in vmOFC and contractions in ACC; prenatal smoking linked to expansions in hippocampus and contractions in postcentral/occipital gyri. Psychosocial family adversity showed the highest mean overlap (0.36), with expansions in subcortical limbic areas and contractions in vmOFC. - Multivariate adversity patterns (PCA): PC1 (lifespan psychosocial family adversities + prenatal smoking) associated predominantly with cortical and subcortical contractions (vmOFC, medial frontal, pre/postcentral, frontal pole/middle frontal, superior frontal, inferior temporal, caudate, occipital), with small expansions in vmPFC/paracingulate/superior frontal/lateral frontal pole. PC2 (early obstetric risk + prenatal stress) showed region-specific mix of contractions (lateral frontal pole, orbital frontal, inferior/superior frontal, middle temporal, fusiform, hippocampus, supramarginal, parietal/occipital) and expansions (medial frontal pole/vmOFC, caudate, superior frontal). PC3 (maternal sensitivity) showed contractions (paracingulate, supramarginal, insula, precuneus, lateral occipital, precentral, supplementary motor) and expansions (middle/superior frontal, vmPFC/perigenual ACC, vmOFC, precentral, angular gyrus, thalamus). Opposing trajectories were evident in vmOFC across PCs. - Lifespan age-related normative model: Using 19,759 individuals, models explained up to ~60% variance in JD development. Age-related expansions observed in limbic subcortical areas, medial cerebellum, ventricles; age-related contractions in frontal regions including vmPFC/vmOFC, ACC, inferior frontal gyrus. - Individual deviations and anxiety: Negative deviations (more contraction than predicted) were widespread; positive deviations (more expansion than predicted) were more focal (brainstem, thalamus, limbic system). Negative deviations predicted anxiety symptoms: at T1 and T2 using T1 deviations (β=0.07, SE=0.02, P=0.00006, η²=0.10; and β=0.06, SE=0.02, P=0.0005, η²=0.06, respectively). Associations remained significant when controlling for depressive symptoms, aggression, attention (β=0.03, SE=0.008, P=0.0003) and all adversities (β=0.06, SE=0.01, P=0.0001, η²=0.07). No significant associations for other symptom domains or positive deviations (P>0.05). - Sensitivity analyses: Results were robust to using unbinned raw adversity scores (r=0.97 with binned version, P<0.001), modeling adversity via PCs, excluding TIV (r=0.95 with original model), excluding obstetric adversity, using adversity-only models, and substituting the outcome with modulated gray matter volume (significant correlation with JD-based model, r=0.25, P<0.001). Prospective life events showed broader and stronger morphometric associations than retrospective trauma in several regions.

The findings demonstrate a robust, whole-brain neurobiological signature of cumulative lifespan adversity that persists from young adulthood into the early thirties and generalizes to an independent cohort. By leveraging voxel-wise normative modeling, the study moves beyond group averages to quantify individual neurobiological heterogeneity and link it to psychopathology. Adversity- and region-specific patterns suggest differential effects on maturation: for example, psychosocial adversities were associated with vmOFC contractions (consistent with accelerated development), while obstetric risks related to vmOFC expansions (suggestive of delayed maturation), echoing literature on preterm-related volumetric increases. Limbic structures (amygdala, hippocampus) also showed patterns consistent with accelerated development under psychosocial adversity versus delayed maturation with obstetric complications. These mechanistic insights reconcile previously inconsistent findings and highlight the necessity of situating adversity effects within normative developmental trajectories. Clinically, individual negative deviations from the adversity-based normative model predicted concurrent and future anxiety, underscoring the potential of individualized neurobiological markers derived from normative modeling to explain and predict psychopathological risk better than unmodeled data. The work suggests that adversity-related neuroplasticity leaves enduring traces in adult brain structure, with potential implications for targeted interventions and prevention strategies.

This study identifies a stable and replicable whole-brain morphometric signature of lifelong adversity exposure that persists through young adulthood. It delineates adversity- and region-specific expansions and contractions, interpretable as accelerated or delayed maturation relative to normative developmental trajectories. Crucially, individualized deviations from the adversity-based normative pattern—particularly excess contractions—predict current and future anxiety symptoms, indicating clinical relevance of neurobiological heterogeneity beyond group-level effects. Future research should: (1) acquire longitudinal neuroimaging across the full lifespan to examine sensitive periods and causal pathways; (2) replicate structure-coefficient profiles and brain–behavior associations in larger and clinical cohorts with harmonized assessments; (3) disentangle subjective versus objective, prospective versus retrospective adversity measures and their distinct neural correlates; and (4) incorporate dimensional adversity frameworks (e.g., threat vs deprivation) and multimodal eco-exposome assessments to refine mechanistic understanding and inform intervention.

- Imaging only in adulthood within the MARS cohort limits inference about timing effects and causal direction; volumetric differences could precede, result from, or interact bidirectionally with adversity exposure. - Structure coefficients are cohort-sensitive; larger samples are needed for more precise adversity-specific effect estimation and improved normative modeling. - The anxiety association with individual deviations requires replication in cohorts with comparable assessments; longitudinal data on the same scanner would facilitate modeling temporal changes (scanner differences at T1 vs T2 in MARS precluded this). - The largely nonclinical sample at T2 constrains generalizability to clinical populations; clinical validation is needed. - Differences between subjective versus objective adversity and prospective versus retrospective reporting could influence findings; while prospective life events showed broader effects than retrospective trauma here, the causes (measurement vs adversity nature) remain unclear. - Adversity dimensions (e.g., threat vs deprivation) were not separable with available instruments; results suggest predominant acceleration in limbic structures for psychosocial risks, but dimensional effects may be more evident in youth. - Comprehensive eco-exposome measures (e.g., intensity, controllability, duration, context; biological mediators like inflammation and HPA axis) were not integrated; future multimodal designs are needed.