Perinatal Factors and Neuroimaging Changes for Brain Maturation Among Adolescents

Eren Kafadar; Margaret Gardner; Lena Dorfschmidt; Jonathan A Berken; Audrey C Luo; Kevin Y Sun; Richard A I Bethlehem; Sara B DeMauro; Ran Barzilay; Varun Warrier; Tyler M Moore; Jakob Seidlitz; Heather H Burris; Theodore D Satterthwaite; Russell T Shinohara; Aaron F Alexander-Bloch

doi:10.1001/jamanetworkopen.2026.27643

. 2026 Aug 7;9(8):e2627643. doi: 10.1001/jamanetworkopen.2026.27643

Perinatal Factors and Neuroimaging Changes for Brain Maturation Among Adolescents

Eren Kafadar ^1,^2,^3,^✉, Margaret Gardner ^1,^2,³, Lena Dorfschmidt ^1,^2,³, Jonathan A Berken ^3,^4,^5,⁶, Audrey C Luo ^1,^3,⁷, Kevin Y Sun ^1,^2,^3,⁷, Richard A I Bethlehem ⁸, Sara B DeMauro ^4,⁵, Ran Barzilay ^1,^2,³, Varun Warrier ^8,⁹, Tyler M Moore ^1,^3,⁷, Jakob Seidlitz ^1,^2,³, Heather H Burris ^4,^5,^10,^11,^12,¹³, Theodore D Satterthwaite ^3,⁷, Russell T Shinohara ^10,^14,¹⁵, Aaron F Alexander-Bloch ^1,^2,^3,^✉

¹Department of Psychiatry, Perelman School of Medicine, University of Pennsylvania, Philadelphia

²Department of Child and Adolescent Psychiatry and Behavioral Science, The Children’s Hospital of Philadelphia (CHOP), Philadelphia, Pennsylvania

³Lifespan Brain Institute of CHOP and Penn, Philadelphia, Pennsylvania

⁴Department of Pediatrics, Perelman School of Medicine, University of Pennsylvania, Philadelphia

⁵Division of Neonatology, Department of Pediatrics, The Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania

⁶Division of Gastroenterology, Hepatology, and Nutrition, Department of Pediatrics, The Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania

⁷Penn Lifespan Informatics and Neuroimaging Center (PennLINC), Department of Psychiatry, Perelman School of Medicine, University of Pennsylvania, Philadelphia

⁸Department of Psychology, University of Cambridge, Cambridge, United Kingdom

⁹Department of Psychiatry, University of Cambridge, Cambridge, United Kingdom

¹⁰Department of Biostatistics, Epidemiology and Informatics, University of Pennsylvania, Philadelphia

¹¹Department of Obstetrics and Gynecology, Perelman School of Medicine at the University of Pennsylvania, Philadelphia

¹²Associate Editor, JAMA Network Open

¹³Leonard Davis Institute of Health Economics, University of Pennsylvania, Philadelphia

¹⁴Penn Statistics in Imaging and Visualization Center, Department of Biostatistics, Epidemiology, and Informatics, University of Pennsylvania, Philadelphia

¹⁵Center for Artificial Intelligence and Data Science for Integrated Diagnostics, University of Pennsylvania, Philadelphia

Accepted for Publication: June 11, 2026.

Published: August 7, 2026. doi:10.1001/jamanetworkopen.2026.27643

Open Access: This is an open access article distributed under the terms of the CC-BY-NC-ND License, which does not permit alteration or commercial use, including those for text and data mining, AI training, and similar technologies. © 2026 Kafadar E et al. JAMA Network Open.

^✉

Corresponding Authors: Eren Kafadar, BS, (eren.kafadar@pennmedicine.upenn.edu), and Aaron F. Alexander-Bloch, MD, PhD (aaron.alexander-bloch@pennmedicine.upenn.edu), Department of Psychiatry, University of Pennsylvania, 3700 Hamilton Walk, B504, Philadelphia, PA 19104.

Author Contributions: Ms Kafadar and Dr Alexander-Bloch had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Kafadar, Bethlehem, Moore, Shinohara, Alexander-Bloch.

Acquisition, analysis, or interpretation of data: Kafadar, Gardner, Dorfschmidt, Berken, Luo, Sun, DeMauro, Barzilay, Warrier, Seidlitz, Burris, Satterthwaite, Shinohara, Alexander-Bloch.

Drafting of the manuscript: Kafadar, Gardner, Moore.

Critical review of the manuscript for important intellectual content: Kafadar, Dorfschmidt, Berken, Luo, Sun, Bethlehem, DeMauro, Barzilay, Warrier, Seidlitz, Burris, Satterthwaite, Shinohara, Alexander-Bloch.

Statistical analysis: Kafadar, Gardner, Moore, Seidlitz, Shinohara.

Obtained funding: DeMauro, Warrier, Satterthwaite, Alexander-Bloch.

Administrative, technical, or material support: Dorfschmidt, Berken, Sun, Seidlitz, Satterthwaite, Alexander-Bloch.

Supervision: Bethlehem, Burris, Shinohara, Alexander-Bloch.

Conflict of Interest Disclosures: Ms Kafadar reported receiving support from National Institutes of Health (NIH) grants during the conduct of the study. Ms Gardner reported having inventorship interest in intellectual property licensed to Centile Bioscience by the Children’s Hospital of Philadelphia outside the submitted work. Dr Bethlehem reported being a cofounder and shareholder of Centile Bioscience outside the submitted work. Dr DeMauro reported receiving grants from the NIH during the conduct of the study and outside the submitted work. Dr Barzilay reported receiving grants from the National Institute of Mental Health during the conduct of the study and personal fees from Taliaz Health outside the submitted work. Dr Seidlitz reported receiving personal fees from Centile Bioscience outside the submitted work. Dr Shinohara reported receiving grants from the NIH during the conduct of the study and personal fees from Octave Bioscience outside the submitted work. Dr Alexander-Bloch reported receiving grants from the NIH during the conduct of the study and receiving personal fees from Centile Biosience; having inventorship interest in intellectual property licensed to Centile Bioscience by the Children’s Hospital of Philadelphia; and owning a patent for Automated System and Method for Two-Dimensional Cranial Biometry With Multi-Metric Quality Control from Medical Imaging Data, pending, outside the submitted work. No other disclosures were reported.

Funding/Support: This work was supported by grants R01MH132934, R01MH133843, and R01MH134896-01 from the NIH (principal investigator: Dr Alexander-Bloch).

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: Dr Burris is an Associate Editor of JAMA Network Open but was not involved in any of the decisions regarding review of the manuscript or its acceptance.

Data Sharing Statement: See Supplement 4.

^✉

Corresponding author.

PMCID: PMC13451985 PMID: 42566216

This cohort study investigates the association of longitudinal change in brain volumes with birth weight, gestational age, and changes in psychopathology among children and adolescents.

Key Points

Question

Are differences in brain maturation trajectories, quantified by novel conditional-longitudinal models, associated with perinatal factors and mental health in adolescence?

Findings

In this cohort study of 10 830 baseline and 7262 follow-up neuroimaging data from the Adolescent Brain Cognitive Development (ABCD) Study, conditional-longitudinal normative models revealed that trajectories of brain maturation in adolescence were associated with birth weight and longitudinal changes in mental health.

Meaning

The findings suggest that conditional-longitudinal models detect interindividual variability in brain maturation, which is associated with both perinatal factors and concurrent changes in psychopathology.

Abstract

Importance

Brain maturation varies between individuals, particularly during dynamic developmental periods such as adolescence. Directly assessing the differences in longitudinal trajectories can reveal deviations from normative patterns.

Objective

To ascertain the association of longitudinal change in brain volumes with birth weight, gestational age, and longitudinal changes in psychopathology.

Design, Setting, and Participants

In this cohort study, cross-sectional and longitudinal normative models were developed for brain volumes from the first 2 neuroimaging data collection time points (baseline: 2016-2018; follow-up: 2019-2021) of the Adolescent Brain Cognitive Development (ABCD) Study, an ongoing community-based longitudinal cohort study at 21 US sites. Longitudinal models indexed an individual’s expected brain volume at follow-up, conditioned on their baseline measurement, and thus were conditional-longitudinal models. Split-half subsets on demographically matched samples were used to fit the models. ABCD Study participants were recruited through the US school systems. Exclusion criteria included severe medical conditions that interfered with the study protocols. The present analysis further excluded the sample based on imaging quality flags and missing data. Data analysis was performed between May 2024 and August 2025.

Exposures

Birth weight and gestational age derived from parent-reported questionnaires. General psychopathology and subfactor scores were calculated using a bifactor model.

Main Outcomes and Measures

Brain volumes and change in volume between time points 1 (baseline) and 2 (follow-up). Cross-sectional and longitudinal centiles were used to quantify deviations in volumes.

Results

The sample included 10 830 ABCD Study participants with neuroimaging data collected at baseline (mean [SD] age, 9.9 [0.62] years; 5609 males [51.8%]) and 7262 with data collected at follow-up (mean [SD] age, 12.0 [0.65] years; 3875 males [53.4%]). Longitudinal centiles were sensitive to individual-specific changes in brain volumes. Lower birth weight was associated with lower longitudinal centiles, suggesting larger decreases in brain volumes over time (n = 27 regions, β range = 0.030-0.083). Lower longitudinal centiles were associated with greater increases in psychopathology, suggesting decreasing brain volumes with increasing psychopathology scores (n = 37 regions, β range = −0.061 to −0.031). Changes in psychopathology were not associated with brain volumes at either time point when indexed by cross-sectional centiles.

Conclusions and Relevance

In this cohort study, conditional-longitudinal models captured individual-level deviations from expected growth trajectories, providing information beyond static positions on a growth curve to assess differences in maturation. Robust associations were observed between individual trajectory deviations, birth weight, and longitudinally assessed mental health symptoms. Conditional-longitudinal models hold promise for applications across psychiatric neuroscience, from development to aging.

Introduction

Normative models of neuroimaging data benchmark interindividual differences in quantitative features against population reference models. Similar to pediatric growth charts of height or weight, such “brain charts,” can boost interpretability and power to detect clinical differences.^1,2,3 Analyses using deviation scores derived from normative models of brain features compared with analyses of raw measures can increase detection of meaningful patterns in populations as well as specific individuals.² Previous normative modeling studies largely model longitudinal data using cross-sectional brain charts, which do not directly address variability in longitudinal change.^4,5,6 Over time, individuals with extreme centiles tend to move closer to the mean, with different expected growth velocities based on baseline centiles.^7,8 Longitudinal normative models can account for the regression to the mean by benchmarking change between time points, which is particularly useful in highly dynamic periods of brain development such as adolescence. Here, we present conditional-longitudinal models where within-individual change is quantified by conditioning follow-up brain-volume measurements on baseline values. Harnessing anatomical brain magnetic resonance imaging (MRI) data from the Adolescent Brain Cognitive Development (ABCD) Study,⁹ we tested for brain trajectory differences associated with perinatal developmental factors (eg, birth weight) and with contemporaneous measures of psychopathology.

Extensive research links perinatal factors, including birth weight and gestational age, to brain MRI differences persisting into adolescence and adulthood. In general, prematurity and low birth weight are associated with smaller regional brain volumes.^{10,11,12,13,14} However, few studies provide a longitudinal perspective on whether outcomes of perinatal adversity are stable or whether they play a role in the dynamics of brain maturation. Prematurity may result in slower brain growth in childhood,^15,16 and there may be age-varying associations between birth weight and brain volume.^17,18 However, prior work has relied on modeling approaches that do not fully disentangle static differences from differences in how the brain changes. For instance, if birth weight is associated with brain volume at baseline and if individuals with larger brains tend to have larger absolute change between time points, then associations with change might simply reflect a propagation of baseline differences rather than meaningful differences in longitudinal trajectories. Conditional-longitudinal models resolve this uncertainty by quantifying degree of expected change for each individual based on their baseline brain volumes.

Adverse perinatal factors have also been linked to increased susceptibility to psychiatric conditions.^19,20,21,22 Adolescence is a critical period for both brain development and sensitivity to mental illness, and individual deviations from normative brain-growth patterns may be related to changes in mental health.^23,24,25 Baseline neuroimaging features in the ABCD Study have previously been linked to overall psychopathology.^26,27 Smaller studies have also suggested that structural brain changes can be factors in severity of psychiatric symptoms over time.^28,29 However, the potential utility of longitudinal normative models remains untested. Understanding the interactions between perinatal factors, brain maturation, and emerging psychopathology in adolescence may elucidate potential links between adverse perinatal factors and elevated risk for psychiatric conditions. In this study, we introduce conditional-longitudinal models to address this gap. Our aim was to ascertain the association of longitudinal change in brain volumes with birth weight, gestational age, and longitudinal changes in psychopathology.

Methods

Overview

This study used baseline (collected 2016-2018) and follow-up (collected 2019-2021) neuroimaging data acquired in the ABCD Study, an ongoing community-based longitudinal cohort study at 21 US sites that includes 11 873 children aged 9 to 10 years at enrollment. Participants are generally recruited through school systems. Caregivers of participants provide informed consent, and study sites receive approval from their respective institutional review board. The ABCD Study excludes participants with severe medical conditions, which could interfere with protocols. We further removed participants based on data availability and scan quality (eMethods 1 in Supplement 1). In accordance with 45 CFR §46.104, this cohort study was exempt from ethics review and informed consent requirement because it was a secondary analysis of deidentified data and was not human participant research. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Imaging phenotypes included regional cortical volumes (Desikan-Killiany parcellation³⁰) and subcortical volumes (Aseg parcellation³¹). Details on MRI image acquisition and processing are described in Casey et al⁹ and eMethods 2 in Supplement 1. Gestational age, birth weight, and neonatal complications (eTable 1 in Supplement 1) were derived from parent-reported questionnaires. Given their high colinearity, birth weight was corrected for gestational age using the growthstandards package, version 0.1.6 in R (R Project for Statistical Computing) to calculate birth weight percentiles.^32,33 Polygenic scores (PGS) for birth weight, which quantify genetic associations with birth weight based on genetic variants identified through large-scale genomewide association studies (GWAS),³⁴ were calculated with polygenic risk scores continuous shrinkage,³⁵ trained on GWAS³⁶ from the Early Growth Genetics,³⁷ and limited to the European ancestry subsample (eMethods 3 in Supplement 1). Psychopathology was quantified via the general p-factor as well as orthogonal subfactors derived using a longitudinal bifactor model described in Moore et al²⁶ (eMethods 4 in Supplement 1).

Normative Models

Normative models using generalized additive models for location, scale, and shape (GAMLSS) were fit to brain volumes via gamlss, version 5.4.22.³⁸ This framework has been reliably applied to model brain growth trajectories, allows for flexible models, and thus ensures the adaptability of our longitudinal modeling methods to future applications and different datasets.^1,3 Distribution family Box-Cox-t was used, considering both cross-sectional and longitudinal models. Model selection followed previous selection methods.^1,3 Covariates included age and sex for cross-sectional models as well as time point 1 (baseline) brain volumes and interscan interval for conditional-longitudinal models. A random effect for site was also included. Conditional-longitudinal models of time point 2 (follow-up) brain volumes included time point 1 volumes for conditioning on baseline measurements and the interscan interval as an optional covariate in the model selection space, yielding participant-specific reference distributions for time point 2 brain volumes given time point 1 brain volumes (eMethods 5 in Supplement 1).

Models were fit in 2 matched split-halves based on the ABCD Reproducible Matched Samples.^39,40 To prevent data leakage, we calculated centiles (ie, participant-specific deviation from normative models) for split-half A using model estimates from split-half B, and vice versa. Demographic characteristics were comparable between the split-half groups (Table; eTable 2 in Supplement 1). Model parameters are provided in eTables 4 to 6 in Supplement 2.

Table. Distribution of Modeling Variables and Variables of Interest for Each Split Sample and Time Point.

Variable	Sample, mean (SD)
	Time point 1 (baseline)		Time point 2 (follow-up)
	Split-half A (n = 5421)	Split-half B (n = 5409)	Split-half A (n = 3674)	Split-half B (n = 3588)
Age, y	9.93 (0.62)	9.91 (0.62)	11.97 (0.65)	11.96 (0.65)
Sex, No. (%)
Female	2648 (49)	2573 (48)	1738 (47)	1649 (46)
Male	2773 (51)	2836 (52)	1936 (53)	1939 (54)
Gestational age, wk	39 (2)	39 (2)	39 (2)	39 (2)
Gestational age, available^a	5347	5330	3628	3546
Birth weight, oz	112 (24)	113 (23)	112 (24)	113 (23)
Birth weight, available^a	5196	5165	3539	3448
Birth weight PGS, z score^b	−0.005158 (1.01)	0.000128 (0.99)	−0.003601 (1.01)	0.010892 (0.99)
Birth weight PGS, available^a^,^b	2413	2367	1779	1703
Neonatal complications, No.	0.394 (0.774)	0.402 (0.789)	0.398 (0.783)	0.410 (0.799)
Neonatal complications, available^a	5344	5337	3627	3588
P-factor, z score	−0.012353 (1.00)	0.002284 (1.00)	−0.009939 (0.99)	0.010219 (1.00)
P-factor, available^a	5421	5409	3574	3452

Open in a new tab

Abbreviation: n, sample size; PGS, polygenic score.

SI conversion factor: To convert ounces to grams, multiply by 28.

^{^a}

Available sample size for models including each variable after excluding participants with missing data.

^{^b}

PGS was calculated on European ancestry subsample only.

Statistical Analysis

Linear regression tested associations between our outcome variable (brain centiles) and our independent variables. Centiles indexed brain size in cross-sectional models and conditional change between time points in longitudinal models. Independent variables in separate regression models included birth weight, gestational age, birth weight percentile, birth weight PGS, neonatal complications, and p-factor for psychopathology. All independent variables were z scored, and centiles were converted into centile z scores using qnorm in R (equivalent to z scores for normally distributed data). To generate a summary value across regional centiles, we calculated a cumulative deviation score across regional centiles: the centile Mahalanobis distance (CMD¹) (eMethods 5 in Supplement 1). Specific covariates depended on the variable being tested. To account for differences in baseline psychopathology scores, we included baseline p-factor scores as a covariate with change in p-factor scores. To account for genetic ancestry effects, we included the first 10 ancestry principal components as covariates with birth weight PGS. To test birth weight associations without the genetic component, we included birth weight PGS as a covariate with birth weight (eMethods 6 in Supplement 1). Across all variables, sensitivity analyses included covariates of parental income and educational level as well as current height and weight. To explore associations above and beyond brain size as a whole, we ran regressions, including an intracranial volume covariate (eMethods 7 in Supplement 1).

Multiple comparisons were corrected using the Benjamini-Hochberg, with false discovery rate–corrected P < .05 considered significant. Analyses were run between May 2024 and August 2025 using publicly available tabular ABCD Data Release 5.1 in R, version 4.2.3. Anatomical visualizations were rendered using ggseg version 1.6.5.⁴¹

Results

The sample included 10 830 participants at baseline (mean [SD] age, 9.9 [0.62] years; 5221 females [48.2%], 5609 males [51.8%]) and 7262 at follow-up (mean [SD] age, 12.0 [0.65] years; 3387 females [46.6%], 3875 males [53.4%]) (Table). A total of 7192 participants at the same site between the 2 time points were included in the conditional-longitudinal models.

Unique Information From Conditional-Longitudinal Normative Models

Conditional-longitudinal models computed at time point 2 were sensitive to individual brain change between time points, yielding distinct information from individual centiles calculated from cross-sectional models computed at time point 1 and time point 2 (Figure 1A, B). Using CMD as a cumulative measure of deviation, cross-sectional CMD time point 1 and CMD time point 2 were highly intercorrelated (r = 0.82, P < 2.20 × 10⁻¹⁶), while their correlations with longitudinal CMD were attenuated (time point 1: r = 0.17, P < 2.20 × 10⁻¹⁶; time point 2: r = 0.18, P < 2.20 × 10⁻¹⁶) (eFigure 1 in Supplement 1). A participant’s conditional-longitudinal centile, or their deviations from expected change in brain size between time points, was dependent on an individual-specific association based on their brain size at baseline and other relevant covariates, such as age and sex (Figure 1C; eFigure 1 in Supplement 1). Conditional-longitudinal models more accurately captured changes between time points, especially by accounting for expected regression to the mean. There were significant inverse correlations between time point 1 centiles and the difference in centiles between time points (maximum r = –0.37, P = 2.20 × 10⁻¹⁶; minimum r = –0.024, P = .04), which were attenuated when using conditional-longitudinal centiles to directly quantify change between time points (maximum r = –0.095, P = 1.00 × 10⁻¹³; minimum r = –0.030, P = .03) (eFigure 1 in Supplement 1).

Three-panel plots of age versus total cortical G M and centile z scores. Panel A title: Participants with total cortical G M volume at time point 1. Horizontal axis labeled Age, y, ranging from 8 to 11. Vertical axis labeled Total cortical G M, mm3, ranging from 450000 to 750000. Multiple thick gray, slightly downward sloping reference curves span ages about 9 to 11 at approximate levels near 540000, 560000, 590000, 620000, 660000, 690000, and 710000. Three colored circular markers: an orange dot near age 10 point 3 and 560000; a light blue dot near age 10 point 2 and 605000; a dark blue dot near age 10 point 6 and 715000. Panel B title: Same participants with total cortical G M volume at time point 2. Horizontal axis labeled Age, y, ranging from 10 to 14. Vertical axis labeled Total cortical G M, mm3, ranging from 450000 to 750000. Similar thick gray, downward sloping reference curves span ages about 11 to 14 at approximate levels near 515000, 535000, 550000, 585000, 620000, 655000, 690000, and 710000. Three colored circular markers: light blue near age 12 and 588000; orange near age 12 point 6 and 555000; dark blue near age 12 point 6 and 722000. Panel C title: Relationship between centiles at time point 2 and each participant’s conditional-longitudinal centile. Legend at upper left labeled Centile at time point 1, percent, with colored dots and values 14 in orange, 41 in light blue, and 97 in dark blue. Horizontal axis labeled Time point 2, z score, centile, ranging from minus 3 to plus 3 with centile annotations: minus 3 less than 1 percent, minus 2 2 percent, minus 1 16 percent, 0 50 percent, plus 1 84 percent, plus 2 98 percent, plus 3 greater than 99 percent. Vertical axis labeled Longitudinal, z score, centile, with the same tick labels and centile annotations. A black dashed diagonal identity line runs from lower left to upper right. Three smooth, steep S shaped curves: orange leftmost, light blue centered, dark blue rightmost. Each curve has one filled dot: orange dot near x about minus 1 and y about 0; light blue dot near x about minus 0 point 3 and y about minus 1; dark blue dot near x about plus 2 and y about plus 1 point 4. — Conditional-longitudinal models characterize participant brain size at time point 2 based on the participant’s brain size at baseline (time point 1). A, Orange, light blue, and dark blue dots represent 3 participants with low (14th), middle (41st), and high (97th) total cortical gray matter (GM) centiles at time point 1. B, Changes to each participant’s cortical GM centiles at time point 2 are shown (14th to 16th, 41st to 34th, 97th to 98th). C, Dots on the curves mark the true time point 2 centile and the corresponding conditional-longitudinal centile for each participant. The x- and y-axes of panel C show centile z scores to illustrate the ends of the centile distribution. Dashed line represents the identity line where time point 2 centile is equal to the longitudinal centile.

Birth Weight Association With Conditional-Longitudinal Centiles

Next, we examined associations between both cross-sectional and conditional-longitudinal centiles and perinatal factors. Smaller birth weight and younger gestational age were each associated with lower cross-sectional centiles at both time points, corresponding to smaller brain volumes (birth weight time point 1: n = 96 regions, β range = 0.023-0.224; time point 2: n = 94 regions, β range = −0.026 to 0.223; gestational age time point 1: n = 65 regions, β range = −0.045 to 0.086; time point 2: n = 37 regions, β range = −0.048 to 0.075). The magnitude of these associations was greater for birth weight and varied regionally (Figure 2). An exception was centiles for ventricles, which showed an inverse association such that ventricular volume was smaller given older gestational ages at birth.

Data figure with brain maps for birth weight and gestational age at two time points. Two labeled panels arranged vertically. Panel A label at upper left: Birth weight. Panel B label at lower left: Gestational age. A horizontal color bar at the top right spans dark blue through white to dark red, labeled beta beneath the bar, with tick labels at minus zero point two, minus zero point one, zero, zero point one, and zero point two. In Panel A, three column headers run left to right: Time point 1, Time point 2, and Conditional-longitudinal centile, each with a small upward-pointing triangle marker above the header line. Under each header are paired cortical surface maps in lateral and medial views, followed by two rows of subcortical section silhouettes at the bottom: one coronal-like slice and one sagittal-like slice. Brain regions are outlined with black parcel boundaries; many parcels are filled with light pink to red tones, with some parcels in pale blue. Several areas are filled solid medium gray, indicating regions without color overlay. In the Conditional-longitudinal centile column of Panel A, most cortical parcels appear medium gray with only scattered pale pink parcels. In Panel B, the same three column headers repeat: Time point 1, Time point 2, and Conditional-longitudinal centile, again with paired lateral and medial cortical maps per column and the same two subcortical slice silhouettes at the bottom. In Panel B, Time point 1 includes widespread pale pink parcels with some medium gray regions; Time point 2 contains more medium gray parcels with fewer pink regions. In the Conditional-longitudinal centile column of Panel B, nearly all cortical parcels are medium gray with minimal pale pink coloration. No numeric axes are present beyond the beta color scale; no scale bars are visible. — Lower birth weight and smaller gestational age were associated with overall lower cross-sectional centiles (corresponding to smaller size), while only lower birth weight was associated with lower conditional-longitudinal centiles (corresponding to greater-than-expected decreases in volume). Regions that did not survive multiple comparisons are in gray. Information on regression statistics across all regions and global tissue volumes is provided in eTables 7 to 9 in Supplement 3.

Testing the association between perinatal factors and brain change, we found that conditional-longitudinal centiles were directly associated with birth weight, and no significant associations were found with gestational age. Smaller birth weight was associated with smaller regional volumes at time point 2 after accounting for the size of regional volumes at time point 1, especially for frontal and temporal cortical volumes, bilateral thalamus, and brainstem volumes (n = 27 regions; β range = 0.030-0.083; FDR-corrected P < .05) (Figure 2).

Given the degree of covariance between birth weight and gestational age, we calculated a birth weight percentile using published intergrowth charts.²⁹ The birth weight percentile provided a measure of birth weight after accounting for gestational age. The association between birth weight percentile and brain volume centiles closely followed results from linear regressions with birth weight itself, showing robustness of birth weight–brain associations when the association between birth weight and gestational age was accounted for (birth weight percentile time point 1: n = 98 regions, β range = 0.018-0.161; time point 2: n = 96 regions, β range = 0.022-0.167; longitudinal: n = 31 regions, β range = 0.023-0.059) (eFigure 2 in Supplement 1). Magnitudes for birth weight percentile were largely robust to including gestational age as a covariate (birth weight percentile time point 1: n = 98 regions, β range = 0.018-0.164; time point 2: n = 96 regions, β range = 0.022-0.170; longitudinal: n = 33 regions, β range = 0.022-0.061) (eTables 7-9 in Supplement 3).

We performed additional analyses to understand underlying factors contributing to birth weight associations. We calculated birth weight PGS, aggregating genetic factors associated with birth weight based on a GWAS³¹ (eMethods 3 in Supplement 1). In support of the generalizability of birth weight PGS to the European ancestry subsample of the ABCD Study (time point 1: n = 4780; time point 2: n = 3482), we confirmed a small effect size but significant association between birth weight PGS and the phenotype of birth weight in our sample (r = 0.14; P < 2.20 × 10⁻¹⁶). Higher birth weight PGS (time point 1: n = 73 regions, β range = 0.031-0.082; time point 2: n = 71 regions, β range = 0.035-0.097) was associated with higher brain centiles but not with conditional-longitudinal centiles (eFigure 3 in Supplement 1).

Additionally, we evaluated a composite score reflecting the burden of neonatal complications to take into account the higher rate of neonatal complications following premature births.⁴² Higher neonatal complications were associated with lower centiles, with the largest effect sizes for subcortical structures observed at time point 1 (n = 26 regions; β range = −0.056 to 0.031; FDR-corrected P < .05), with some regions also found to have significant associations at time point 2 (n = 3 regions; β range = −0.048 to −0.043; FDR-corrected P < .05) (eFigure 4 in Supplement 1). Conditional-longitudinal centiles were not significantly associated with neonatal complications. Findings for birth weight and gestational age were largely robust to controlling for neonatal complications (eTables 7-9 in Supplement 3).

We conducted sensitivity analyses to confirm the robustness of these findings. First, given the possible associations between birth-related factors and socioeconomic status, we included income level and parental educational level as covariates in the linear regressions and found our results to be largely robust (birth weight time point 1: n = 94 regions, β range = −0.016 to 0.217; time point 2: n = 90 regions, β range = −0.025 to 0.212; longitudinal: n = 25 regions, β range = −0.025 to 0.082) (eFigure 5 in Supplement 1). Second, given the possible associations between perinatal factors and adolescent body size, we included height and weight as covariates and similarly found convergent results (birth weight time point 1: n = 96 regions, β range = −0.027 to 0.231; time point 2: n = 93 regions, β range = −0.035 to 0.230; longitudinal: n = 35 regions, β range = −0.019 to 0.083) (eFigure 6 in Supplement 1).

Given that regional outcomes could have an overarching association with intracranial size, we included centile scores of intracranial volume as a covariate. Associations with birth weight, gestational age, and birth weight percentile, but not for birth weight PGS, were robust to controlling for intracranial volume (birth weight time point 1: n = 81 regions, β range = −0.049 to 0.142; time point 2: n = 76 regions, β range = −0.055 to 0.147; longitudinal: n = 27 regions, β range = −0.024 to 0.083) (eFigure 7 in Supplement 1).

Association of Increases in Psychopathology Over Time With Lower Conditional-Longitudinal Brain Centiles

To understand how the association between birth weight and brain maturation may relate to emerging psychopathology, we examined the associations between perinatal factors and p-factor. We found that at baseline, while birth weight (r = 0.042; P = 3.78 × 10⁻⁴) and gestational age (r = 0.027; P = .02) showed minimal inverse associations with psychopathology, higher burden of neonatal complications showed a somewhat higher magnitude of direct association (r = 0.093; P = 4.30 × 10⁻¹⁵). Next, we tested the association between p-factor and brain centiles from our models. We found that at both baseline and follow-up time points, p-factor was inversely associated with cross-sectional centiles, meaning that smaller brain volumes were associated with higher psychopathology scores (p-factor time point 1: n = 95 regions, β range = −0.080 to −0.020; time point 2: n = 81 regions, β range = −0.075 to −0.028) (Figure 3). To assess longitudinal change, we looked at how within-individual change in p-factor between time points was associated with longitudinal brain maturation and found an inverse relationship. Larger increases in p-factor were associated with smaller conditional-longitudinal centiles across multiple frontal, parietal, and occipital cortical regions as well as in the cerebellum and bilateral hippocampus (p-factor longitudinal: n = 37 regions; β range = −0.061 to −0.031; FDR-corrected P < .05) (Figure 3). In contrast, within-individual changes in p-factor had no significant associations with brain size, indexed by cross-sectional centiles. This finding underscores the increased sensitivity of conditional-longitudinal models to person-level changes in longitudinally measured behavioral factors (eTables 7-9 in Supplement 3). In extended exploratory analyses, several psychopathology subfactors mirrored the associations between the general p-factor and cross-sectional centiles, while increases in the psychosis subfactor were associated with smaller longitudinal centiles across the brain (eMethods 4, eTable 3, and eFigure 8 in Supplement 1).

Data figure with brain maps for time point 1, time point 2, and centile. Horizontal color bar at the top right with a blue to white to red gradient; tick labels from minus zero point two, minus zero point one, zero, zero point one, to zero point two. Centered under the bar, the Greek letter beta. Below, three labeled columns separated by wide spacing: left header Time point 1, middle header Time point 2, right header Conditional-longitudinal centile. A small upward-pointing triangle marker appears above each header line. Each column contains six stylized brain renderings arranged in three rows by two columns. In each row, the left image is a lateral cortical surface view and the right image is a medial cortical surface view; the same arrangement repeats for the three columns. Cortical parcels are outlined with thin black borders and filled primarily with very light blue shading; some parcels are unfilled white. Multiple regions across panels are filled solid medium gray, distinct from the blue shading. In the top row of each column, gray regions occupy large medial areas on the right-hand brain view, including a broad central medial region, with smaller gray patches on lateral views. In the middle row, gray regions again cover extensive medial surfaces on the right-hand views and smaller lateral patches on the left-hand views. In the bottom row of each column, two schematic slice views appear: a coronal slice on the left and a sagittal slice on the right, each with a thick dark gray outer rim and internal light blue regions; small internal structures are outlined in black, and a brainstem-like shape appears in light blue in the sagittal view. No numeric axes are present beyond the beta color scale, and no scale bars are visible. — Higher psychopathology measured at time points 1 and 2 was associated with corresponding lower cross-sectional centiles, and larger increases in psychopathology between time points was associated with lower conditional-longitudinal centiles. Regions that did not survive multiple comparisons are in gray. Information on regional statistics and regions not visualized here is provided in eTables 7 to 9 in Supplement 3.

Discussion

We believe that, by introducing the conditional-longitudinal normative models, we extended the use of normative modeling datasets by specifically targeting variability in trajectories of brain development. This framework uniquely allows assessment of individual deviations in developmental trajectories beyond baseline differences in size, which was not feasible with previous cross-sectional normative approaches to longitudinal neuroimaging datasets. The ABCD Study demonstrated that conditional-longitudinal centiles were associated with both perinatal factors and longitudinal changes in psychopathology scores during early adolescence.

Our analyses with birth weight and gestational age showed minimal association with the general psychopathologic factor, despite prior studies showing some association. This finding could be due to the general p-factor not being well-suited to capture higher effect sizes with perinatal factors or due to low sampling of participants with very low birth weight and gestational age in our cohort.

We found that both perinatal and psychopathology factors were associated with brain maturation. Consistent with previous cross-sectional studies,^17,43,44,45 lower cross-sectional centiles, indexing smaller volumes, were associated with lower gestational age, lower birth weight, and higher burden of neonatal complications. Birth weight had the most robust associations, with cross-sectional brain centiles; lower birth weight was also associated with lower conditional-longitudinal centiles in the frontal and temporal cortex, bilateral thalamus, and brainstem. The genetic contribution to birth weight, indexed by birth weight PGS, was associated total brain size but not regional volumes or longitudinal change. Our sample size for birth weight PGS was limited to the European ancestry subsample of the ABCD Study due to ancestry bias in the underlying GWAS, which remains common in genetic studies, and may lead to insufficient power.⁴⁶ However, it is also plausible that the association between birth weight and longitudinal brain change may be through nongenetic pathways. For example, intrauterine conditions leading to lower birth weight may also lead to epigenetic changes,⁴⁷ which continue to affect brain growth through metabolic and other cellular mechanisms.

Greater psychopathology was also associated with smaller brain volumes, as previously reported.²⁷ Our novel analyses found that greater within-individual increases in psychopathology were associated with lower conditional-longitudinal centiles in frontal, parietal, and occipital cortex, cerebellum, and bilateral hippocampus. These findings were largely robust to controlling for intracranial volume as well as adolescent body size and socioeconomic status.

Cortical areas whose longitudinal maturation was specifically linked to birth weight appeared primarily in association cortices, which have a more prolonged developmental trajectory and thus could be more susceptible to altered brain maturation dynamics related to lower birth weight.^1,48 Slower childhood growth in thalamus has been previously reported in very preterm children, which continues to mature through adolescence, especially in connectivity with association cortices, which are implicated in working memory, cognitive flexibility, and executive functioning.^49,50 The associations between birth weight and continued development of these brain regions could explain lower scores on cognitive tasks in children and adolescents with perinatal adversity.⁵¹

Regions associated with changes in psychopathology also included association areas in frontal and parietal lobes, which is consistent with previous reports of accelerated thinning and volume reduction in the frontal lobes with emerging depressive symptoms in adolescents.^29,52 Additionally, slower growth in the hippocampus and cerebellum have been linked to increases in depressive symptoms and emotional problems in this age group.^52,53 Increases in psychopathology have been associated with smaller longitudinal centiles in sensory areas, which could signal more widespread network-level disruptions in cortical maturation underlying increases in psychopathology.^54,55 Our extended results on subfactors of psychopathology support this view. Increases in the psychosis subfactor specifically are associated with greater decreases in brain volumes, which is consistent with theories of increased synaptic pruning during adolescence as an underlying developmental mechanism for schizophrenia.^56,57

Limitations

This study has limitations. We reported on brain volumes within a single cortical parcellation during adolescence. Future work should investigate other morphometric features, alternative parcellations, and studies of the aging brain, where conditional-longitudinal centiles may be particularly useful. While we focused on conditional growth charts rather than velocity growth charts, velocity charts—which are based on the rate of change at fixed time intervals—may be another important application of normative models to longitudinal data.⁵⁸ Finally, while we used publicly released tabular data from the ABCD Study to maximize reproducibility and followed standard quality control procedures, biases related to suboptimal image quality may still persist in these data.⁵⁹

Conclusions

In this cohort study of participants with neuroimaging data at baseline and 2-year follow-up, we quantified interindividual variability in brain maturation by fitting longitudinal models conditioned on the baseline measurements. Centiles calculated from these models yielded insights into long-debated questions about how perinatal factors may affect subsequent brain development and how within-participant brain changes play a role in mental health during adolescence. These models may ultimately inform early identification of adolescents at risk for atypical brain development, providing a framework for preventive interventions.

All data and code are publicly available to facilitate future applications. Our approach may prove useful for investigating deviations from normative trajectories in future clinical and preclinical samples, to ascertain how trajectories of psychiatric conditions relate to the dynamics of brain development. Furthermore, the framework of conditional-longitudinal models can be widely applied to longitudinal data from both developing and aging populations, to improve characterization of how the brain continues to change throughout the lifespan.

Supplement 1.

eMethods 1. Sample

eMethods 2. Imaging-Derived Phenotypes (IDPs)

eMethods 3. Perinatal Factors

eMethods 4. P-Factor

eMethods 5. Normative Modeling

eMethods 6. Linear Regressions

eMethods 7. Additional Covariates in Sensitivity Analyses

eResults. Post Hoc Analyses With Psychopathology Subfactors

eFigure 1. Conditional Longitudinal Centiles Are Distinct From Cross-Sectional Centiles and Account for Regression to the Mean

eFigure 2. Birth Weight Percentile Shows Findings Consistent With Birth Weight Results

eFigure 3. PGS for Birth Weight Mirrors Associations With Birth Weight for Cross-Sectional Centiles

eFigure 4. More Neonatal Complications Are Associated With Smaller Cross-Sectional Centiles

eFigure 5. Results Are Robust to Controlling for Socioeconomic Factors

eFigure 6. Results Are Largely Robust to Controlling for Anthropometric Factors

eFigure 7. Results Are Robust to Controlling for Intracranial Volume, Except for Polygenic Score

eFigure 8. Psychopathology Subfactors Are Differentially Associated With Brain Size, and Increases in Psychosis Are Associated With Smaller Brain Volumes Over Time

eTable 1. ABCD Developmental Questions for Deriving the Neonatal Complication Burden Measure

eTable 2. Distribution of Covariates for Sensitivity Analyses in Each Split Sample and Timepoint

eTable 3. Subfactor Reliability

eReferences

jamanetwopen-e2627643-s001.pdf^{(4.1MB, pdf)}

Supplement 2.

eTable 4. Time Point 1 Model Parameters

eTable 5. Time Point 2 Model Parameters

eTable 6. Conditional-Longitudinal Model Parameters

jamanetwopen-e2627643-s002.xlsx^{(142.4KB, xlsx)}

Supplement 3.

eTable 7. Regressions With Time Point 1 Centiles

eTable 8. Regressions With Time Point 2 Centiles

eTable 9. Regressions With Conditional-Longitudinal Centiles

jamanetwopen-e2627643-s003.xlsx^{(376KB, xlsx)}

Supplement 4.

Data Sharing Statement

jamanetwopen-e2627643-s004.pdf^{(17.7KB, pdf)}

References

1.Bethlehem RAI, Seidlitz J, White SR, et al. ; 3R-BRAIN; AIBL; Alzheimer’s Disease Neuroimaging Initiative; Alzheimer’s Disease Repository Without Borders Investigators; CALM Team; Cam-CAN; CCNP; COBRE; cVEDA; ENIGMA Developmental Brain Age Working Group; Developing Human Connectome Project; FinnBrain; Harvard Aging Brain Study; IMAGEN; KNE96; Mayo Clinic Study of Aging; NSPN; POND; PREVENT-AD Research Group; VETSA . Brain charts for the human lifespan. Nature. 2022;604(7906):525-533. doi: 10.1038/s41586-022-04554-y [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Rutherford S, Barkema P, Tso IF, et al. Evidence for embracing normative modeling. Elife. 2023;12:e85082. doi: 10.7554/eLife.85082 [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Dorfschmidt L, White S, Gardner M, et al. Charting structural brain asymmetry across the human lifespan. bioRxiv. Preprint posted online July 24, 2025. doi: 10.1101/2025.07.21.665924 [DOI]
4.Di Biase MA, Tian YE, Bethlehem RAI, et al. Mapping human brain charts cross-sectionally and longitudinally. Proc Natl Acad Sci U S A. 2023;120(20):e2216798120. doi: 10.1073/pnas.2216798120 [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Ge R, Yu Y, Qi YX, et al. ; ENIGMA Lifespan Working Group . Normative modelling of brain morphometry across the lifespan with CentileBrain: algorithm benchmarking and model optimisation. Lancet Digit Health. 2024;6(3):e211-e221. doi: 10.1016/S2589-7500(23)00250-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Rehak Buckova B, Fraza C, Rehák R, et al. Using normative models pre-trained on cross-sectional data to evaluate intra-individual longitudinal changes in neuroimaging data. Elife. 2025;13:RP95823. doi: 10.7554/eLife.95823.4 [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Wei Y, He X. Conditional growth charts. Ann Stat. 2006;34(5):2069-2097. doi: 10.1214/009053606000000623 [DOI] [Google Scholar]
8.Cole TJ. The development of growth references and growth charts. Ann Hum Biol. 2012;39(5):382-394. doi: 10.3109/03014460.2012.694475 [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Casey BJ, Cannonier T, Conley MI, et al. ; ABCD Imaging Acquisition Workgroup . The Adolescent Brain Cognitive Development (ABCD) Study: imaging acquisition across 21 sites. Dev Cogn Neurosci. 2018;32:43-54. doi: 10.1016/j.dcn.2018.03.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
10.van Soelen ILC, Brouwer RM, Peper JS, et al. Effects of gestational age and birth weight on brain volumes in healthy 9 year-old children. J Pediatr. 2010;156(6):896-901. doi: 10.1016/j.jpeds.2009.12.052 [DOI] [PubMed] [Google Scholar]
11.Bjuland KJ, Rimol LM, Løhaugen GCC, Skranes J. Brain volumes and cognitive function in very-low-birth-weight (VLBW) young adults. Eur J Paediatr Neurol. 2014;18(5):578-590. doi: 10.1016/j.ejpn.2014.04.004 [DOI] [PubMed] [Google Scholar]
12.Karolis VR, Froudist-Walsh S, Kroll J, et al. Volumetric grey matter alterations in adolescents and adults born very preterm suggest accelerated brain maturation. Neuroimage. 2017;163:379-389. doi: 10.1016/j.neuroimage.2017.09.039 [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Wheater E, Shenkin SD, Muñoz Maniega S, et al. Birth weight is associated with brain tissue volumes seven decades later but not with MRI markers of brain ageing. Neuroimage Clin. 2021;31:102776. doi: 10.1016/j.nicl.2021.102776 [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Nivins S, Kennedy E, McKinlay C, Thompson B, Harding JE; Children with Hypoglycemia and Their Later Development (CHYLD) Study Team . Size at birth predicts later brain volumes. Sci Rep. 2023;13(1):12446. doi: 10.1038/s41598-023-39663-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Thompson DK, Matthews LG, Alexander B, et al. Tracking regional brain growth up to age 13 in children born term and very preterm. Nat Commun. 2020;11(1):696. doi: 10.1038/s41467-020-14334-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Kelly CE, Thompson DK, Adamson CL, et al. Cortical growth from infancy to adolescence in preterm and term-born children. Brain. 2024;147(4):1526-1538. doi: 10.1093/brain/awad348 [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Walhovd KB, Krogsrud SK, Amlien IK, et al. Fetal influence on the human brain through the lifespan. Elife. 2024;12:RP86812. [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Thalhammer M, Seidlitz J, Neubauer A, et al. Heterogeneous, temporally consistent, and plastic brain development after preterm birth. Nat Commun. 2025;16(1):8269. doi: 10.1038/s41467-025-63967-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Abel KM, Wicks S, Susser ES, et al. Birth weight, schizophrenia, and adult mental disorder: is risk confined to the smallest babies? Arch Gen Psychiatry. 2010;67(9):923-930. doi: 10.1001/archgenpsychiatry.2010.100 [DOI] [PubMed] [Google Scholar]
20.Dooley N, Clarke M, Cotter D, Cannon M. Birth weight and childhood psychopathology in the ABCD cohort: association is strongest for attention problems and is moderated by sex. Res Child Adolesc Psychopathol. 2022;50(5):563-575. doi: 10.1007/s10802-021-00859-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Dooley N, Healy C, Cotter D, Clarke M, Cannon M. The persistent effects of foetal growth on child and adolescent mental health: longitudinal evidence from a large population-based cohort. Eur Child Adolesc Psychiatry. 2023;32(10):2067-2076. doi: 10.1007/s00787-022-02045-z [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Balit J, Collet O, Yang S, et al. Preterm birth and risk of psychiatric disorders: a register-linkage cohort study: liens entre la naissance prématurée et le risque de troubles psychiatriques: une étude de cohorte avec couplage de registres. Can J Psychiatry. 2026;71(2):93-105. doi: 10.1177/07067437251389872 [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Kang K, Seidlitz J, Bethlehem RAI, et al. Study design features increase replicability in brain-wide association studies. Nature. 2024;636(8043):719-727. doi: 10.1038/s41586-024-08260-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Paus T, Keshavan M, Giedd JN. Why do many psychiatric disorders emerge during adolescence? Nat Rev Neurosci. 2008;9(12):947-957. doi: 10.1038/nrn2513 [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Casey BJ, Jones RM. Neurobiology of the adolescent brain and behavior: implications for substance use disorders. J Am Acad Child Adolesc Psychiatry. 2010;49(12):1189-1201. doi: 10.1097/00004583-201012000-00005 [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Moore TM, Kaczkurkin AN, Durham EL, et al. Criterion validity and relationships between alternative hierarchical dimensional models of general and specific psychopathology. J Abnorm Psychol. 2020;129(7):677-688. doi: 10.1037/abn0000601 [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Romer AL, Ren B, Pizzagalli DA. Brain structure relations with psychopathology trajectories in the ABCD Study. J Am Acad Child Adolesc Psychiatry. 2023;62(8):895-907. doi: 10.1016/j.jaac.2023.02.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Oostermeijer S, Whittle S, Suo C, et al. Trajectories of adolescent conduct problems in relation to cortical thickness development: a longitudinal MRI study. Transl Psychiatry. 2016;6(6):e841. doi: 10.1038/tp.2016.111 [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Bos MGN, Peters S, van de Kamp FC, Crone EA, Tamnes CK. Emerging depression in adolescence coincides with accelerated frontal cortical thinning. J Child Psychol Psychiatry. 2018;59(9):994-1002. doi: 10.1111/jcpp.12895 [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Desikan RS, Ségonne F, Fischl B, et al. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage. 2006;31(3):968-980. doi: 10.1016/j.neuroimage.2006.01.021 [DOI] [PubMed] [Google Scholar]
31.Fischl B, Salat DH, Busa E, et al. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron. 2002;33(3):341-355. doi: 10.1016/S0896-6273(02)00569-X [DOI] [PubMed] [Google Scholar]
32.Hafen R. growthstandards: Anthropometric growth standard calculations. 2025. Accessed July 20, 2026. https://ki-tools.github.io/growthstandards/
33.Villar J, Cheikh Ismail L, Victora CG, et al. ; International Fetal and Newborn Growth Consortium for the 21st Century (INTERGROWTH-21st) . International standards for newborn weight, length, and head circumference by gestational age and sex: the Newborn Cross-Sectional Study of the INTERGROWTH-21st Project. Lancet. 2014;384(9946):857-868. doi: 10.1016/S0140-6736(14)60932-6 [DOI] [PubMed] [Google Scholar]
34.Kullo IJ, Lewis CM, Inouye M, Martin AR, Ripatti S, Chatterjee N. Polygenic scores in biomedical research. Nat Rev Genet. 2022;23(9):524-532. doi: 10.1038/s41576-022-00470-z [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Ge T, Chen CY, Ni Y, Feng YA, Smoller JW. Polygenic prediction via Bayesian regression and continuous shrinkage priors. Nat Commun. 2019;10(1):1776. doi: 10.1038/s41467-019-09718-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Warrington NM, Beaumont RN, Horikoshi M, et al. ; EGG Consortium . Maternal and fetal genetic effects on birth weight and their relevance to cardio-metabolic risk factors. Nat Genet. 2019;51(5):804-814. doi: 10.1038/s41588-019-0403-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Middeldorp CM, Felix JF, Mahajan A, McCarthy MI; EArly Genetics Lifecourse Epidemiology (EAGLE) consortium; Early Growth Genetics (EGG) consortium . The Early Growth Genetics (EGG) and EArly Genetics and Lifecourse Epidemiology (EAGLE) consortia: design, results and future prospects. Eur J Epidemiol. 2019;34(3):279-300. doi: 10.1007/s10654-019-00502-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Rigby RA, Stasinopoulos DM. Generalized additive models for location, scale and shape. J R Stat Soc Ser C Appl Stat. 2005;54(3):507-554. doi: 10.1111/j.1467-9876.2005.00510.x [DOI] [Google Scholar]
39.Cordova M, Doyle O, Conan G, et al. ABCD Reproducible Matched Samples (ARMS) software. October 5, 2020. Accessed August 22, 2025. https://osf.io/7xn4f/
40.Feczko E, Conan G, Marek S, et al. Adolescent Brain Cognitive Development (ABCD) community MRI collection and utilities. bioRxiv. Preprint posted online July 11, 2021. doi: 10.1101/2021.07.09.451638 [DOI]
41.Mowinckel AM, Vidal-Piñeiro D. Visualisation of brain statistics with R-packages ggseg and ggseg3d. arXiv. Preprint posted online December 17, 2019.
42.Manuck TA, Rice MM, Bailit JL, et al. ; Eunice Kennedy Shriver National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network . Preterm neonatal morbidity and mortality by gestational age: a contemporary cohort. Am J Obstet Gynecol. 2016;215(1):103.e1-103.e14. doi: 10.1016/j.ajog.2016.01.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Rimol LM, Bjuland KJ, Løhaugen GCC, et al. Cortical trajectories during adolescence in preterm born teenagers with very low birthweight. Cortex. 2016;75:120-131. doi: 10.1016/j.cortex.2015.12.001 [DOI] [PubMed] [Google Scholar]
44.El Marroun H, Zou R, Leeuwenburg MF, et al. Association of gestational age at birth with brain morphometry. JAMA Pediatr. 2020;174(12):1149-1158. doi: 10.1001/jamapediatrics.2020.2991 [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Ma Q, Wang H, Rolls ET, et al. Lower gestational age is associated with lower cortical volume and cognitive and educational performance in adolescence. BMC Med. 2022;20(1):424. doi: 10.1186/s12916-022-02627-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Schultz LM, Merikangas AK, Ruparel K, et al. Stability of polygenic scores across discovery genome-wide association studies. HGG Adv. 2022;3(2):100091. doi: 10.1016/j.xhgg.2022.100091 [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Quinn EB, Hsiao CJ, Maisha FM, Mulligan CJ. Low birthweight is associated with epigenetic age acceleration in the first 3 years of life. Evol Med Public Health. 2023;11(1):251-261. doi: 10.1093/emph/eoad019 [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Larsen B, Sydnor VJ, Keller AS, Yeo BTT, Satterthwaite TD. A critical period plasticity framework for the sensorimotor-association axis of cortical neurodevelopment. Trends Neurosci. 2023;46(10):847-862. doi: 10.1016/j.tins.2023.07.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Loh WY, Anderson PJ, Cheong JLY, et al. Longitudinal growth of the basal ganglia and thalamus in very preterm children. Brain Imaging Behav. 2020;14(4):998-1011. doi: 10.1007/s11682-019-00057-z [DOI] [PubMed] [Google Scholar]
50.Sydnor VJ, Bagautdinova J, Larsen B, et al. Human thalamocortical structural connectivity develops in line with a hierarchical axis of cortical plasticity. Nat Neurosci. 2025;28(8):1772-1786. doi: 10.1038/s41593-025-01991-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
51.van Houdt CA, Oosterlaan J, van Wassenaer-Leemhuis AG, van Kaam AH, Aarnoudse-Moens CSH. Executive function deficits in children born preterm or at low birthweight: a meta-analysis. Dev Med Child Neurol. 2019;61(9):1015-1024. doi: 10.1111/dmcn.14213 [DOI] [PMC free article] [PubMed] [Google Scholar]
52.Whittle S, Lichter R, Dennison M, et al. Structural brain development and depression onset during adolescence: a prospective longitudinal study. Am J Psychiatry. 2014;171(5):564-571. doi: 10.1176/appi.ajp.2013.13070920 [DOI] [PubMed] [Google Scholar]
53.Wang Y, Ma L, Chen R, et al. Emotional and behavioral problems change the development of cerebellar gray matter volume, thickness, and surface area from childhood to adolescence: a longitudinal cohort study. CNS Neurosci Ther. 2023;29(11):3528-3548. doi: 10.1111/cns.14286 [DOI] [PMC free article] [PubMed] [Google Scholar]
54.Royer J, Kebets V, Piguet C, et al. Multimodal neural correlates of childhood psychopathology. Elife. 2024;13:e87992. doi: 10.7554/eLife.87992 [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Sripada C, Angstadt M, Taxali A, et al. Widespread attenuating changes in brain connectivity associated with the general factor of psychopathology in 9- and 10-year olds. Transl Psychiatry. 2021;11(1):575. doi: 10.1038/s41398-021-01708-w [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Feinberg I. Schizophrenia: caused by a fault in programmed synaptic elimination during adolescence? J Psychiatr Res. 1982;17(4):319-334. doi: 10.1016/0022-3956(82)90038-3 [DOI] [PubMed] [Google Scholar]
57.Germann M, Brederoo SG, Sommer IEC. Abnormal synaptic pruning during adolescence underlying the development of psychotic disorders. Curr Opin Psychiatry. 2021;34(3):222-227. doi: 10.1097/YCO.0000000000000696 [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Tanner JM, Whitehouse RH. Clinical longitudinal standards for height, weight, height velocity, weight velocity, and stages of puberty. Arch Dis Child. 1976;51(3):170-179. doi: 10.1136/adc.51.3.170 [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Elyounssi S, Kunitoki K, Clauss JA, et al. Addressing artifactual bias in large, automated MRI analyses of brain development. Nat Neurosci. 2025;28(8):1787-1796. doi: 10.1038/s41593-025-01990-7 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials