Infant subcortical brain volumes associated with maternal obesity and diabetes: a large multicohort human study Authors: Ann M. Alex, Jerod M. Rasmussen, Jetro J. Tuulari, Julie Nihouarn Sigurdardottir, Claudia Buss, Kirsten A. Donald, A. David Edwards, Sonja Entringer, John H. Gilmore, Nynke A. Groenewold, Hasse Karlsson, Linnea Karlsson, Katherine E. Lawrence, Inka Mattila, Dan J. Stein, Martin Styner, Paul M. Thompson, Pathik D. Wadhwa, Heather J. Zar, Xi Zhu, Gustavo de los Campos, Rebecca C. Knickmeyer, Shan Luo, ENIGMA-ORIGINs working group Published: 2026 ABSTRACT BackgroundMaternal diabetes and obesity are established risk factors for adverse offspring health. Emerging evidence suggests that these fetal programming effects vary by sex, yet it remains unclear whether these factors independently or interactively influence early brain development.MethodsThis prospective study included 1,965 infants from six international cohorts. Infant MRI was used to derive subcortical volumes (thalamus, amygdala, hippocampus, pallidum, putamen, caudate). ComBat harmonization was applied. Multiple linear regression tested main and interaction effects of maternal obesity, maternal diabetes, and sex, controlling for covariates with false discovery rate (FDR) corrections.ResultsOf the sample, 46% were female (N = 909), 9% were exposed to maternal diabetes (N = 172), 22% to maternal obesity (N = 386), and 3% to both (N = 61). MRI scans were performed at 25.9 ± 18.8 days. Maternal diabetes was associated with smaller thalamic volume (standardized β = -0.09, 95%CI –0.16 to -0.01, FDR P = 0.020), but this association was attenuated after adjusting for maternal obesity. Maternal obesity was associated with smaller hippocampal (standardized β = -0.13, 95%CI -0.21 to –0.05, FDR P = 0.009) and thalamic volumes (standardized β = -0.09, 95%CI -0.14 to –0.03, FDR P = 0.007).Sex-specific associations were observed. In females, maternal obesity was associated with smaller hippocampal (standardized β = -0.24, 95%CI -0.36 to -0.13, FDR P < 0.001) and amygdala volumes (standardized β = -0.18, 95% CI = -0.30 to –0.06, FDR P = 0.016). A three-way interaction (diabetes x obesity x sex) was observed for thalamus volume (standardized β = -0.50, 95%CI -0.81 to -0.18, FDR P = 0.017). In males, combined exposure had lower thalamic volume compared to those with one or neither exposure (all Ps < 0.05). In females, maternal obesity (standardized β = -0.12, 95%CI -0.20 to –0.04, FDR P = 0.015) and diabetes (standardized β = -0.16, 95%CI -0.30 to –0.02, FDR P = 0.042) showed independent associations with thalamic volume, without a significant interaction (standardized β = 0.23, 95%CI -0.01 to 0.46, FDR P = 0.092).ConclusionsMaternal obesity shows stronger associations with infant subcortical volumes than maternal diabetes. It is associated with smaller hippocampal and amygdala volumes in females, while combined exposure to maternal diabetes and obesity is associated with smaller thalamic volumes in males. These findings highlight the role of maternal metabolic health and infant sex in early neurodevelopment.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12916-026-04860-6. Background The intrauterine environment plays a critical role in fetal brain development [1, 2], which begins around the second gestational week and continues to birth [3]. Several conditions during pregnancy including obesity and diabetes have been shown to confer increased health risks in offspring [4–7] but their neurological contributions have not been robustly established. Preliminary studies suggest an influence of maternal diabetes exposure on the brain. In rats, offspring born to mothers with diabetes show morphometric malformations of the hypothalamus [8] while in humans, exposure to maternal diabetes prenatally is associated with hypothalamic gliosis [9, 10], smaller subcortical structures (e.g., thalamus, caudate, putamen), lower cortical thickness (CT) [11] and smaller cortical gray matter (GM) volume during late childhood [12]. However, the existing studies have limitations. First, studies in humans have been limited to later periods of life (ages 7 – 11 years) making it difficult to disentangle the contributing effects of prenatal exposures from postnatal environmental factors [9–11, 13–16]. Second, maternal obesity in pregnancy has also been associated with alterations in offspring brains at birth and during infancy [17–23] but prior studies have not disentangled the effects of maternal obesity from those driven independently by maternal diabetes [10, 11, 16, 24–26] and those resulting from combined diabetes and obesity exposure. Furthermore, there is a strong rationale for investigating sex-specific effects in the context of maternal obesity/maternal diabetes as fetal development constitutes a critical period [27] during which steroid hormone exposure drives sex-specific maternal placental function [28], and epigenetic modification [29]. Indeed, these mechanisms underpin observed sex-specific differences in offspring brain outcomes including rodent models demonstrating maternal diabetes/maternal obesity-related sex-specific gene expression changes [30–32] in the brain, and human studies identifying sex-specific associations between maternal obesity and offspring hippocampal volume during childhood [33]. Yet, most human clinical studies to date [9, 13–15, 24–26] have used small sample sizes (N < 160) and homogeneous populations, calling into question the reproducibility and generalizability of findings. To address these knowledge gaps, we leveraged mega-analytics using a large international cohort (N = 1,965) to investigate associations between prenatal exposure to maternal diabetes and/or maternal obesity and infant subcortical volumes, testing for moderating effects of sex. Based on prior studies [11, 14, 16], we hypothesized that prenatal exposure to maternal diabetes or maternal obesity would be associated with smaller subcortical region volumes, and combined exposure would exacerbate these effects. We also hypothesized that these associations may vary in a sex-dependent manner. Methods Participant characteristics The ENIGMA-ORIGINs working group [34, 35] contributed data from five cohorts: the Drakenstein Child Health Study (South Africa), UNC Early Brain Development Study (EBDS), UNC Gut Microbiome and Infant Anxiety (GMIA), the University of California, Irvine, USA, and the developing Human Connectome Project (UK). Additional data was provided by FinnBrain (Finland). Approval was obtained from respective local ethical review boards and informed consent from parents/legal guardians were obtained before data collection. Cohort characteristics Drakenstein Child Health Study (DCHS) This cohort was established to investigate the determinants of child growth, health, and development in a stable, semi-urban, low socioeconomic status community in South Africa. Exclusion criteria were minimal and focused on unavailability for follow-up. Detailed information is available elsewhere [36]. UNC Early Brain Development Study (EBDS) The cohort included an exceptionally large sample of typically developing infants, children at risk for schizophrenia and bipolar illness, a "structural" high risk group (children with prenatal isolated mild ventriculomegaly), and a sample of twins. Major medical illness in the mother, abnormality on ultrasound, and current substance abuse were exclusion criteria at enrollment. Detailed information is available elsewhere [37–39]. UNC Gut Microbiome and Infant Anxiety (GMIA) This is a prospective, longitudinal study of the microbiome-gut-brain axis. Recruitment was restricted to infants without major dietary or medical interventions. Inclusion criteria were infants born at term, vaginally delivered, and exclusively breastfed for the first 2 weeks. Exclusion criteria were admission to neonatal intensive care unit, weighing < 2500 g at birth, receiving antibiotics in the first 2 weeks, or metal in the body. Infants were excluded if the mother received antibiotics in the two weeks before delivery. Detailed information is available elsewhere [40]. University of California, Irvine This is a prospective, longitudinal, follow-up study in a population-based cohort. Exclusion criteria included preterm birth (< 34 completed weeks gestation), psychotropic medication, corticosteroids, smoking, and drug use during pregnancy (self-reports verified by urinary cotinine and drug toxicology), congenital or genetic disorders, and major neurologic disorder at birth. Detailed information is available elsewhere [34]. Developing Human Connectome Project (dHCP) This project aims to study neurodevelopment in a typical urban population sample from 20 to 44 weeks post-conception. Exclusion criteria include mothers or infants with contraindications to MR imaging, preterm infants too unwell to undergo scanning, and language barriers to obtaining consent. Detailed information is available elsewhere [41]. FinnBrain This is a birth cohort study recruiting Mother-infant dyads in Finland. Inclusion criteria were gestational age at birth ≥ 35 weeks and birth weight > 1500 g. Previously diagnosed central nervous system anomalies or abnormal findings in a previous MRI scan were exclusion criteria. Detailed information is available elsewhere [42, 43]. Imaging measures Each site acquired and processed structural 3D T1-weighted and T2-weighted volumetric brain MRI scans of participants. Images were acquired at a field strength of 3 T. Quality control was performed locally. Acquisition parameters and segmentation protocol details are provided in Additional File 1: eTable 1. Exposure Information on maternal diabetes exposure during pregnancy including gestational diabetes mellitus, pre-gestational type 1 diabetes, and pre-gestational type 2 diabetes was either extracted from electronic medical records (EMR; in the EBDS, UC Irvine, dHCP, GMIA, and FinnBrain cohorts) and/or self-reported. Only the UC Irvine and FinnBrain cohorts distinguished the type of MD. Pre-pregnancy BMI ≥ 30 kg/m2 was used to classify maternal obesity, and < 30 kg/m2 classified as non-obese. Statistical analysis Participants with missing values (≤ 1.3% overall, e.g., maternal education: 1.32%, gestational age: 0.05%, birthweight: 0.05%) were excluded using listwise deletion to maintain consistency across models. Data harmonization To investigate the effects of maternal diabetes and maternal obesity exposure on neonatal brain volumes, we used ComBat-harmonized linear regression models. ComBat harmonization, implemented via the sva package [44] in R, was applied to adjust for site-specific differences while preserving biological variability associated with exposures and covariates. Scanner upgrades within sites (i.e., EBDS, GMIA) were not treated as distinct sites to preserve statistical power and avoid over-parameterization. Although imaging processing pipelines differed across cohorts and regions were not registered to a single common atlas, regional measures were defined using standardized anatomical parcellations within each site. Regions with identical anatomical labels were treated as homologous across cohorts and harmonized using ComBat. Statistical models We used the harmonized data to test for the association between harmonized brain volume phenotypes (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}_{i}$$\end{document}Yi) and prenatal exposures. First, we tested for the association of brain volumes with exposure to maternal diabetes (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M{D}_{i}$$\end{document}MDi, 1 for diabetic mother, 0 otherwise) controlling for child’s age (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Ag{e}_{i})$$\end{document}Agei), sex (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Se{x}_{i}$$\end{document}Sexi, 1 if male, 0 for female), pre-term (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P{T}_{i}$$\end{document}PTi 1 if pre-term), low-birthweight (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$LB{W}_{i}$$\end{document}LBWi, 1 if low birthweight), and maternal education (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$MED{U}_{i}$$\end{document}MEDUi, 1 primary and secondary, 0 for tertiary). Thus, our first model was Model 1 (Maternal Diabetes): \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}_{ij}={\beta }_{0} +Ag{e}_{ij}{\times \beta }_{1}+Se{x}_{i}\times {\beta }_{2}+P{T}_{i}\times {\beta }_{3}+LB{W}_{i}\times {\beta }_{4}+MED{U}_{i}\times {\beta }_{5}+M{D}_{i}{\beta }_{6}+{\varepsilon }_{i}.$$\end{document}Yij=β0+Ageij×β1+Sexi×β2+PTi×β3+LBWi×β4+MEDUi×β5+MDiβ6+εi. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}_{ij}={\beta }_{0} +Ag{e}_{ij}{\times \beta }_{1}+Se{x}_{i}\times {\beta }_{2}+P{T}_{i}\times {\beta }_{3}+LB{W}_{i}\times {\beta }_{4}+MED{U}_{i}\times {\beta }_{5}+M{D}_{i}{\beta }_{6}+{\varepsilon }_{i}.$$\end{document}Yij=β0+Ageij×β1+Sexi×β2+PTi×β3+LBWi×β4+MEDUi×β5+MDiβ6+εi. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}_{ij}={\beta }_{0} +Ag{e}_{ij}{\times \beta }_{1}+Se{x}_{i}\times {\beta }_{2}+P{T}_{i}\times {\beta }_{3}+LB{W}_{i}\times {\beta }_{4}+MED{U}_{i}\times {\beta }_{5}+M{D}_{i}{\beta }_{6}+{\varepsilon }_{i}.$$\end{document}Yij=β0+Ageij×β1+Sexi×β2+PTi×β3+LBWi×β4+MEDUi×β5+MDiβ6+εi. To examine independent effects of maternal diabetes and maternal obesity exposure, we expanded Model 1 by adding the effect of exposure to maternal obesity (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M{O}_{i}$$\end{document}MOi, 1 if obese, 0 otherwise)–we refer to this model as Maternal Obesity + Maternal Diabetes. Model 2 (Maternal Obesity + Maternal Diabetes): \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}_{ij}={\beta }_{0} +Ag{e}_{ij}{\times \beta }_{1}+Se{x}_{i}\times {\beta }_{2}+P{T}_{i}\times {\beta }_{3}+LB{W}_{i}\times {\beta }_{4}+MED{U}_{i}\times {\beta }_{5}+M{D}_{i}{\beta }_{6}+M{O}_{i}{\beta }_{7} +{\varepsilon }_{i}.$$\end{document}Yij=β0+Ageij×β1+Sexi×β2+PTi×β3+LBWi×β4+MEDUi×β5+MDiβ6+MOiβ7+εi. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}_{ij}={\beta }_{0} +Ag{e}_{ij}{\times \beta }_{1}+Se{x}_{i}\times {\beta }_{2}+P{T}_{i}\times {\beta }_{3}+LB{W}_{i}\times {\beta }_{4}+MED{U}_{i}\times {\beta }_{5}+M{D}_{i}{\beta }_{6}+M{O}_{i}{\beta }_{7} +{\varepsilon }_{i}.$$\end{document}Yij=β0+Ageij×β1+Sexi×β2+PTi×β3+LBWi×β4+MEDUi×β5+MDiβ6+MOiβ7+εi. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}_{ij}={\beta }_{0} +Ag{e}_{ij}{\times \beta }_{1}+Se{x}_{i}\times {\beta }_{2}+P{T}_{i}\times {\beta }_{3}+LB{W}_{i}\times {\beta }_{4}+MED{U}_{i}\times {\beta }_{5}+M{D}_{i}{\beta }_{6}+M{O}_{i}{\beta }_{7} +{\varepsilon }_{i}.$$\end{document}Yij=β0+Ageij×β1+Sexi×β2+PTi×β3+LBWi×β4+MEDUi×β5+MDiβ6+MOiβ7+εi. Subsequently, we evaluated possible interactions between exposures and infant sex by adding interactions terms to Model 2. Model 3 (Maternal Obesity × Maternal Diabetes): This model included, in addition to the effects in Model 2, an interaction term between maternal obesity and diabetes. Models 4 and 5 (Sex × Maternal Obesity and Sex × Maternal Diabetes, respectively): We tested first-order interactions of sex with each exposure by including either interaction terms in the model. Model 6 (Sex × Maternal Obesity × Maternal Diabetes): Finally, we tested a model that included, in addition to the predictors in Maternal Obesity + Maternal Diabetes all two-way interactions between Sex, maternal obesity and maternal diabetes, plus the three-way interaction between Sex and the two prenatal maternal exposures. The models described above were used for intracranial volume (ICV), for sub-cortical volumes we also included in the model the effect of ICV as a covariate. To determine significance while controlling for multiple testing, we used a (within model) false discovery rate (FDR) of 0.05, with FDR computed using Benjamini–Hochberg’s (BH) method [45]. When an exposure-by-sex interaction was found to be significant, sex-stratified analyses were performed. Pairwise comparisons between groups (defined by combinations of multiple exposures) were performed using least-squares means method (lsmeans) and emmeans package [46]. Sensitivity analyses We performed eight sensitivity analyses to check the robustness of the results. Sensitivity analysis 1: During the first 100 days, when the brain is developing at an exponential rate, ICV and subcortical structures may have unique growth patterns, and adjusting for ICV may have different effects on each regional volume. To address this, we performed a sensitivity analysis by removing ICV from the models for subcortical volumes. Sensitivity analysis 2: We removed birthweight from the main model because birthweight is correlated with gestational term. Sensitivity analysis 3: To assess potential collider bias from adjusting for gestational age and birthweight—variables that may act as mediators—we conducted a sensitivity analysis excluding these covariates. Sensitivity analysis 4: Since preterm children constitute around a quarter of the sample and Maternal Obesity and Maternal Diabetes are known risk factors for preterm birth, we conducted separate analyses in both full-term-only and preterm-only subsamples. For preterm-only models, the putamen and pallidum regions were excluded due to insufficient sample size to support stable model estimation. Furthermore, we performed mediation analysis to assess whether preterm birth mediated the association between exposures and neonatal brain volumes. These analyses were done using the mediation package [47] in R, applying a nonparametric bootstrap approach with percentile confidence intervals to evaluate preterm birth as a potential mediator of the association between maternal obesity and neonatal brain volumes. To account for differences in developmental maturity at the time of imaging, we repeated the primary model, replacing postnatal age at MRI with postmenstrual age (PMA; defined as gestational age at birth plus postnatal age at scan) as the age covariate in a subset of children where PMA data were available. Of the 436 (23.7%) preterm infants, 291 (15.8%) were late preterm (34–36 weeks), 80 (4.4%) were moderate preterm (32–33 weeks), 59 (3.2%) were very preterm (28–31 weeks), and 6 (0.3%) were extremely preterm (< 28 weeks). Sensitivity analysis 5: To account for residual socioeconomic confounding, in a subset of participants with available family income data, we repeated the analysis by including income as an additional covariate in the regression models. Sensitivity analysis 6: We replaced the categorical maternal obesity variable with maternal pregnancy BMI as a continuous predictor to evaluate linear exposure–response relationships. Sensitivity analysis 7: The DCHS cohort did not specifically have pre-pregnancy BMI, but rather maternal BMI. To investigate if the effects of exposure to pre-pregnancy BMI were similar to exposure to maternal BMI, we performed analyses excluding the DCHS cohort from the dataset. Sensitivity analysis 8: We conducted a cohort-wise analysis to assess the replicability of results. Results A total of 1,965 participants from 6 international cohorts were included (Table 1). Around 46% of the participants were female, ~ 9% of infants were exposed to maternal diabetes, ~ 22% were exposed to maternal obesity, and ~ 3% were exposed to both maternal diabetes and maternal obesity prenatally. MRI scans were performed between 1 and 99 days old (mean = 25.87; SD = 18.77). The age distribution is shown in Additional File 2: eFigure 1. Across the full sample, the mean (SD) volumes were: ICV 468.83 (63.41) cm3; hippocampus 1996.41 (250.89) mm3; amygdala 746.43 (105.34) mm3; thalamus 8470.45 (1069.77) mm3; caudate 3500.757 (522.24) mm3; putamen 4136.9 (326.63) mm3; and pallidum 918.73 (63.93) mm3. Table 1Socio-demographic and birth outcome distribution of participantsCohortAllCape TowndHCPFinnBrainGMIAUC IrvineEBDSN196513960512510683907Age at MRI scan, days (mean ± SD)25.86 ± 18.7625.86 ± 18.7625.86 ± 18.7625.86 ± 18.7625.86 ± 18.7625.86 ± 18.7625.86 ± 18.76SexMale1056 (53.7%)71 (51.1%)322 (55.0%)70(56.0%)55 (51.9%)49 (59.0%)489 (53.9%)Female909 (46.3%)68 (48.9%)283 (45.0%)55(44.0%)51 (48.1%)34 (41.0%)418 (46.1%)Diabetes ExposureYes172 (8.8%)1(0.7%)49(8.1%)16(13.0%)5(4.7%)6(7.2%)95 (10.5%)No1790 (91.2%)138 (99.3%)556 (91.9%)107 (87.0%)101 (95.3%)77 (92.8%)811 (89.4%)Obesity ExposureYes386 (21.1%)51 (37.2%)74(12.2%)13(10.4%)10(9.5%)21 (25.3%)217 (27.8%)No1447 (78.9%)86 (62.8%)531 (87.8%)110 (89.6%)95 (94.5%)62 (74.7%)563 (62.2%)GestationPreterm492 (25.1%)12(8.6%)137 (22.6%)0(0.0%)0(0.0%)6 (7.3%)337 (37.2%)Full-term1472 (74.9%)127 (91.4%)468 (77.4%)125 (100.0%)106 (100.0%)76 (92.7%)570 (62.8%)BirthweightLow427 (21.7%)6(4.3%)136 (22.5%)0(0.0%)0(0.0%)4(4.8%)281 (31.0%)Normal1537 (78.3%)133 (95.7%)469 (77.5%)125 (100.0%)106 (100.0%)79 (95.2%)625 (69.0%)Maternal EducationPrimary-Secondary580 (29.9%)132 (95.0%)34(5.6%)72(59.5%)13 (12.3%)56 (67.5%)302 (33.3%)Tertiary1359 (70.1%)7(5.0%)549 (94.4%)49(40.5%)93 (87.7%)27 (32.5%)605 (66.7%)Family IncomeLow518 (40.5%)47 (33.8%)NA44(36.4%)28 (26.4%)45 (57.0%)354 (42.4%)Medium–High761 (59.5%)92 (66.2%)77(63.6%)78 (73.6%)34 (43.0%)480 (57.6%) Table 1Socio-demographic and birth outcome distribution of participantsCohortAllCape TowndHCPFinnBrainGMIAUC IrvineEBDSN196513960512510683907Age at MRI scan, days (mean ± SD)25.86 ± 18.7625.86 ± 18.7625.86 ± 18.7625.86 ± 18.7625.86 ± 18.7625.86 ± 18.7625.86 ± 18.76SexMale1056 (53.7%)71 (51.1%)322 (55.0%)70(56.0%)55 (51.9%)49 (59.0%)489 (53.9%)Female909 (46.3%)68 (48.9%)283 (45.0%)55(44.0%)51 (48.1%)34 (41.0%)418 (46.1%)Diabetes ExposureYes172 (8.8%)1(0.7%)49(8.1%)16(13.0%)5(4.7%)6(7.2%)95 (10.5%)No1790 (91.2%)138 (99.3%)556 (91.9%)107 (87.0%)101 (95.3%)77 (92.8%)811 (89.4%)Obesity ExposureYes386 (21.1%)51 (37.2%)74(12.2%)13(10.4%)10(9.5%)21 (25.3%)217 (27.8%)No1447 (78.9%)86 (62.8%)531 (87.8%)110 (89.6%)95 (94.5%)62 (74.7%)563 (62.2%)GestationPreterm492 (25.1%)12(8.6%)137 (22.6%)0(0.0%)0(0.0%)6 (7.3%)337 (37.2%)Full-term1472 (74.9%)127 (91.4%)468 (77.4%)125 (100.0%)106 (100.0%)76 (92.7%)570 (62.8%)BirthweightLow427 (21.7%)6(4.3%)136 (22.5%)0(0.0%)0(0.0%)4(4.8%)281 (31.0%)Normal1537 (78.3%)133 (95.7%)469 (77.5%)125 (100.0%)106 (100.0%)79 (95.2%)625 (69.0%)Maternal EducationPrimary-Secondary580 (29.9%)132 (95.0%)34(5.6%)72(59.5%)13 (12.3%)56 (67.5%)302 (33.3%)Tertiary1359 (70.1%)7(5.0%)549 (94.4%)49(40.5%)93 (87.7%)27 (32.5%)605 (66.7%)Family IncomeLow518 (40.5%)47 (33.8%)NA44(36.4%)28 (26.4%)45 (57.0%)354 (42.4%)Medium–High761 (59.5%)92 (66.2%)77(63.6%)78 (73.6%)34 (43.0%)480 (57.6%) Socio-demographic and birth outcome distribution of participants 70 (56.0%) 55 (44.0%) 1 (0.7%) 49 (8.1%) 16 (13.0%) 5 (4.7%) 6 (7.2%) 74 (12.2%) 13 (10.4%) 10 (9.5%) 12 (8.6%) 0 (0.0%) 0 (0.0%) 6 (4.3%) 0 (0.0%) 0 (0.0%) 4 (4.8%) 34 (5.6%) 72 (59.5%) 7 (5.0%) 49 (40.5%) 44 (36.4%) 77 (63.6%) Main effects of maternal diabetes and maternal obesity Maternal diabetes Prenatal exposure to maternal diabetes was negatively associated with infant thalamus volume (standardized β = −0.09, 95% CI −0.16 to −0.01; FDR P = 0.020) only before controlling maternal obesity exposure. Associations of maternal diabetes exposure with ICV and subcortical volumes were not significant, after adjusting for maternal obesity (Table 2, Additional File 1: Extended Table 2, Additional File 1: eTable 2). Table 2Standardized regression coefficients for exposures (CI) and FDR p-values by model and brain volumesModelEffects*ICVHippocampusAmygdalaThalamusMDMaternal diabetes−0.12(−0.25 – 0.01);0.063−0.11(−0.22 – 0.00);0.060(−0.12 – 0.12);0.991−0.09(−0.16 – −0.01);0.02MO + MDMaternal obesity−0.16(−0.25 – −0.07);0.007−0.13(−0.21 – −0.05);0.009−0.02(−0.10 – 0.07);0.924−0.09(−0.14 – −0.03);0.007Maternal diabetes−0.06(−0.20 – 0.07);0.628−0.08(−0.20 – 0.04);0.566−0.01(−0.14 – 0.11);0.924−0.07(−0.14 – 0.01);0.291MO × MDMaternal obesity−0.14(−0.24 – −0.04);0.035−0.15(−0.24 – −0.06);0.021−0.03(−0.12 – 0.06);0.745−0.08(−0.13 – −0.02);0.035Maternal diabetes−0.01(−0.18 – 0.16);0.921−0.14(−0.29 – 0.01);0.383−0.06(−0.22 – 0.09);0.745−0.05(−0.14 – 0.05);0.745Maternal obesity × Maternal diabetes−0.15(−0.42 – 0.13);0.7450.16(−0.09 – 0.41);0.7450.14(−0.12 – 0.39);0.745−0.05(−0.21 – 0.10);0.764Sex × MOMaternal obesity−0.21(−0.34 – −0.08);0.011−0.24(−0.36 – −0.12);0.011−0.15(−0.27 – −0.03);0.049−0.09(−0.17 – −0.02);0.049Maternal diabetes−0.07(−0.20 – 0.07);0.462−0.08(−0.20 – 0.04);0.334−0.02(−0.14 – 0.11);0.858−0.07(−0.14 – 0.01);0.243Sex Male × Maternal obesity0.1(−0.08 – 0.28);0.4220.21(0.05 – 0.38);0.0470.26(0.09 – 0.42);0.0140.02(−0.08 – 0.12);0.858Sex × MDMaternal obesity−0.16(−0.25 – −0.07);0.011−0.13(−0.21 – −0.05);0.014−0.02(−0.10 – 0.07);0.925−0.09(−0.14 – −0.03);0.011Maternal diabetes−0.06(−0.26 – 0.13);0.925−0.11(−0.28 – 0.07);0.828−0.04(−0.22 – 0.14);0.925−0.07(−0.19 – 0.04);0.773Sex Male × Maternal diabetes0(−0.26 – 0.26);0.9950.05(−0.19 – 0.29);0.9250.05(−0.19 – 0.30);0.9250.01(−0.13 – 0.16);0.935Sex × MO × MDMaternal obesity−0.23(−0.37 – −0.09);0.014−0.28(−0.41 – −0.15);0.014−0.2(−0.33 – −0.07);0.017−0.12(−0.20 – −0.04);0.021Maternal diabetes−0.09(−0.33 – 0.15);0.535−0.2(−0.42 – 0.02);0.17−0.17(−0.39 – 0.06);0.226−0.15(−0.28 – −0.01);0.143Maternal obesity × Maternal diabetes0.11(−0.30 – 0.52);0.6790.33(−0.04 – 0.70);0.1810.43(0.05 – 0.81);0.1420.22(−0.02 – 0.45);0.17Sex Male × Maternal obesity0.16(−0.03 – 0.35);0.1920.25(0.08 – 0.42);0.030.32(0.14 – 0.50);0.0140.08(−0.03 – 0.19);0.248Sex Male × Maternal diabetes0.16(−0.17 – 0.49);0.440.12(−0.18 – 0.42);0.5350.2(−0.11 – 0.51);0.3040.19(0.00 – 0.38);0.143Sex Male × Maternal obesity × Maternal diabetes−0.48(−1.03 – 0.08);0.189−0.31(−0.81 – 0.19);0.307−0.55(−1.07 – −0.04);0.143−0.5(−0.81 – −0.18);0.017Positive estimate indicates larger volume associated with Maternal Diabetes or Maternal ObesityBoldface indicate significance at FDR-Adjusted P < 0.05 Table 2Standardized regression coefficients for exposures (CI) and FDR p-values by model and brain volumesModelEffects*ICVHippocampusAmygdalaThalamusMDMaternal diabetes−0.12(−0.25 – 0.01);0.063−0.11(−0.22 – 0.00);0.060(−0.12 – 0.12);0.991−0.09(−0.16 – −0.01);0.02MO + MDMaternal obesity−0.16(−0.25 – −0.07);0.007−0.13(−0.21 – −0.05);0.009−0.02(−0.10 – 0.07);0.924−0.09(−0.14 – −0.03);0.007Maternal diabetes−0.06(−0.20 – 0.07);0.628−0.08(−0.20 – 0.04);0.566−0.01(−0.14 – 0.11);0.924−0.07(−0.14 – 0.01);0.291MO × MDMaternal obesity−0.14(−0.24 – −0.04);0.035−0.15(−0.24 – −0.06);0.021−0.03(−0.12 – 0.06);0.745−0.08(−0.13 – −0.02);0.035Maternal diabetes−0.01(−0.18 – 0.16);0.921−0.14(−0.29 – 0.01);0.383−0.06(−0.22 – 0.09);0.745−0.05(−0.14 – 0.05);0.745Maternal obesity × Maternal diabetes−0.15(−0.42 – 0.13);0.7450.16(−0.09 – 0.41);0.7450.14(−0.12 – 0.39);0.745−0.05(−0.21 – 0.10);0.764Sex × MOMaternal obesity−0.21(−0.34 – −0.08);0.011−0.24(−0.36 – −0.12);0.011−0.15(−0.27 – −0.03);0.049−0.09(−0.17 – −0.02);0.049Maternal diabetes−0.07(−0.20 – 0.07);0.462−0.08(−0.20 – 0.04);0.334−0.02(−0.14 – 0.11);0.858−0.07(−0.14 – 0.01);0.243Sex Male × Maternal obesity0.1(−0.08 – 0.28);0.4220.21(0.05 – 0.38);0.0470.26(0.09 – 0.42);0.0140.02(−0.08 – 0.12);0.858Sex × MDMaternal obesity−0.16(−0.25 – −0.07);0.011−0.13(−0.21 – −0.05);0.014−0.02(−0.10 – 0.07);0.925−0.09(−0.14 – −0.03);0.011Maternal diabetes−0.06(−0.26 – 0.13);0.925−0.11(−0.28 – 0.07);0.828−0.04(−0.22 – 0.14);0.925−0.07(−0.19 – 0.04);0.773Sex Male × Maternal diabetes0(−0.26 – 0.26);0.9950.05(−0.19 – 0.29);0.9250.05(−0.19 – 0.30);0.9250.01(−0.13 – 0.16);0.935Sex × MO × MDMaternal obesity−0.23(−0.37 – −0.09);0.014−0.28(−0.41 – −0.15);0.014−0.2(−0.33 – −0.07);0.017−0.12(−0.20 – −0.04);0.021Maternal diabetes−0.09(−0.33 – 0.15);0.535−0.2(−0.42 – 0.02);0.17−0.17(−0.39 – 0.06);0.226−0.15(−0.28 – −0.01);0.143Maternal obesity × Maternal diabetes0.11(−0.30 – 0.52);0.6790.33(−0.04 – 0.70);0.1810.43(0.05 – 0.81);0.1420.22(−0.02 – 0.45);0.17Sex Male × Maternal obesity0.16(−0.03 – 0.35);0.1920.25(0.08 – 0.42);0.030.32(0.14 – 0.50);0.0140.08(−0.03 – 0.19);0.248Sex Male × Maternal diabetes0.16(−0.17 – 0.49);0.440.12(−0.18 – 0.42);0.5350.2(−0.11 – 0.51);0.3040.19(0.00 – 0.38);0.143Sex Male × Maternal obesity × Maternal diabetes−0.48(−1.03 – 0.08);0.189−0.31(−0.81 – 0.19);0.307−0.55(−1.07 – −0.04);0.143−0.5(−0.81 – −0.18);0.017Positive estimate indicates larger volume associated with Maternal Diabetes or Maternal ObesityBoldface indicate significance at FDR-Adjusted P < 0.05 Standardized regression coefficients for exposures (CI) and FDR p-values by model and brain volumes Positive estimate indicates larger volume associated with Maternal Diabetes or Maternal Obesity Boldface indicate significance at FDR-Adjusted P < 0.05 Maternal obesity Significant negative associations were observed between maternal obesity exposure and ICV (standardized β = −0.16, 95% CI −0.25 to −0.07, FDR P = 0.007), hippocampus volumes (standardized β = −0.13, 95% CI −0.21 to −0.05, FDR P = 0.009), and thalamus volumes (standardized β = −0.09, 95% CI −0.14 to −0.03, FDR P = 0.007), independent of maternal diabetes exposure (Table 2, Additional File 1: Extended Table 2, Additional File 1: eTable 3). Interaction of maternal diabetes and maternal obesity Interactions of maternal diabetes and maternal obesity were not significant on ICV and subcortical volumes (Table 2, Additional File 1: Extended Table 2, Additional File 1: eTable 4). Sex-specific effects Interaction of maternal obesity and sex Significant maternal obesity-by-sex interactions were observed in the hippocampus (standardized β = 0.21, 95% CI 0.05 to 0.38, FDR P = 0.047) and amygdala (standardized β = 0.26, 95% CI 0.09 to 0.42, FDR P = 0.014; Table 2, Fig. 1, Additional File 1: eTable 5). When stratified by sex, maternal obesity was significantly associated with smaller hippocampal (standardized β = −0.24, 95% CI −0.36 to −0.13, FDR P < 0.001) and amygdala (standardized β = −0.18, 95% CI −0.30 to −0.06, FDR P = 0.016) volumes in female infants, whereas in male infants, main effects of maternal obesity were not observed (hippocampus: standardized β = −0.04, 95% CI −0.16 to 0.08, FDR P = 0.646; amygdala: standardized β = 0.12, 95% CI −0.01 to 0.24, FDR P = 0.174), independent of maternal diabetes (Additional File 1: eTable 6a, Additional File 1: eTable 6b).Fig. 1Pairwise comparisons of hippocampal and amygdala volume in male and female children exposed/unexposed to maternal obesity. Scatter plot showing individual hippocampal and amygdala volumes (color-coded by cohort) with model-adjusted means ± 95% confidence intervals (CI) derived from linear regression models including infant age at scan, gestational age, birthweight, maternal education, and intracranial volume as covariates. Data are ComBat-harmonized. MO- denotes unexposed to maternal obesity, and MO + denotes exposed to maternal obesity. The number of observations in each category is as follows. Male MO- (N = 701), Male MO + (N = 181), Female MO- (N = 590), Female MO + (N = 166). The pooled violin plots are displayed for visual presentation purposes, which emphasize overall distributional patterns. *** P < 0.001 Pairwise comparisons of hippocampal and amygdala volume in male and female children exposed/unexposed to maternal obesity. Scatter plot showing individual hippocampal and amygdala volumes (color-coded by cohort) with model-adjusted means ± 95% confidence intervals (CI) derived from linear regression models including infant age at scan, gestational age, birthweight, maternal education, and intracranial volume as covariates. Data are ComBat-harmonized. MO- denotes unexposed to maternal obesity, and MO + denotes exposed to maternal obesity. The number of observations in each category is as follows. Male MO- (N = 701), Male MO + (N = 181), Female MO- (N = 590), Female MO + (N = 166). The pooled violin plots are displayed for visual presentation purposes, which emphasize overall distributional patterns. *** P < 0.001 Interaction of maternal diabetes and sex Interactions of maternal diabetes with sex were not significantly associated with ICV and subcortical volumes (Table 2, Additional File 1: Extended Table 2, Additional File 1: eTable 7). Interaction of maternal diabetes, maternal obesity, and sex Three-way interactions of maternal diabetes, maternal obesity and sex emerged in the offspring thalamus (standardized β = −0.50, 95% CI −0.81 to −0.18, FDR corrected P = 0.017, Fig. 2, Table 2, Additional File 1: eTable 8).Fig. 2Pairwise comparisons of thalamic volume in male and female children exposed/unexposed to maternal obesity/diabetes. Scatter plots display thalamic volumes stratified by sex. Individual data points are color-coded by cohort, and overlaid points and error bars represent model-adjusted means ± 95% CI from linear regression models including infant age at scan, gestational age, birthweight, maternal education, and intracranial volume as covariates. The models tested the interaction between maternal obesity and diabetes exposures within each sex. Data are ComBat-harmonized. MO/MD- denotes unexposed to maternal obesity or maternal diabetes; MO-/MD + denotes exposed to maternal diabetes and not obesity; MO/MD- denotes exposed to maternal obesity and not diabetes; MO + /MD + denotes exposed to both maternal diabetes and obesity. The number of observations in each category is as follows. Males: MO-/MD- (N = 658), MO-/MD + (N = 43), MO + /MD- (N = 151), + / + (N = 30). Females: MO-/MD- (N = 550), MO-/MD + (N = 40), MO + /MD- (N = 143), MO + /MD + (N = 23). The pooled violin plots are displayed for visual presentation purposes, which emphasize overall distributional patterns. * P < 0.05, ** P < 0.01 Pairwise comparisons of thalamic volume in male and female children exposed/unexposed to maternal obesity/diabetes. Scatter plots display thalamic volumes stratified by sex. Individual data points are color-coded by cohort, and overlaid points and error bars represent model-adjusted means ± 95% CI from linear regression models including infant age at scan, gestational age, birthweight, maternal education, and intracranial volume as covariates. The models tested the interaction between maternal obesity and diabetes exposures within each sex. Data are ComBat-harmonized. MO/MD- denotes unexposed to maternal obesity or maternal diabetes; MO-/MD + denotes exposed to maternal diabetes and not obesity; MO/MD- denotes exposed to maternal obesity and not diabetes; MO + /MD + denotes exposed to both maternal diabetes and obesity. The number of observations in each category is as follows. Males: MO-/MD- (N = 658), MO-/MD + (N = 43), MO + /MD- (N = 151), + / + (N = 30). Females: MO-/MD- (N = 550), MO-/MD + (N = 40), MO + /MD- (N = 143), MO + /MD + (N = 23). The pooled violin plots are displayed for visual presentation purposes, which emphasize overall distributional patterns. * P < 0.05, ** P < 0.01 In male infants, a synergistic maternal obesity x maternal diabetes interaction in the thalamus was identified (standardized β = −0.30, 95% CI −0.51 to −0.09, FDR P = 0.015, Additional File 1: eTable 9a). Post hoc analysis showed that male infants with combined exposures had significantly lower thalamic volumes than those exposed to maternal diabetes alone (cohen’s d = 0.84, 95% CI 0.37 to 1.32, P = 0.003), maternal obesity alone (cohen’s d = 0.56, 95% CI 0.17 to 0.96, P = 0.027), or neither exposure (cohen’s d = 0.69, 95% CI 0.32 to 1.07, P = 0.002, Additional File 1: eTable 10b). In contrast, exposure to maternal diabetes alone (cohen’s d = −0.15, 95% CI −0.46 to 0.16, P = 0.777) or maternal obesity alone (cohen’s d = 0.13, 95% CI −0.05 to 0.31, P = 0.502) did not differ significantly from the unexposed group (Additional File 1: eTable 10b). In female infants, maternal obesity exposure was associated with smaller thalamic volume (standardized β = −0.12, 95% CI −0.20 to −0.04, FDR P = 0.015), and as well as maternal diabetes exposure (standardized β = −0.16, 95% CI −0.30 to 0.01, FDR P = 0.042). The interaction between maternal obesity and maternal diabetes was not significant in females (standardized β = 0.23, 95% CI: –0.01 to 0.46, FDR P = 0.092; Additional File 1: eTable 9b), although post hoc analysis indicated that female infants exposed to maternal obesity without diabetes had smaller thalamic volumes than unexposed (Cohen’s d = −0.28, 95% CI: −0.47 to −0.09, P = 0.022; Additional File1: eTable 10b). Maternal diabetes-by-maternal obesity-by-sex interactions were not significant in other subcortical volumes (Additional File 1: eTable 8). Sensitivity analysis results Results remained largely the same after completing various sensitivity analyses, with robust associations observed between maternal obesity and thalamus (Additional File 1: eTable 11–14, 16–19), hippocampus (Additional File 1: eTable 11–14, 16–18), and ICV (Additional File 1: eTable 12–13, 15, 17–18). Substitution of maternal BMI as a continuous IV in place of maternal obesity found similar associations with thalamic and hippocampal volumes (Additional File 1: eTable 18). Notably, sensitivity analyses found that associations of maternal obesity and amygdala and caudate volumes were significant without adjustment for ICV (Additional File 1: eTable 11). They also revealed associations of maternal diabetes with hippocampus and thalamus in models including only full-term children (Additional File 1: eTable 14), with ICV in models with PMA (Additional File 1: eTable 16), and with thalamus in models with additional adjustment for family income (Additional File 1: eTable 17). Given three-way significant correlations among maternal obesity, preterm birth, and ICV and thalamus, mediation analysis was performed and revealed only a small proportion of the total effect mediated for ICV (14.9%) and thalamus (21.6%), suggesting that preterm birth did not substantially mediate the association between maternal obesity and the brain volumes (Additional File 1: eTable 20). Cohort-wise analysis exhibited consistent patterns across cohorts (Additional File 2: eFigure 2). Discussion In this large, well-powered, multi-cohort study we identified independent, interactive sex-dependent associations between in utero exposure to maternal diabetes and/or maternal obesity and thalamic, hippocampal and amygdala volumes. To our knowledge, this is the largest study to establish associations between prenatal exposure to maternal diabetes and/or maternal obesity with subcortical brain volumes in early infancy. While ours is not the first study to report additive effects driven by maternal diabetes and maternal obesity exposure [48, 49], it is the first to robustly establish this link with offspring brain development in early infancy. Maternal diabetes and maternal obesity are the most common medical conditions during pregnancy and have separately been associated with structural and functional alterations in offspring's brain during gestation and infancy [17–23, 25, 26]. However, prior studies have failed to account for their frequent co-occurrence in exploring relationships with offspring brain outcomes. Our study marks an important advancement, explicitly recognizing this critical confound and disentangling the influence of maternal diabetes from that of maternal obesity. Maternal obesity exposure was independently associated with smaller ICV, thalamic, and hippocampal volumes in infants. Maternal diabetes showed no significant correlation with ICV or subcortical volumes after controlling maternal obesity exposure. These results suggest that maternal obesity exposure has a stronger impact on infant subcortical volumes than diabetes exposure, implying that prior studies may have overestimated the association of maternal diabetes by failing to isolate the effect of maternal obesity with offspring brain outcomes. Interestingly, the most prominent findings of our study are sex specific associations of maternal diabetes and maternal obesity with offspring thalamic, hippocampal and amygdala volumes. In females, maternal diabetes and maternal obesity were independently and negatively related to thalamus volumes, but combined exposures were not. In contrast, in males we observed that combined exposure was associated with lower thalamic volumes with moderate to large effect sizes, but maternal diabetes or maternal obesity alone were not. We also found a significant interaction between maternal obesity exposure and sex in infant hippocampal and amygdala volumes, which was driven by lower volumes in female offspring exposed to maternal obesity. These findings are noteworthy as sex differences in the thalamus, hippocampus and amygdala, regions involved in central regulation of metabolism, neuroendocrine, and stress-related processes, could provide the biological basis for studies that have reported sex differences in cardiometabolic disorders and neurodevelopmental disorders following prenatal exposure to maternal diabetes and maternal obesity [50]. While the current study was not designed to identify the biological mechanisms underlying these sex-specific effects, there is a rich body of evidence demonstrating that prenatal steroid hormone exposure drives sex-specific placental function [28] and epigenetic modification [29] in the developing brain. For example, sex hormones such as testosterone and estradiol bind to DNA through receptors and epigenome coactivators (e.g., Histone Acetyltransferases) and corepressors (e.g., Histone Deacetylases) that modify chromatin structure and gene expression [29]. In addition, sex-differences in placental immune responsivity are well established, and given that maternal obesity/diabetes is associated with a heightened inflammatory milieu, this also supports inflammatory pathways as contributors to sex-specific programming of placental function and brain development [51]. It then follows that the sex-specific findings reported here may result from an interaction between upstream hormonally programmed sex differences in the brain and exposure to diabetes/obesity-related ligands. Further studies would be needed to explore this possibility. Regardless, the current findings illustrate the important regulatory role of sex in offspring brain outcomes and emphasize the need to factor sex into analyses exploring these metrics. According to Developmental Origins of Health and Disease theory [52], environmental exposures act in the earliest stage of life to influence health and disease later in life and the developing brain has been shown to be a target of these exposures. In our study, the thalamus, hippocampus and amygdala show the greatest changes under the influence of maternal diabetes or maternal obesity. Thalamic development starts as early as 8–14 weeks of gestation and has the most protracted period of development among subcortical regions [34], making it highly susceptible to environmental insults. Moreover, our results are in concordance with a recent study in which 9–10-year-old children with prenatal exposure to maternal diabetes had smaller thalamic volumes than un-exposed children [12]. However, that study did not account for maternal obesity exposure, nor did it examine interactions with sex. Our results extend this work and provide a more robust scaffold for our understanding of this developmental trajectory. The hippocampus has also been reported to be susceptible to environmental insults in infant imaging studies [52–56] but the findings have been more mixed and there is thus a lack of consensus. In contrast with our work, a study in 7–11-year-old children exposed to maternal obesity prenatally showed that males but not females had lower hippocampal volumes [33], but this study was limited by sample size and studied older children, whereas our study included one of the largest samples of infant imaging near birth without confounding contributors of postnatal exposures. In a separate study, lower hippocampal thickness was observed in 7–11-year-old children exposed to maternal diabetes prenatally [14], while three other studies in children with similar age ranges did not observe such associations [12, 14, 16]. Our study aligns with the latter, confirming the absence of prenatal influence of maternal diabetes exposure on hippocampal volume. Our findings are the first to demonstrate negative associations of maternal obesity exposure with amygdala volume in female infants near birth. This contrasts with a recent study showing non-significant overall and sex-specific associations of maternal pre-pregnancy BMI and amygdala volumes among 10-year-olds. The discrepancy in results is likely attributable to the two studies examining different developmental time points (infancy near birth versus 10-year-olds), where the impact of the prenatal obesity exposure on brain structure may attenuate or change expression over time. Moreover, although a prior study in 9–10-year-old children reported significant negative relationships between prenatal exposure to maternal diabetes and dorsal striatum volumes [12] (i.e., putamen, caudate), we did not find such relationships in our younger cohort. We speculate that prior significant findings may be diminished after controlling for maternal obesity exposure given our data demonstrating that relationships between maternal diabetes exposure and infant subcortical volumes were attenuated after adjusting for maternal obesity exposure. Furthermore, our sensitivity analyses emphasize the robustness of our findings. Our study has strengths and limitations. First, a key strength is that we investigated relationships between prenatal exposure to maternal diabetes and/or maternal obesity and the developing brain in early infancy, thereby reducing confounding postnatal factors. Second, as diabetes and obesity often go together during pregnancy, we studied both independent and combined associations of maternal diabetes and maternal obesity with subcortical volumes. Third, we leveraged data from six international cohorts, providing the statistical power necessary to demonstrate reproducibility and estimates of effect sizes in the context of maternal diabetes/maternal obesity in heterogeneous populations. Finally, we ruled out possible cohort effects by analyzing each cohort separately and observing the direction of effects – in turn demonstrating the robustness of results. Nonetheless, our study is limited by the self-reported status of maternal diabetes and maternal obesity during pregnancy in some cohorts, lack of classification of types of diabetes during pregnancy, lack of information on maternal diet, medication use, and lack of metabolic markers in offspring. The sample was over-represented for pre-term infants. Initial mediation analysis showed that pre-term birth potentially mediated effects of maternal obesity on ICV and thalamus volumes. These results need to be interpreted with caution given that the sample size for pre-term subgroup was limited. We cannot exclude other confounders such as maternal mental health in pregnancy which has also been associated with infant outcomes. Our primary statistical analyses controlled site-specific effects, rather than scanner. Scanner upgrades in two sites may contribute to residual variance. We also restricted our analysis to subcortical volumes. Investigating other neuroimaging phenotypes, including thickness and surface area of cortical regions is an important future direction. Conclusions In a diverse sample of 1,965 infants, maternal obesity exposure was independently correlated with smaller ICV, thalamic, and hippocampal volumes near birth. Conversely, maternal diabetes exposure demonstrated no significant associations with offspring ICV or subcortical volumes. Prenatal exposure to maternal diabetes or maternal obesity was independently associated with lower thalamus volumes in females; Males with concomitant maternal diabetes and maternal obesity exposure demonstrated the lowest thalamic volumes. Females exposed to maternal obesity, but not males, exhibited lower hippocampus and amygdala volumes. These results support the hypothesis that gestational exposure to maternal diabetes and/or maternal obesity influences the developing brain, and sex moderates these effects. Supplementary Information Additional file 1: Extended Table 2, eTable 1 – eTable 20. Extended Table 2: Summary of results from regression models using ComBat adjusted data. eTable 1: Image acquisition and processing. eTable 2: Model 1- Effect of exposure to maternal diabetes on neonatal brain volumes using ComBat corrected data. eTable 3: Model 2 - Main effect of exposure to maternal diabetes and maternal obesity on neonatal brain volumes using ComBat corrected data. eTable 4: Model 3 - Effect of exposure to maternal diabetes and maternal obesity and their interaction with each other on neonatal brain volumes using ComBat adjusted data. eTable 5: Model 4 - Interaction of sex with maternal obesity and their effect neonatal brain volumes using ComBat adjusted data. eTable 6: Model 5 - Interaction of sex with maternal diabetes and their effect neonatal brain volumes using ComBat adjusted data. eTable 7: Model 6 - 3 way interaction of sex with maternal diabetes with maternal obesity and their effect neonatal brain volumes using ComBat adjusted data. eTable 8a: Model (MO + MD) - Main effect of exposure to maternal diabetes and maternal obesity on hippocampal and amygdala volumes using ComBat corrected data in males. eTable 8b: Model (MO + MD) - Main effect of exposure to maternal diabetes and maternal obesity on hippocampal and amygdala volumes using ComBat corrected data in females. eTable 9a: Model (MO X MD) - Effect of exposure to maternal diabetes and maternal obesity and their interaction with each other on thalamic volume using ComBat adjusted data in males. eTable 9b: Model (MO X MD) -Effect of exposure to maternal diabetes and maternal obesity and their interaction with each other on thalamic volume using ComBat adjusted data in females. eTable 10a: lsmeans and pairwise comparison for hippocampal and amygdala volume. eTable 10b: Sex stratified results of lsmeans and pairwise comparison for thalamic volume. eTable 11: Sensitivity analysis without including ICV using ComBat adjusted data. eTable 12: Sensitivity analysis without including birthweight in the model using ComBat adjusted data. eTable 13: Main effect of exposure to maternal diabetes with maternal obesity and their effect neonatal brain volumes using ComBat adjusted data without gestational age and birthweight as covariates. eTable 14: Sensitivity analysis with only fullterm born children using ComBat adjusted data. eTable 15: Sensitivity analysis with only pre-term born children using ComBat adjusted data. eTable 16: Sensitivity analysis with post-menstrual age using ComBat adjusted data. eTable 17: Main effect of exposure to maternal diabetes and maternal obesity on neonatal brain volumes using ComBat adjusted data adjusting for family income. eTable 18: Main effect of exposure to maternal diabetes and maternal obesity on neonatal brain volumes ComBat adjusted data with BMI as a continuous variable. eTable 19: Sensitivity analysis excluding the Cape Town cohort using ComBat adjusted data. eTable 20: Causal Mediation Analysis on effect of Obesity on associated brain volumes through preterm birth.Additional file 2: eFigure 1: Distribution of participant ages at MRI across cohorts. Violin plots shows the distribution of neonatal ages in days at the time of MRI acquisition for each cohort. Individual observations are shown as jittered points overlaid on the violins to illustrate sample density and variability. Cohort names are shown along the y-axis. eFigure 2. Cohort-specific associations between maternal obesity and diabetes exposure and neonatal brain volumes. Forest plots display cohort-wise regression estimates (β) and 95% confidence intervals for the associations between maternal obesity and maternal diabetes exposures and neonatal brain volumes (A) intracranial volume (ICV), (B) hippocampus, (C) amygdala, and (D) thalamus. Each point represents an exposure effect estimate from a cohort-specific linear regression model adjusted for infant age at scan, sex, gestational age at birth, low birth weight, maternal education, and intracranial volume (for non-ICV outcomes). The scanner model was included as an additional covariate in the GMIA and EBDS cohorts. Exposure variables were excluded from a model if the exposed group size was too small or had no variation within the cohort. The dashed vertical line represents the null (β = 0). Negative β values (left of the vertical dashed line) indicate lower volume associated with the maternal exposure, whereas positive β values (right of the vertical dashed line) indicate higher volume. Cohort names are shown along the y-axis, and colors distinguish maternal obesity and maternal diabetes exposures. Additional file 1: Extended Table 2, eTable 1 – eTable 20. Extended Table 2: Summary of results from regression models using ComBat adjusted data. eTable 1: Image acquisition and processing. eTable 2: Model 1- Effect of exposure to maternal diabetes on neonatal brain volumes using ComBat corrected data. eTable 3: Model 2 - Main effect of exposure to maternal diabetes and maternal obesity on neonatal brain volumes using ComBat corrected data. eTable 4: Model 3 - Effect of exposure to maternal diabetes and maternal obesity and their interaction with each other on neonatal brain volumes using ComBat adjusted data. eTable 5: Model 4 - Interaction of sex with maternal obesity and their effect neonatal brain volumes using ComBat adjusted data. eTable 6: Model 5 - Interaction of sex with maternal diabetes and their effect neonatal brain volumes using ComBat adjusted data. eTable 7: Model 6 - 3 way interaction of sex with maternal diabetes with maternal obesity and their effect neonatal brain volumes using ComBat adjusted data. eTable 8a: Model (MO + MD) - Main effect of exposure to maternal diabetes and maternal obesity on hippocampal and amygdala volumes using ComBat corrected data in males. eTable 8b: Model (MO + MD) - Main effect of exposure to maternal diabetes and maternal obesity on hippocampal and amygdala volumes using ComBat corrected data in females. eTable 9a: Model (MO X MD) - Effect of exposure to maternal diabetes and maternal obesity and their interaction with each other on thalamic volume using ComBat adjusted data in males. eTable 9b: Model (MO X MD) -Effect of exposure to maternal diabetes and maternal obesity and their interaction with each other on thalamic volume using ComBat adjusted data in females. eTable 10a: lsmeans and pairwise comparison for hippocampal and amygdala volume. eTable 10b: Sex stratified results of lsmeans and pairwise comparison for thalamic volume. eTable 11: Sensitivity analysis without including ICV using ComBat adjusted data. eTable 12: Sensitivity analysis without including birthweight in the model using ComBat adjusted data. eTable 13: Main effect of exposure to maternal diabetes with maternal obesity and their effect neonatal brain volumes using ComBat adjusted data without gestational age and birthweight as covariates. eTable 14: Sensitivity analysis with only fullterm born children using ComBat adjusted data. eTable 15: Sensitivity analysis with only pre-term born children using ComBat adjusted data. eTable 16: Sensitivity analysis with post-menstrual age using ComBat adjusted data. eTable 17: Main effect of exposure to maternal diabetes and maternal obesity on neonatal brain volumes using ComBat adjusted data adjusting for family income. eTable 18: Main effect of exposure to maternal diabetes and maternal obesity on neonatal brain volumes ComBat adjusted data with BMI as a continuous variable. eTable 19: Sensitivity analysis excluding the Cape Town cohort using ComBat adjusted data. eTable 20: Causal Mediation Analysis on effect of Obesity on associated brain volumes through preterm birth. Additional file 2: eFigure 1: Distribution of participant ages at MRI across cohorts. Violin plots shows the distribution of neonatal ages in days at the time of MRI acquisition for each cohort. Individual observations are shown as jittered points overlaid on the violins to illustrate sample density and variability. Cohort names are shown along the y-axis. eFigure 2. Cohort-specific associations between maternal obesity and diabetes exposure and neonatal brain volumes. Forest plots display cohort-wise regression estimates (β) and 95% confidence intervals for the associations between maternal obesity and maternal diabetes exposures and neonatal brain volumes (A) intracranial volume (ICV), (B) hippocampus, (C) amygdala, and (D) thalamus. Each point represents an exposure effect estimate from a cohort-specific linear regression model adjusted for infant age at scan, sex, gestational age at birth, low birth weight, maternal education, and intracranial volume (for non-ICV outcomes). The scanner model was included as an additional covariate in the GMIA and EBDS cohorts. Exposure variables were excluded from a model if the exposed group size was too small or had no variation within the cohort. The dashed vertical line represents the null (β = 0). Negative β values (left of the vertical dashed line) indicate lower volume associated with the maternal exposure, whereas positive β values (right of the vertical dashed line) indicate higher volume. Cohort names are shown along the y-axis, and colors distinguish maternal obesity and maternal diabetes exposures.