Puberty, pregnancy, and menopause show shared and distinct structural changes across the lifespan

Sophie R van ’t Hof; Marieke G N Bos; Milou Straathof; Eveline A Crone; Elseline A Hoekzema

doi:10.1038/s41467-026-76755-2

. 2026 Sep 8;17:9360. doi: 10.1038/s41467-026-76755-2

Puberty, pregnancy, and menopause show shared and distinct structural changes across the lifespan

Sophie R van ’t Hof ^1,^✉, Marieke G N Bos ^2,³, Milou Straathof ¹, Eveline A Crone ⁴, Elseline A Hoekzema ¹

PMCID: PMC13554251 PMID: 42711329

Abstract

The female brain undergoes substantial reorganization during major hormonal transitions, yet whether puberty, pregnancy, and menopause engage shared or distinct neuroplastic mechanisms remains unknown. We compared longitudinal structural brain changes across all three major hormonal life events using identical analytical methods within a single cohort framework (n = 1095), with stable control groups to control for normative developmental and aging trajectories. Both pre-to-post menarche and pre-to-post pregnancy groups showed widespread cortical gray matter reductions, with cross-cohort comparisons revealing divergent profiles across more than half of cortical regions alongside convergence in prefrontal, parietal, and temporal association cortices. Pre-to-post menopausal women showed a fundamental divergence from both puberty and pregnancy, as they showed no volumetric loss, while stable pre- and postmenopausal control groups did. Puberty and pregnancy thus share both convergent and divergent cortical remodeling, while menopause represents a qualitatively distinct pattern, defined by the absence rather than acceleration of volumetric change.

Subject terms: Neuroscience, Brain

Brain volumes decline during puberty and pregnancy, with both shared and distinct regional patterns, while menopause appears to briefly pause rather than speed up ongoing volume loss, as shown by this neuroimaging study directly comparing these three transitions.

Introduction

The female brain can undergo three major hormonal life events across the lifespan: puberty, pregnancy, and menopause. Each event involves profound shifts in sex steroid hormones, particularly estrogens and progesterone, which bind to widespread receptors throughout the brain. Longitudinal neuroimaging studies have begun to characterize structural brain changes during puberty and pregnancy, but evidence for menopause is limited to a single study that did not examine whole-brain gray matter structure. Whether these three events engage a shared neuroplastic program or each triggers distinct patterns of remodeling have not yet been directly examined.

Puberty represents the first major hormonal life event, characterized by dramatic increases in sex steroid hormones that coincide with extensive brain remodeling. Although most studies focus on the entire adolescence period, emerging research has begun to isolate puberty-specific effects on brain structure. Longitudinal neuroimaging studies document reductions in cortical gray matter volume across puberty^1–7. Recent evidence suggests that menarche, the first menstruation, may mark a particularly salient transition point within this broader developmental window, as machine learning algorithms can classify pre- versus post menarche status from brain structure with high accuracy⁸. These findings suggest that the menarche transition itself represents a discrete neuroplastic event warranting specific investigation, rather than simply one point along a continuous pubertal trajectory.

Similarly, pregnancy induces pronounced structural brain changes. Longitudinal studies comparing pre-conception to pregnancy and post-pregnancy brain scans in women becoming first-time mothers by our and other groups reveal decreases in gray matter volume^9–14. Reductions are linked to estradiol levels during pregnancy¹⁰, and some persist for at least six years postpartum¹⁵. Recent work from our group shows similar changes occur during a second pregnancy, with some specific differences¹¹. The pregnancy-related changes predominantly affect regions of the default mode and theory-of-mind networks, both implicated in social processes such as social cognition, suggesting they may represent adaptive remodeling that supports maternal caregiving behaviors and mother-infant bonding^9,10,16,17. Remarkably, in a previous study, we have demonstrated that the magnitude and pattern of brain morphometric changes during a first pregnancy closely resembles those occurring in pubertal girls, with both groups showing comparable rates of monthly volumetric reduction¹⁸.

In striking contrast to the emerging puberty and pregnancy research, menopause remains profoundly understudied. To date, only a single longitudinal neuroimaging study has examined gray matter volume changes across the menopausal transition¹⁹, focusing exclusively on hippocampal atrophy. Cross-sectional studies suggest menopause affects frontal and temporal cortical regions²⁰, but the lack of longitudinal whole-brain data limits understanding of the trajectory of cortical change during the menopausal transition, the regional specificity of these changes, and whether hormonal decline produces patterns opposite to the hormonal increases of puberty and pregnancy.

A recent review of structural MRI evidence across the full female lifespan suggests that each transition is associated with region-specific gray matter changes²¹, yet the absence of longitudinal whole-brain comparisons across all three transitions within a single framework leaves fundamental questions about shared versus distinct neuroplastic mechanisms unresolved. These transitions may engage overlapping neuroplastic processes driven by sex steroid fluctuations, predicting broadly shared spatial patterns, though potentially with reversed directionality when hormones decline versus increase. In addition, each transition may engage distinct mechanisms tailored to stage-specific adaptive demands: neurodevelopmental specialization in puberty, maternal caregiving preparation in pregnancy, and adaptation to reproductive senescence in menopause. These distinct adaptive demands may predict different regional targets despite involving the same hormonal systems. Distinguishing the extent of shared versus unique mechanisms requires direct comparison of spatial extent, regional specificity, and magnitude of changes across transitions.

To address these questions, we compared brain structural changes across all three major female hormonal life events using longitudinal neuroimaging data and identical analytical methods applied within a single study. We apply FreeSurfer-based morphometric analyses consistently across all cohorts, enabling quantitative comparison of effect sizes, spatial patterns, and regional specificity. Critically, we include not only the pre-to-post menarche, pregnancy, menopause groups but also stable control groups, to help isolate transition-specific changes from normal developmental or aging trajectories. This leads to three groups per cohort: in the puberty cohort, a pre-to-post menarche group, with stable premenarchal and stable postmenarchal groups serving as controls. In the pregnancy cohort, women becoming first- or second-time mothers are compared with nulliparous controls scanned at similar intervals. In the menopause cohort, a pre-to-post menopause group consists of women who reported not having reached menopause at the first timepoint and reported being postmenopausal at the second timepoint, with stable premenopausal and stable postmenopausal groups as controls.

In this work, we find that both pre-to-post menarche as well as pre-to-post first and second pregnancy show accelerated cortical gray matter reductions relative to their respective controls, with cross-cohort comparisons revealing divergent profiles across more than half of cortical regions, including a stepwise gradient in sensorimotor and cingulate cortex where pubertal decline exceeded pregnancy, which in turn exceeded menopause. Convergence between puberty and pregnancy was nonetheless evident across prefrontal, parietal, and temporal association cortices. The menopausal transition shows a qualitatively distinct pattern: rather than the volumetric loss observed in stable premenopausal and postmenopausal women, pre-to-post menopausal women show no significant reduction in total and cortical gray matter volume. These findings demonstrate that puberty and pregnancy share regions of convergent cortical remodeling alongside regional divergence, while menopause represents a distinct pattern defined not by what changes, but by what does not.

Results

Changes in total, cortical and subcortical gray matter volume

To characterize the magnitude of structural brain changes during puberty, we first compared monthly rates of gray matter volume change between groups within our: 1. Pubertal cohort (n = 142), with menarche status based on self-report, which consisted of a stable pre-menarche group (n = 49), a pre-to-post menarche group (n = 34) and a post-menarche group (n = 59) 2. Pregnancy cohort (n = 110), consisting of a group pre-to-post first pregnancy (n = 40), pre-to-post second pregnancy (n = 30), and nulliparous control group (n = 40) 3. Menopause cohort (n = 843), with menopausal status based on self-report, with a stable pre-menopausal group (n = 49), a pre-to-post menopause group (n = 120), and a stable post-menopausal group (n = 674). All rates reflect longitudinal change computed over each participant’s assigned interval (T1-T2 or T2–T3 for the puberty cohort, T1–T2 for the pregnancy cohort, and T1-T2 for the menopause cohort), expressed as monthly percentage change in eTIV-adjusted bilateral gray matter volume.

Pubertal cohort

One-way ANOVAs revealed significant group differences for total gray matter volume (F(2,139) = 4.79, p = 0.0010, η² = 0.060, % CI = 0.01, 1.00), cortical volume (F(2,139) = 5.49, p = 0.005, η² = 0.070, % CI = 0.01, 1.00), and also for subcortical volume (F(2,139) = 4.82, p = 0.009, η² = 0.060, % CI = 0.01, 1.00). Post-hoc Tukey HSD tests indicated that for total gray matter volume, the stable premenarchal group differed significantly from the pre-to-post group (mean difference = 0.23 mm³/month, p = 0.007, d = 0.69, % CI = 0.23, 1.13), with a similar pattern for cortical volume (mean difference = 0.247 mm³/month, p = 0.004, d = 0.73, % CI = 0.38, 1.18).

For subcortical volume, the stable premenarchal group differed significantly from the stable postmenarchal group (mean difference = 0.189 mm³/month, p = 0.025, % CI = −0.90, −0.12), as well as the pre-to-post menarche group (mean difference = 0.22, p = 0.021, % CI = 0.15, 1.05), while the pre-to-post menarche group did not differ from the stable postmenopause group; see Supplementary Table 4 for detailed statistics.

To characterize the direction of these group differences, one-sample t-tests were used to assess whether monthly rates of change differed significantly from zero within each group, see Supplementary Table 5 and Fig. 1 (Bonferroni-corrected threshold: p < 0.0055). The pre-to-post group exhibited significant monthly reductions in total gray matter (t(33) = −5.88, p < 0.001, M = −0.13 mm³/month, % CI = −0.17, −0.08) and cortical volume (t(33) = −6.84, p < 0.001, M = −0.16 mm³/month, % CI = −0.21, −0.11), whereas neither the stable premenarchal nor the stable postmenarchal group showed significant monthly changes from zero for these measures. For subcortical volume, the stable premenarchal group showed a significant monthly increase (t(48) = 3.05, p = 0.004, M = 0.19 mm³/month, % CI = 0.065, 0.32), while neither the pre-to-post nor the stable postmenarchal group differed significantly from zero. No significant associations between age at baseline and monthly rate of change were observed in any group, see Supplementary Table 6.

Fig. 1 — Violin plots show the distribution of monthly percentage rates of change in total (top), cortical (middle), and subcortical (bottom) gray matter volume for all groups within the pubertal, pregnancy, and menopause cohorts. Dots indicate group medians. Asterisks denote groups whose rates differ significantly from zero (Bonferroni-corrected). Brackets with asterisks indicate significant between-group differences within cohort. Rates reflect eTIV-adjusted bilateral volumes computed between available timepoints and expressed as monthly percentage change. Note all have age-matched control groups (premenarchal and menarche transition; first-time mothers and nulliparous control; premenopausal and menopause transition). Source data are provided as a Source Data file. Artwork by make.piccs.

Pregnancy cohort

To characterize the magnitude of structural brain changes during pregnancy, we first compared monthly rates of gray matter volume change between groups. One-way ANOVAs revealed significant group differences for total gray matter volume (F(2,106) = 28.67, p < 0.001, η² = 0.35, % CI = 0.01, 1.00) and cortical volume (F(2,106) = 22.83, p < 0.001, η² = 0.30, % CI = 0.01, 1.00), but not for subcortical volume (F(2,106) = 3.47, p = 0.035, η² = 0.06, % CI = 0.01, 1.00), which did not survive Bonferroni correction. Post-hoc Tukey HSD tests indicated that for both total gray matter and cortical volume, first-time and second-time mothers differed significantly from nulliparous controls (total GM: t(107) = −6.80, p < 0.0001, η² = 0.30, % CI = [0.19, 1.00 and t(107) = −6.03, p < 0.0001, η² = 0.25, % CI = 0.14, 1.00; cortical: t(107) = −6.01, p < 0.0001, η² = 0.25, % CI = 0.14, 1.00 and t(107) = −5.50, p < 0.0001, η² = 0.22, % CI = 0.12, 1.00), but not from each other (total GM: t(107) = −0.23, p = 0.97, η² = 0.0005, % CI = 0.00, 1.00; cortical: t(107) = −0.04, p = 0.999, η² = 0.00001, % CI = 0.00, 1.00); see Supplementary Table 4.

To characterize the direction of these group differences, one-sample t-tests were used to assess whether monthly rates of change differed significantly from zero within each group (Bonferroni-corrected threshold: p < 0.0055), see Supplementary Table 5 and Fig. 1. Both first-time mothers (t(39) = −8.24, p < .001, M = −0.12 mm³/month, % CI = −0.14, −0.88) and second-time mothers (t(29) = −8.55, p < 0.001, M = −0.11 mm³/month, % CI = −0.14, 0.085) exhibited significant monthly reductions in total gray matter volume, whereas nulliparous controls did not differ significantly from zero (t(39) = 1.093, p = 0.28, M = 0.16 mm³/month, % CI = −0.14, 0.047). A highly similar pattern was observed for cortical volume, with significant reductions in first-time (t(39) = −6.38, p < 0.001, M = −0.12 mm³/month, % CI = −0.16, −0.082) and second-time mothers (t(29) = −8.33, p < 0.001, M = −0.12 mm³/month, % CI = −0.15, −0.090) but not in controls (t(39) = 1.60, p = 0.12, M = 0.033 mm³/month, % CI = −0.0088, 0.075). For subcortical volume, significant reductions were observed across all three groups (first-time mothers: t(39) = −5.75, p < 0.001, % CI = −0.10, −0.050; second-time mothers: t(29) = −5.86, p < 0.001, % CI = −0.93, −0.45; nulliparous controls: t(38) = −4.11, p < 0.001, % CI = −0.057, −0.019). No significant associations between age at baseline and monthly rate of change were observed in any group; see Supplementary Table 6.

Menopause cohort

To characterize the magnitude of structural brain changes around the menopause, we first compared monthly rates of gray matter volume change between groups. One-way ANOVAs revealed no significant group differences for total gray matter volume (F(2, 840) = 2.11, p = 0.122, η² <0.01, % CI = 0.00, 1.00), cortical volume (F(2, 840) = 1.93, p = 0.146, η² <0.01, % CI = 0.00, 1.00), or subcortical volume (F(2, 840) = 0.057, p = 0.944, η² <0.01, % CI = 0.00, 1.00).

To characterize the direction of change within each group, one-sample t-tests were used to assess whether monthly rates of change differed significantly from zero (Bonferroni-corrected threshold: p < 0.0055), see Supplementary Table 5. For total gray matter volume, significant monthly reductions were observed in the stable premenopausal group (t(48) = −3.61, p < 0.001, M = −0.033 mm³/month, % CI = −0.052, −0.015) and the stable postmenopausal group (t(673) = −9.28, p < 0.001, M = −0.025 mm³/month, % CI = −0.030, −0.020), but not in the pre-to-post group (t(119) = −2.43, p = 0.017, M = −0.013 mm³/month, % CI = −0.12, −0.0024; not significant after Bonferroni correction). A similar pattern was observed for cortical volume, with significant reductions in the stable premenopausal group (t(48) = −3.25, p = 0.002, M = −0.033 mm³/month, % CI = −0.054, −0.013) and the stable postmenopausal group (t(673) = −10.22, p < 0.001, M = −0.028 mm³/month, % CI = −0.033, −0.022), but not in the pre-to-post group (t(119) = −2.35, p = 0.021, M = −0.013 mm³/month, % CI = −0.024, −0.0020; not significant after Bonferroni correction). For subcortical volume, significant reductions were observed across all three groups (stable premenopausal: t(48) = −3.56, p < 0.001, % CI = −0.058, −0.016; pre-to-post: t(119) = −5.99, p < 0.001, % CI = −0.049, −0.025; stable postmenopausal: t(673) = −13.49, p < 0.001, % CI = −0.045, −0.034).

No significant associations between age at baseline and monthly rate of change were observed in any group; see Supplementary Table 6. Analyses excluding participants using HRT at any timepoint did not change the results; Supplementary Table 7.

Region-specific analyses

Pubertal cohort

The cortical PCA of 74 cortical regions revealed that the first principal component explained 55.18% of the total variance in cortical brain change, with a clear drop-off to subsequent components (PC2: 3.98%, PC3: 2.80%); see Supplementary Fig. 3. All 74 regions showed moderate loadings (0.10–0.15); see Fig. 2 and see Supplementary Table 8. The predominant direction of loadings was negative, consistent with coordinated gray matter reduction across these regions, in line with previous research.

Fig. 2 — Top 20 regions with the highest loading of the principal component analysis per cohort and distribution of PC1 scores across groups for the A puberty, B pregnancy, and C menopause cohorts. Left: Brain depicting these cortical regions colored by PCA loading, with the color scale. Right: Bar plot with center line showing median; box limits = 25th and 75th percentiles (IQR); whiskers extend to 1.5× IQR; points represent individual observations beyond whiskers. Error bars show 95% confidence intervals. A Pubertal cohort, n = 142: stable pre-menarche (n = 49), pre-to-post menarche (n = 34), and stable post-menarche (n = 59) groups; stable groups served as controls. B Pregnancy cohort, n = 110: pre-to-post first-pregnancy (n = 40), pre-to-post second-pregnancy (n = 30), and nulliparous control (n = 40) groups. C Menopause cohort (UK Biobank), n = 843: stable pre-menopausal (n = 49), pre-to-post menopause (n = 120), and stable post-menopausal (n = 674) groups; stable groups served as controls. In all panels, each individual contributed two structural MRI scans (baseline and follow-up), from which an individual monthly rate of change was derived; replicates are biological, and the unit of study is the subject’s monthly rate of change rather than the individual scan. Group differences in PC1 scores were assessed using one-way ANOVA with Tukey HSD post-hoc tests (two-sided), correcting for the three pairwise contrasts within each cohort (* p < 0.05, ** p < 0.01, *** p < 0.001). A F(2,138) = 8.07, p <0.001, η² = 0.10; pre-to-post vs. stable premenarche, p = 0.00004; vs. stable postmenarchal, p = 0.038; premenarche vs. postmenarchal, p = 0.143. B F(2,106) = 23.38, p <0.001, η² = 0.31; first-time vs. nulliparous controls, p <0.001; second-time vs. nulliparous controls, p <0.001; first-time vs. second-time, p = 0.97. C F(2,840) = 3.47, p = 0.031, η² = 0.0082; pre-to-post vs. stable postmenopausal, p = 0.030; vs. stable premenopausal, p = 0.15; premenopausal vs. postmenopausal, p = 0.90. Source data are provided as a Source Data file.

One-way ANOVA revealed significant group differences in PC1 scores (F(2,138) = 8.07, p < 0.001, η² = 0.10, % CI = 0.03, 1.00). Post-hoc Tukey HSD comparisons indicated that girls in the pre-to-post menarche group showed significantly higher PC1 scores compared to stable premenarchal girls (t(138) = 3.96, p < 0.001, d = 0.89, % CI = 0.43, 1.35) and stable postmenarchal girls (t(138) = 1.90, p = 0.038, d = 0.41, % CI = −0.02, 0.84), indicating a stronger expression of the dominant pattern of coordinated cortical change. No significant difference was observed between the stable premenarchal and stable postmenarchal groups (t(138) = 2.48, p = 0.143, d = 0.48, % CI = 0.09, 0.87); see Supplementary Table 9.

To test whether age or follow-up duration influenced PC1 scores, linear models were estimated with PC1 as the outcome and group as the predictor, with each covariate added separately. Neither the main effect of age (β = −0.292, p = 0.33, % CI = 0.89, 0.30) nor the group × age interaction (F(2,134) = 0.47, p = 0.628) was significant, suggesting that age did not confound the group differences in PC1 scores. However, follow-up duration showed a significant main effect on PC1 scores (β = 0.384, p <0.001, % CI = 0.19, 0.58), and follow-up duration differed significantly across groups (F(2,138) = 8.69, p <0.001, η² = 0.12, % CI = 0.04, 1.00). The group × follow-up duration interaction was not significant (F(2,135) = 0.58, p = 0.561). As significant group differences in total subcortical gray matter were observed only in the pubertal cohort, monthly rates of change were examined within 8 bilateral subcortical structures. Kruskall-Wallis revealed significant group differences for the thalamus (H(2) = 12.64, p_fdr = 0.012, η² = 0.090,% CI = 0.028, 1.00), hippocampus (H(2) = 8.15, p_fdr = 0.034, η² = 0.058, % CI = 0.016, 1.00), putamen (H(2) = 9.89, p_fdr = 0.019, η = 0.070, % CI = 0.021, 1.00), and ventral DC (H(2) = 11.59, p_fdr = 0.012, η² = 0.082, % CI = 0.026, 1.00); see Supplementary Table 10 and Supplementary Fig. 4. Post-hoc analyses show for thalamus, a significant difference between the stable premenarchal and the pre-to-post (W = 1211, p = 0.0014, r = 0.454, % CI = 0.23, 0.63) and stable postmenarchal (W = 1880, p = 0.022, r = 0.301, % CI = 0.090, 0.49), but not between the pre-to-post and postmenarchal group (W = 951, p = 1.00, r = −0.052, % CI = −0.29, 0.19). For the hippocampus, there was only a significant difference between the two stable groups (W = 1874, p = 0.025, r = 0.296, % CI = 0.086, 0.48]), but not between the pre-to-post and stable premenarchal W = 1048, p = 0.14, r = 0.258, % CI = 0.010, 0.48]) and postmenarchal group (W = 1121, p = 1.00, r = 0.12, % CI = −0.13, 0.35). For the putamen, there was a significant difference between the stable premenarchal group and pre-post group (W = 1162, p = 0.0071, r = 0.40, % CI = 0.16, 0.57), but not between the stable postmenarchal group and the pre-post group W = 837, p = 0.56, r = −0.17, % CI = 0.39, 0.078) or premenarchal group (W = 1782, p = 0.11, r = 0.23, % CI = 0.017, 0.43). For the ventral DC, there is a significant difference between the stable premenarchal and stable postmenarchal group (W = 1913, p = 0.012, r = 0.32, % CI = 0.12, 0.50) and pre-to-post and stable postmenarchal group (W = 1305, p = 0.049, r = 0.30, % CI = 0.066, 0.51) but not between the stable premenarchal and pre-to-post group (W = 1041, p = 0.16, r = 0.25, % CI = 0.001, 0.47).

Wilcoxon rank test assessing whether monthly rates of change differed from zero indicated a significant positive monthly rate of change in the thalamus in the stable premenarchal group (H(2) = 12.64, p = 0.0018, η² = 0.090, % CI = 0.035, 1.00), surviving Bonferroni correction (p < 0.0021). No other subcortical structures showed significant monthly change in any group, see Supplementary Table 11. Also, no significant correlations between age and monthly rate of change were observed for any subcortical structure in any group; see Supplementary Table 12.

Pregnancy cohort

The cortical PCA revealed that PC1 explained 48.59% of the total variance in cortical brain change, with a clear drop-off to subsequent components (PC2: 5.64%, PC3: 4.21%); see Supplementary Fig. 3. Examination of PC1 loadings revealed three regions with strong contributions (loading > 0.15): the precuneus, superior temporal sulcus, and superior parietal lobule, broadly consistent with previous reports of pregnancy-related brain changes (see Fig. 2). An additional 52 regions showed moderate loadings (0.10–0.15); see Supplementary Table 13. The predominant direction of loadings was negative, consistent with coordinated gray matter reduction across these regions, in line with previous research.

One-way ANOVA revealed significant group differences in PC1 scores (F(2,106) = 23.38, p < 0.001, η² = 0.31, CI = 0.19, 1.00). Post hoc Tukey HSD comparisons indicated that both first time mothers (PRG₁ vs. control: t(106) = 6.17, p <0.001, d = 1.39, % CI = 0.94, 1.84) and second time mothers (PRG₂ vs. control: t(106) = 5.50, p <0.001, d = 1.34, % CI = 0.85, 1.82) showed significantly higher PC1 scores compared to nulliparous controls, indicating a stronger expression of the dominant pattern of coordinated cortical change in pregnant women. First-time and second-time mothers did not differ significantly from each other (t(106) = 0.22, p = 0.97, d = 0.05, % CI = −0.43, 0.53), suggesting that the core pattern of pregnancy-related brain changes is similar regardless of parity; see Supplementary Table 14.

To test whether age influenced PC1 scores, a linear model was estimated with PC1 as the outcome, group as a categorical predictor, and age as a covariate. The main effect of age was non-significant (t(105) = −1.41, p = 0.162, η² = 0.01, % CI = 0.00, 1.00). A significant group by age interaction was detected (F(2, 103) = 5.34, p = 0.006, η² = 0.06, % CI = 0.00, 1.00) suggesting that the relationship between age and cortical change patterns differed across groups. Examination of within-group age slopes revealed a significant negative relationship between age and PC1 scores in first-time mothers (t(103) = −3.43, p = 0.001, b = −0.75, % CI = −1.18, −0.32), indicating that older maternal age was associated with a stronger expression of the dominant cortical change pattern. No significant age effects were observed in second-time mothers (t(103) = 0.67, p = 0.506, b = 0.26, % CI = −0.51, 1.02), or nulliparous controls (t(103) = 0.80, p = 0.425, b = 0.17, % CI = −0.26, 0.60); see Supplementary Fig. 5.

Menopause cohort

The cortical PCA revealed that PC1 explained 23.55% of the total variance in cortical brain change, with a clear drop-off to subsequent components (PC2: 10. 68%, PC3: 6.35%); see Supplementary Fig. 3. Examination of PC1 loadings revealed 17 regions with strong contributions (loading > 0.15), including the superior frontal gyrus, middle frontal gyrus, and orbital gyrus (see Fig. 2). An additional 26 regions showed moderate loadings (0.10–0.15); see Supplementary Table 15. The predominant direction of loadings was positive, in contrast to the pubertal and pregnancy cohort.

One way ANOVA revealed significant group differences in PC1 scores (F(2, 840) = 3.47, p = 0.031, η² = 0.0082, % CI = 0.00, 1.00). Post hoc Tukey HSD comparisons indicated that women in the pre-to-post menopause group showed significantly higher PC1 scores compared to the stable postmenopausal group (t(840) = 2.55, p = 0.030, η² = 0.0077, % CI = 0.08, 2.02). No significant differences were observed between the stable premenopausal and pre-to-post groups, (t(840) = −1.85, p = 0.15, η² = 0.0041, % CI = −2.96, 0.35), or between the stable premenopausal and stable postmenopausal groups (t(840) = −0.42, p = 0.91, η² = 0.0002, % CI = −1.70, 1.19); see Supplementary Table 16.

To test whether age, follow-up duration, HRT use, or parity influenced PC1 scores, linear models were estimated with PC1 as the outcome, group as the predictor, and each covariate added separately. Neither the main effect of age (t(839) = 0.51, p = 0.61, b = 0.030, η² = 0.0003, % CI = −0.08, 0.14), nor the group by age interaction (F(2, 837) = 2.56, p = 0.078, η² = 0.0061), was significant. Similarly, neither the main effect of follow-up duration (t(839) = 1.91, p = 0.056, b = 0.0006, η² = 0.0043, % CI = 0.00, 0.001), nor the group by follow up duration interaction (F(2, 837) = 0.88, p = 0.42, η² = 0.0021) was significant. Neither the main effect of HRT (t(839) = 1.60, p = 0.11, b = 0.54, η² = 0.0030, % CI = −0.12, 1.21), nor the group by HRT interaction (F(2, 837) = 1.59, p = 0.21, η² = 0.0038), were significant. The main effect of parity (t(839) = 0.69, p = 0.49, b = 0.086, η² = 0.0006, % CI = −0.16, 0.33) and the group by parity interaction (F(2, 837) = 2.00, p = 0.14, η² = 0.0048) were also not significant, suggesting that none of these factors confounded the group differences in PC1 scores.

Comparison of regional localization across cohorts

To assess whether cohort differences in pre-to-post groups gray matter change were convergent or divergent across the full cortical parcellation, Kruskal-Wallis tests were conducted across all 74 bilateral Destrieux cortical regions. Results revealed significant differences in pre-to-post-specific monthly rates of gray matter change across the three cohorts in all 74 regions after FDR correction (all p_FDR ≤ 0.001, η² range = 0.073–0.563; median η² = 0.376), see Supplementary Tables 17 and 18. The near-universal omnibus significance, combined with predominantly large effect sizes across all 74 regions, indicates that the pattern of divergence identified in the PCA-selected regions generalizes across the full cortical parcellation.

Pairwise follow-up tests, each FDR-corrected across all 74 regions, revealed a consistent directional pattern, see Supplementary Table 19. In 67 of the 74 significant regions, the pre-to-post groups in the pubertal and pregnancy cohorts both showed control-subtracted rates below zero (i.e., greater decline than controls), while the pre-to-post menopausal group showed rates at or near zero relative to stable pre-menopausal women. This pattern was for puberty and pregnancy across all 67 regions; the menopausal pre-to-post group showed rates above zero in 60 of these regions, and where negative, rates were negligible relative to the other two pre-to-post groups. These findings confirm that puberty and pregnancy are both characterized by accelerated cortical volume loss relative to controls, whereas the pre-to-post menopause is not.

Menarche & pregnancy > menopause

In 34 regions, the pre-to-post menopause group differed from both the pre-to-post menarche and pregnancy, which did not differ significantly from each other, with predominantly large effect sizes (η² median = 0.38, range = 0.14–0.54; Fig. 3). These regions were broadly distributed across lateral and orbital frontal cortex, superior and inferior temporal gyri and sulci, inferior and superior parietal cortex, posterior cingulate, precuneus, fusiform and parahippocampal cortex, and occipital-temporal regions, spanning association cortex bilaterally.

Fig. 3 — Cortical regions where pairwise follow-up tests revealed significant differences between the menopausal transitioning group and both the pubertal and pregnancy transitioning groups, with no significant difference between the pubertal and pregnancy group. *Above:* Mean control-subtracted monthly rates of gray matter volume change per region for pre-to-post menarche group minus stable premenarchal; pre-to-post first-pregnancy group minus nulliparous control group; pre-to-post menopause group minus stable premenopausal group, grouped into brain parts. *Below:* Brain depicting each of the plotted cortical regions colored by effect size (η² = 0.10–0.55). The anatomical labels, rates of changes and effect sizes for all regions are provided in the Supplementary Tables 17–19. Source data are provided as a Source Data file.

Menarche > pregnancy > menopause

In a further 33 regions, all three pairwise comparisons reached significance (η² median = 0.410, range = 0.211–0.563; Fig. 4A), with the pre-to-post menarche showing greater decline than pregnancy, and pregnancy in turn showing greater decline than menopause. These regions spanned sensorimotor, cingulate, frontal, temporal, and occipital cortex, encompassing pre- and postcentral gyri, paracentral and subcentral gyri and sulci, superior frontal gyrus and sulcus, anterior and mid-cingulate, the intraparietal sulcus, superior temporal gyrus and transverse gyrus, parieto-occipital sulcus, lateral fissure, and superior occipital regions. Note that one region, the subcallosal gyrus, differed only between menopause and puberty (η² = 0.073).

Menarche > pregnancy & menopause

In six regions, the pre-to-post menarche showed greater decline than both pregnancy and the menopausal transition, while pregnancy and menopause did not differ significantly from each other (Fig. 4B; η² range = 0.117–0.347). These regions included the long insular gyrus and central opercular sulcus, the short insular gyrus, central sulcus, inferior circular insular sulcus, lateral orbital sulcus, and transverse temporal sulcus. The central sulcus and insular regions had the largest effect sizes within this group, suggesting that the pubertal trajectory in these regions is qualitatively distinct from that of both other transitions and does not share the broader convergence with pregnancy seen across most of the cortex.

Discussion

This study compares brain structural changes across three major female hormonal transitions, puberty, pregnancy, and menopause, using longitudinal neuroimaging data and identical analytical methods applied within a single framework. This approach reveals distinct degrees of brain plasticity that differ markedly in magnitude, spatial distribution, and direction of change.

We first examined monthly rates of total, cortical, and subcortical gray matter change across all three cohorts. In the puberty cohort, girls transitioning through menarche showed significant total and cortical gray matter decline (−0.13/−0.16%/month), a pattern not observed in age-matched stable premenarchal or stable postmenarchal groups. Group comparisons confirmed significant differences, specifically between menarche transitioning and the age-matched stable premenarchal group. Women becoming first- or second-time mothers also showed rates of total and cortical gray matter decline (−0.12/−0.11%/month). In line with our previous work¹⁸, these findings point to comparable rates of gray matter volume change during pregnancy and adolescence. However, our more detailed approach here presented also divergence in total cortex volume, with higher rates of gray matter volume change in girls transitioning from pre- to post-menarche compared to women becoming mothers for the first or second time.

For the menopause cohort, we observed a different pattern: women from pre-to-post menopause showed no significant cortical or total gray matter changes, whereas both stable premenopausal and stable postmenopausal groups showed significant decline. This divergent pattern, in which pre-to-post menopausal women did not show significant volumetric change while stable groups did, suggests a possible attenuation of age-related decline during the transition itself, though direct group-by-time contrasts were not significant and this interpretation should be treated as preliminary. In summary, puberty and pregnancy both showed transition-specific cortical volumetric decline, while menopause showed an absence of decline during the transition itself.

These findings are consistent with directional associations between sex steroid hormone changes and brain volumetric change. Both puberty and pregnancy are characterized by substantial increases in estrogens and progesterone, and previous research has demonstrated that brain changes during these transitions are directly linked to rising sex steroid levels^9,22,23. The coordinated volumetric reductions observed in both menarche and pregnancy transitioning groups align with this hormonal trajectory. Menopause, conversely, is characterized by declining sex steroid hormones, particularly estrogens and progesterone. We observed significant changes in total and cortical gray matter in women that were either premenopausal or postmenopausal and both timepoints, but did not see this in women that were premenopausal at the first timepoint and postmenopausal at the second timepoint, suggesting that hormonal transitions alter the trajectory of gray matter change relative to normative aging. Notably, this hormonal decline appears associated with a distinct degree of brain plasticity: whereas puberty and pregnancy showed volumetric reductions during hormonal increases, our findings point to an attenuation of ongoing age-related volumetric decline observed in stable premenopausal and postmenopausal groups. Thus, the directionality of hormonal change (increases during puberty and pregnancy versus decreases during menopause) appears linked to the direction of volumetric change, consistent with hormonal modulation of neuroplastic processes.

Regional analysis of the puberty cohort revealed that girls transitioning through menarche showed larger PC1 scores, reflecting greater alignment with the cortex-wide pattern of volumetric decline, compared to both stable premenarchal and stable postmenarchal groups. The loadings underlying this component were broadly distributed across the cortex, spanning frontal, temporal, parietal, sensorimotor, and occipital regions, without clear regional specificity. This is consistent with the cross-cohort findings showing widespread cortical volumetric decline across all 74 regions in the transitioning group relative to stable premenarchal peers, and suggests that the menarche transition is associated with an acceleration of broad cortical remodeling rather than a regionally circumscribed reorganization. This is consistent with evidence that menarche status can be distinguished from general age-related brain maturation on the basis of structural MRI data⁸, supporting the hypothesis that the transition carries a neurobiological signature beyond puberty-related development more broadly. To our knowledge, no prior longitudinal study has directly isolated the menarche transition as a discrete neurobiological event by comparing transitioning girls against both premenarchal and postmenarchal stable groups. This design allows the present findings to characterize this specific window of cortical change.

Regional analysis of the pregnancy cohort revealed that first-time mothers showed a stronger expression of the dominant pattern of coordinated cortical change compared to nulliparous controls, across broadly distributed regions spanning frontal, temporal, parietal, sensorimotor, and occipital cortices. The regions with the highest loadings on this component were clustered in posterior parietal and superior temporal areas, including the precuneus, superior temporal sulcus, and superior parietal lobule, consistent with regions showing the largest reductions in previous longitudinal work^9–14. Notably, the strength of this coordinated parieto-temporal signature varied with maternal age in first-time mothers, with older age associated with stronger expression of the dominant change pattern; this relationship was absent in second-time mothers and nulliparous controls.

Regional analysis of the menopause cohort revealed a pattern qualitatively distinct from both other transitions. Principal component analysis identified a coordinated component with loadings concentrated in frontal and temporal regions, including superior and middle frontal gyri, orbital gyri, and multiple temporal regions, that distinguished the transitioning pre-post group from stable postmenopausal women. Notably, however, the variance explained by this component was markedly lower than in the other two cohorts (23.4% versus 55.2% and 48.6% for puberty and pregnancy, respectively), and the overall effect size was small (η² = 0.008), suggesting that coordinated regional change is considerably less pronounced during the menopausal transition than during puberty or pregnancy. We note that the PCA was conducted on all three groups pooled, meaning the covariance structure underlying the solution is more heavily influenced by the stable postmenopausal group, which constitutes the large majority of the menopause cohort. This is a consequence of the group size imbalance inherent to the UK Biobank sample. Unlike the ANOVA and Kruskal-Wallis tests used in all other analyses, which do not assume equal group sizes and are not adversely affected by this imbalance, PCA is sensitive to the relative contribution of each observation to the estimated covariance matrix. As a result, the principal component is preferentially oriented along directions of variance that are most pronounced in the stable postmenopausal group, and the separation of the transitioning group along this component should be interpreted in that light: it reflects the degree to which the transitioning group deviates from the dominant pattern of variance in the pooled sample, rather than an independently defined dimension.

Cross-cohort comparison of gray matter changes in cortical regions revealed that for more than half of the cortical areas, the pubertal and pregnancy transitions showed distinct rather than convergent profiles of volumetric change. In 33 regions, a stepwise gradient was observed across all three transitions: puberty showed the greatest decline relative to controls, followed by pregnancy, followed by menopause. These stepwise regions were concentrated in motor and premotor cortex, and in cingulate cortex, while also spanning portions of prefrontal, temporal, and occipital areas. In six further regions, puberty showed greater decline than both pregnancy and menopause, while the latter two did not differ from each other. These were predominantly insular regions (i.e., the long insular gyrus and central opercular sulcus, the short insular gyrus, and the inferior circular insular sulcus) alongside the central sulcus, the lateral orbital sulcus, and the transverse temporal sulcus. The insular finding is of particular note: while the anterior and superior insular sulci showed the stepwise pattern of the larger cluster, the inferior circular insular sulcus and the short and long insular gyri showed a pubertal-specific pattern not shared by pregnancy or menopause, indicating that insular subregions differ meaningfully in their sensitivity across transitions.

Nonetheless, a substantial degree of convergence between puberty and pregnancy was also evident across 34 regions, in which control-subtracted rates in transitioning pubertal girls and first-time mothers did not differ significantly, while both cohorts showed greater decline than the menopausal transition. These convergent regions were concentrated in prefrontal cortex, parietal cortex, and temporal cortex, with additional occipital involvement. Notably absent from the convergent cluster were motor and premotor cortex and the full extent of cingulate cortex, which showed exclusively stepwise divergence. This anatomical dissociation suggests that the degree of divergence between puberty and pregnancy is greatest in regions subserving sensorimotor and interoceptive functions, while higher-order association cortex (i.e., prefrontal, parietal, and temporal) tends toward convergence across the two transitions.

The cross-cohort comparison further underscores that menopause was characterized not by accelerated volumetric loss but by its absence. Whereas puberty and pregnancy showed negative control-subtracted rates across the vast majority of the cortex, the menopausal transitioning group showed rates at or above those of stable premenopausal peers across 53 of 60 regions where menopause differed significantly from the other cohorts. The seven regions where nominally negative rates were observed had control-subtracted values close to zero (range −0.003 to −0.017 %/month), negligible in comparison to the pubertal and pregnancy rates in the same regions. Together, these findings indicate that puberty and pregnancy engage both shared as well as different mechanisms of cortical remodeling. We use the term brain plasticity here descriptively rather than mechanistically, as our structural data cannot determine whether this attenuation reflects heightened plasticity or simply a slowing of age-related decline, nor whether it carries opportunities or vulnerabilities comparable to those described during puberty and pregnancy. Menopause nonetheless represents a qualitatively distinct degree of brain plasticity, defined not by what changes, but by what does not.

Subcortical findings were limited to the puberty cohort, as for pregnancy and menopause there was a significant decline in subcortical gray matter volume. Although prior research has documented subcortical structural changes during pregnancy and the menopausal transition^9,19, those findings concern individual substructures; the aggregate total subcortical volume examined here may be insufficiently sensitive to detect such region-specific effects. The absence of subcortical effects in the menopause cohort warrants particular consideration. A growing body of literature has implicated specific medial temporal structures as selectively sensitive to the menopausal transition, with longitudinal studies reporting hippocampal gray matter reductions and increased Alzheimer’s disease biomarker burden in women transitioning through menopause¹⁹. Surgical menopause, which produces an abrupt rather than gradual decline in ovarian hormones, has been associated with smaller amygdala volumes and thinner parahippocampal-entorhinal cortex later in life²⁴, suggesting that medial temporal integrity may be sensitive to the pace of hormonal change across the transition. The present study excluded women who had undergone bilateral oophorectomy and examined only total subcortical volume, a composite metric that may dilute region-specific effects concentrated in medial temporal substructures. A similar caveat applies to the pregnancy cohort, where longitudinal studies have reported changes in structures including the hypothalamus²⁵ and hippocampus^9,10.

In contrast to the cortical pattern, where the pre-to-post group showed the most pronounced changes, the subcortical findings in the pubertal cohort revealed a qualitatively different trajectory. Whereas no significant subcortical changes were observed in the pregnancy or menopause cohorts, the pubertal cohort showed significant group differences in total subcortical volume driven by the stable premenarchal group, who showed higher rates of change than the other two groups. Regional analyses of eight bilateral subcortical structures revealed that this pattern was carried by four specific regions: the thalamus, putamen, ventral diencephalon, and hippocampus. Across all four, the stable premenarchal group showed the highest mean monthly rates of change, consistent with an ongoing subcortical growth phase that characterizes early puberty prior to menarche.

The timing at which this growth appears to attenuate, however, differed across structures. For the thalamus, putamen, and ventral diencephalon, the stable premenarchal group differed significantly from the pre-to-post group, while the pre-to-post and stable postmenarchal groups did not differ from each other. This could suggest that the attenuation of growth in these structures coincides with or begins around the menarche transition itself. This is consistent with prior longitudinal evidence that the putamen and thalamus show nonlinear trajectories in females specifically, characterized by an early peak followed by stabilization across later pubertal stages^26,27. A similar pattern of premenarchal attenuation was observed for the ventral diencephalon, a FreeSurfer-defined region comprising the substantia nigra, subthalamic nucleus, and lateral geniculate nucleus that has received limited attention in the puberty literature; prior work has found associations between pubertal timing and ventral diencephalon volume primarily in males⁷. For the hippocampus, the pattern was somewhat different: the significant group difference was between the two stable groups, with the pre-to-post group falling intermediate but not significantly different from either, suggesting that the attenuation of hippocampal growth may extend beyond the menarche transition and continue more gradually into the postmenarchal period. This is broadly consistent with evidence that hippocampal volume continues to increase across puberty, peaking around mid-to-late adolescence^27,28, and with indications from prior work that striatal volume peaks earlier in development than hippocampal volume^28,29.

The remaining four structures (i.e., amygdala, caudate, pallidum, and accumbens) did not show significant group differences after correction. Nevertheless, the stable premenarchal group showed consistently positive mean monthly rates across all eight structures examined, suggesting that the directional pattern of premenarchal volumetric growth and subsequent attenuation may be a broader subcortical phenomenon that the present study was insufficiently powered to detect across all regions. Together, these findings paint a picture of subcortical development during puberty as a region-specific process in which volumetric growth, ongoing in the premenarchal period, attenuates at different points across the pubertal transition depending on the structure.

Our findings address a critical gap in our understanding of the female brain. Despite comprising half the global population, women remain substantially underrepresented in neuroscience research, with only 2–6% of studies conducted exclusively in female samples and fewer than 5% treating sex as an analytic variable^30,31. This blind spot is particularly consequential given the female-specific hormonal transitions that shape the brain across the lifespan. Research on pregnancy-related neuroplasticity only emerged as a field roughly a decade ago, and menopause has received even less attention, with only one prior longitudinal study focusing exclusively on hippocampal changes¹⁹, leaving whole-brain cortical and subcortical change uncharacterized until now. Here, we present a longitudinal examination of whole-brain cortical and subcortical changes across the menopausal transition, directly comparing these changes with those occurring during puberty and pregnancy.

We defined transitions based on observable developmental markers, first menstruation for puberty and final menstrual period for menopause, to enable direct cross-cohort comparisons using identical analytical frameworks. Both markers were based on single self-report items rather than clinical staging, and should therefore be understood as operationally defined transition points rather than precise biological staging. While substantial hormonal changes precede both menarche and menopause onset, self-reported menarche and final menstrual period represent discrete, easily recalled events that can be consistently assessed across cohorts using a single item, in contrast to more continuous or clinically intensive staging approaches. In the puberty cohort, girls classified as premenarchal based on self-report may already be undergoing pubertal hormonal shifts, as significant hormonal changes precede the first menstruation. Similarly, the self-reported menopause question cannot distinguish pre-menopause from perimenopause, meaning that some women classified as premenopausal at T1 may already have been in the menopausal transition. In the puberty cohort, the transitioning group’s T2 scan was obtained on average 1.27 ± 0.91 years after reported menarche, and the stable postmenarchal group’s T1 scan was 2.15 ± 1.31 years after menarche. In the menopause cohort, the pre-to-post group’s T2 scan was obtained on average 2.16 ± 2.14 years after reported menopause, and the stable postmenopausal group’s T1 scan was 5.85 ± 4.42 years after menopause. These intervals reflect the temporal proximity of each scan to the reported transition, though the substantial variation, particularly in the menopause cohort, indicates considerable heterogeneity in how recently participants had transitioned at the time of scanning. Null findings in the menopausal cohort should therefore be interpreted cautiously, as reflecting the available operationally defined groups rather than as evidence of an absence of brain change across the full menopausal transition.

The inclusion of stable pre- (age-matched with transitioning groups) and post-transition groups helps to distinguish transition-specific changes from ongoing developmental or age-related processes, though this interpretation is qualified by the possibility that some women in the stable premenopausal group were already perimenopausal. Few studies have examined menarche itself as a discrete transition marker for brain structural change. However, recent machine learning approaches demonstrate that pre- versus postmenarchal status can be classified from brain structure with significant accuracy⁸. For menopause, only one prior study has examined women during perimenopause and post menopause longitudinally¹⁹, focusing on hippocampal gray matter reductions. The current study extends this work by examining total gray matter and cortical changes longitudinally with two control groups: women premenopausal at both timepoints and women postmenopausal at both timepoints.

We want to point out that we did not include males as controls, as menstruation, pregnancy, and menopause are sex-specific biological processes tied to female reproductive physiology, and any female-male comparison would conflate sex-based differences in brain aging with the hormonal transition effects under study here. Females and males have fundamentally different brain aging trajectories, with the strength of age-related volume decline differing between sexes and varying across cortical and subcortical regions, and the menopausal transition further modulating these sex differences during midlife³².

Several limitations warrant consideration. First, we used monthly rates of change to account for varying interscan intervals across participants. While this approach enables fair comparison, it necessarily simplifies known nonlinear trajectories, particularly in puberty and pregnancy, where changes may be most rapid during specific windows. Future work with more frequent assessments could better characterize these nonlinear patterns.

Second, datasets were acquired on different MRI scanners. We did not harmonize across cohorts, as cross-cohort harmonization would require assuming that menarche, pregnancy, and menopause affect the brain through sufficiently similar mechanisms to justify treating them as variants of a single biological process, an assumption that contradicts our hypothesis of distinct degree of brain plasticity. Furthermore, because cohorts were acquired on different scanners, absolute rate comparisons across cohorts should be interpreted with caution. To minimize software-dependent variability, all cohorts were processed with FreeSurfer version 7, and although minor subversion differences exist across cohorts, the FreeSurfer release notes confirm that recon-all produces identical aparc.a2009s and aseg output across 7.X releases when no new features are employed, meaning these subversion differences do not affect our results. Our primary cross-cohort inferences are based on transition-specific effect sizes, the difference between transitioning and control group rates within each cohort, rather than raw volumetric magnitudes, which reduces but does not eliminate the influence of scanner- and software-related measurement differences.

Third, while this study provides the first direct longitudinal comparison of brain structural changes across all three major female hormonal transitions, several categories of data that would further enrich the interpretation were unavailable across one or more cohorts. Although endocrine measures have been collected and published in the pregnancy cohort, harmonized hormonal data were not available across all three cohorts, meaning that while hormonal directionality can be inferred from the well-established endocrine profiles of these transitions, direct hormone-brain associations could not be systematically assessed in the current cross-cohort framework. Systematic data on hormonal contraceptive use were not consistently collected across sub cohorts, and the available pill use variables were insufficiently complete to support a formal sensitivity analysis. We cannot exclude the possibility that hormonal contraceptive use may have contributed to the observed cortical gray matter reductions, especially in prefrontal and cingulate areas, where previous studies have found associations between oral contraceptive use and cortical volume and thickness. Harmonized measures of additional covariates such as socioeconomic status, mental health, and stress levels were similarly unavailable given the independent origins and different data collection protocols of the three cohorts; individual-level variation in these factors may contribute to within-group variability in the brain change rates reported here. Relatedly, while we included parity as a covariate in the menopause cohort, harmonized reproductive history variables such as age at menarche were not available across all cohorts, precluding a systematic assessment of how cumulative reproductive history may shape brain structural change across these transitions. We hope that the current results will be followed by prospective studies with harmonized, multimodal data collection that can directly address these questions, quantify hormone-brain relationships across the female lifespan, and examine the relative contributions of reproductive, psychological, and social factors to the patterns of brain structural change observed here.

This study compares brain structural changes across three major female hormonal transitions using longitudinal data and identical analytical methods applied within a single cohort framework. By including age-matched stable control groups alongside transitioning groups, we demonstrate that the observed volumetric changes are transition-specific rather than reflecting continuous developmental or aging trajectories.

Our findings reveal both convergence and divergence across transitions. Cross-cohort comparison of cortical regions presents for more than half of the regions, pre-to-post menarche group showed higher rates of gray matter reductions than pre-to-post pregnancy, with sensorimotor and cingulate regions showing a stepwise gradient in which pubertal decline exceeded pregnancy, which in turn exceeded menopause. A smaller set of regions showed a puberal trajectory distinct from both other transitions. Nonetheless, a substantial degree of convergence between puberty and pregnancy was evident across prefrontal, parietal, and temporal association cortices, where both transitions showed comparable rates of volumetric decline relative to their controls, both markedly greater compared to menopause.

Menopause presented a fundamentally different pattern: rather than accelerated volumetric loss, women transitioning through menopause showed rates at or above those of stable premenopausal peers across the vast majority of the cortex, in striking contrast to both earlier transitions. Indeed, whereas both stable premenopausal and postmenopausal women showed significant reductions in total and cortical gray matter volume, the transitioning group did not. This suggests that the transition from pre- to postmenopausal is neurobiologically distinct from the menarche and pregnancy transition, defined not by what changes, but by what does not.

Together, these results establish that the female brain undergoes multiple forms of reorganization across these hormonal transitions across the female lifespan, with puberty and pregnancy sharing a broadly convergent signature of cortical remodeling, and menopause representing a qualitatively distinct degree of brain plasticity defined by the attenuation rather than acceleration of volumetric change. That these patterns align with the direction of hormonal change across transitions opens new questions about the mechanisms linking endocrine flux to cortical plasticity across the female lifespan.

Methods

Participants

For this study, we used three different cohorts (puberty, pregnancy, menopause) that comprised of longitudinal structural MRI data (n = 1095). For an overview of the number of participants, age, and time between sessions per group, see Table 1. Throughout this manuscript, we use sex-related terminology (female/male) when referring to the biological category assessed via self-reported sex at study intake, and use “women” and “girls” rescriptively to refer to participants within the female cohorts, consistent with how these cohorts were recruited and described in their original publications. As menarche, pregnancy, and menopause are reproductive processes tied to female biological sex, our use of these terms throughout refers to sex rather than gender identity, which was not assessed in any of the three cohorts.

Table 1.

Demographics and other variables of interest of our all three groups of all three cohorts. Mean and SD are shown. T1 is the first timepoint and T2 is the second timepoint

Characteristic	Pubertal cohort			Pregnancy cohort			Menopause cohort
Characteristic	pre-pre	pre-post	post-post	prg₁	prg₂	ctr	pre-pre	pre-post	post-post
N	49	34	59	40	30	40	49	120	674
Age at T1	12.70 ± 2.35	12.45 ± 0.95	14.64 ± 1.31	29.35 ± 3.51	32.03 ± 2.33	29.33 ± 3.57	50.49 ± 1.84	51.39 ± 2.39	55.43 ± 2.52
Months T1–T2	16.79 ± 4.11	21.44 ± 4.65	19.04 ± 5.63	16.74 ± 5.22	16.51 ± 4.11	15.04 ± 2.69	32.76 ± 14.93	43.50 ± 20.54	32.95 ± 13.71

Open in a new tab

Puberty cohort

For the puberty cohort, we used data from three studies obtained at Leiden University, BrainLinks (n = 142), BrainTime (n = 299)^6,33, and Self-Concept (n = 201)³⁴. Participants were recruited from a community sample of children, adolescents, and young adults through local schools and advertisements. At each time point, informed consent was obtained from participants or from parents in case of minors. All three studies were approved by the Institutional Review Board at Leiden University Medical Center. Participants received a financial reimbursement for their participation in the study. All participants were fluent in Dutch, right-handed, and had normal or corrected-to-normal vision, and none reported neurological or mental health problems or use of psychotropic medication at timepoint 1 based on self-report. All participants were invited to participate in three consecutive waves of assessment and neuroimaging approximately every 2 years.

We included participants who indicated “female” in response to the question about their sex (n = 319). Menarche status was determined at each timepoint based on the question “Have you gotten your first period?”. For each participant, we evaluated menarche status changes across two possible intervals: between timepoints 1 (T1) and timepoint 2 (T2), and between T2 and timepoint 3 (T3). Participants were classified as “stable pre-menarche” if menarche had not yet occurred at either timepoint, “stable post-menarche” if menarche had already occurred at both timepoints, and “pre-to-post menarche” if menarche occurred between the two timepoints.

Some participants could potentially be classified into multiple groups depending on which interval was considered. To resolve this, we applied a predefined assignment hierarchy. Participants from pre-to-post menarche between T2 and T3 were assigned to the pre-to-post menarche group using their T2–T3 interval; only if a participant had not transitioned during T2–T3 were they eligible for assignment as transitioning using T1–T2. For the stable pre-menarche group, T1–T2 intervals were used where possible (provided participants were 16 years or younger at T1); T2–T3 intervals were used only if the T1–T2 interval was unavailable or did not meet the age criterion. This ensured that each participant contributed data from exactly one interval to exactly one group. For an overview, see Supplementary Fig. 1.

Participants were excluded if they lacked sufficient longitudinal data. One participant in the stable pre-menarche group with missing data at timepoint 1 but available data at timepoints 2 and 3 was retained using the T2-T3 interval instead. Exclusions were made for participants missing timepoint 2 data (n = 12: 5 stable pre-menarche, 1 stable post-menarche, 6 pre-to-post), as including T1-T3 intervals would create outliers in follow-up duration. Additional exclusions included participants with only one available timepoint (n = 9: 3 stable pre-menarche at T1 only, 1 stable pre-menarche at T3 only, 1 pre-to-post at T1 only, 1 stable pre-menarche at T2 only, 2 pre-to-post at T2 only, 5 pre-to-post at T3 only) and those missing data at multiple consecutive timepoints (n = 3: 2 stable pre-menarche missing T2-T3, 1 stable pre-menarche missing T1-T2).

After applying the exclusions, age restrictions, and quality control procedures (see below), the final sample comprised 142 participants: 59 in the stable post-menarche group (30 from Self-Concept, 29 from BrainTime, 0 from BrainLinks), 49 in the stable pre-menarche group (3 from Self-Concept, 9 from BrainTime, 37 from BrainLinks), and 34 in the pre-to-post group (9 from Self-Concept, 25 from BrainTime, 0 from BrainLinks).

For more demographics and variables of interest, see Supplementary Table 1. After testing for normality and homogeneity of variance, we used a Kruskal-Wallis test, which revealed a significant difference in baseline age across groups (H(2) = 40.06, p < 0.001, η² = 0.28, % CI = 0.19, 0.41). Dunn post-hoc tests with Bonferroni correction indicated that the stable post-menarche group was significantly older than both the stable pre-menarche group (Z = 4.79, p < 0.001, r = −0.47, % CI = −0.63, −0.29) and the transitioning group (Z = 5.73, p < 0.001, r = −0.80, % CI = −0.88, −0.70). There was no significant difference in baseline age between the stable pre-menarche and transitioning groups (Z = 1.38, p = 0.250, r = 0.07, % CI = −0.18, 0.31). See Supplementary Fig. 2A for an age distribution plot.

Follow-up duration also differed significantly between groups (H(2) = 9.05, p = 0.011, η² = 0.06, % CI = 0.02, 0.16). Dunn post-hoc tests with Bonferroni correction indicated that the transitioning group differed significantly from both the stable pre-menarche group (Z = −2.89, p = 0.006, r = −0.40, % CI = −0.59, −0.17) and the stable post-menarche group (Z = −2.43, p = 0.023, r = 0.28, % CI = 0.04, 0.49). The stable pre-menarche and stable post-menarche groups did not differ from each other (Z = 0.63, p = 0.789, r = −0.05, % CI = −0.27, 0.16). To characterize the temporal proximity of each scan to the reported menarche, we computed the interval between self-reported age at menarche and age at the postmenarchal scan. In the pre-to-post group, the T2 scan was obtained on average 1.27 ± 0.91 years after reported menarche. In the stable postmenarchal group, the T1 scan was obtained on average 2.15 ± 1.31 years after menarche (see Supplementary Table 1).

Pregnancy cohort

We used data from a cohort of three groups of participants obtained at Leiden University Medical Center^10,11. We obtained ethics approval from the Ethics Review Board of the Leiden University Medical Center. All participants signed the informed consent forms before any study-related measurement and received monetary compensation. The study was a prospective pre-conception cohort study. There was a group with nulliparous women, of which some had the intent to become pregnant and others did not. This led to a cohort of 40 women who had become mothers between T1 and T2 (age at T1 = 29.35 ± 3.51), and a group of 40 women who did not become mothers between T1 and T2 (age at T1 = 29.33 ± 3.57). In addition, there was a group of 30 women who already had one child at T1 that had become mothers for the second time at T2 (age at T1 = 32.03 ± 2.33).

For demographics and other variables of interest, see Supplementary Table 2. Age at baseline differed significantly between groups. As the assumption of normality was violated in two of the three groups, we used a Kruskal-Wallis test, which revealed a significant difference in age across groups (H(2) = 19.90, p < 0.001, η² = 0.18, % CI = 0.08, 0.32). Dunn post-hoc tests with Bonferroni correction indicated a significant difference in age between the first-time and second-time mother groups (Z = −3.86, p < 0.001, r = −0.55, % CI = −0.71, −0.33) and between the second-time mothers and nulliparous control group (Z = 4.04, p < .001, r = 0.56, % CI = 0.34, 0.72), but not between the first-time mothers and nulliparous control group (Z = 0.24, p = 1.00, r = 0.04, % CI = −0.21, 0.28). See Supplementary Fig. 2B for an age distribution plot. There was no significant difference in time between sessions across groups (H(2) = 1.59, p = 0.451, η² = 0.01, % CI = 0.00, 0.11).

Menopause cohort. Data for analysis were obtained from the UK Biobank (www.ukbiobank.ac.uk). The UK Biobank is an international open-access resource that recruited over 500,000 individuals between 40 and 69 years between 2006 and 2010. In 2014, UK Biobank began inviting back 100,000 of the original participants for magnetic resonance imaging (MRI), including brain scans. Written informed consent was obtained from all participants and is available at www.ukbiobank.ac.uk. The UK Biobank approved the study application (Project ID = 509769).

We selected participants from whom a brain MRI was obtained at two timepoints and who assigned the sex ‘female’ to themselves. Menopause status was defined as the self-reported response to the question: “Have you had your menopause (periods stopped)?” (data field 2724). We excluded women who had undergone a hysterectomy (n = 17) or bilateral oophorectomy (n = 19) at either timepoint. Three groups were defined based on menopause status between timepoints. Our experimental group, the pre-to-post menopause group (n = 120), consisted of women who selected “No” at T1 (classified as pre- or perimenopausal) and “Yes” at T2 (classified as postmenopausal), allowing assessment of brain structural changes from pre- to postmenopausal. Note that this single-item classification does not conform to STRAW + 10 reproductive staging criteria³⁵ and cannot distinguish perimenopause from pre-menopause within the ‘No’ response category; accordingly, some women classified as premenopausal at T1 may already have been in the menopausal transition. The mean age was 51.39 ± 2.39 years at T1, see Supplementary Fig. 2C for an age distribution plot.

Our first control group, the stable premenopausal group (n = 49): Women who selected “No” at both timepoints, remaining pre- or perimenopausal throughout the study period. The mean age at T1 was 50.49 ± 1.84 years. Our second control group, the stable postmenopausal group (n = 674), consisted of women who were postmenopausal (selected ‘Yes’) at both timepoints. To minimize confounding by age, we restricted this group to women aged 47–59 years at baseline, and 51–64 years at T2, matching the age range of the pre-to-post group. This age restriction follows the approach of Than et al.³², who demonstrated that age × menopause interactions persisted even when restricting analyses to women younger than 58 years, confirming that steeper brain volume decline in postmenopausal women was not simply due to age differences between groups. This resulted in 674 age-matched controls from an initial pool of 1865 postmenopausal women, allowing for comparison of women at similar chronological ages who differ primarily in menopausal status. The mean age at T1 was 55.43 ± 2.52 years.

For demographics and other variables of interest, see Supplementary Table 3. After testing for normality and homogeneity of variance, we used Kruskal-Wallis tests, which revealed a significant difference in age across groups (H(2) = 248.33, p < 0.001, η²= 0.29, % CI = 0.26, 0.33). Dunn post-hoc tests with Bonferroni correction indicated that the stable postmenopausal group was significantly older than both the stable premenopausal group (Z = −10.38, p < 0.001, r = −0.87, % CI = −0.90, −0.82) and the pre-to-post group (Z = −12.85, p < 0.001, r = −0.74, % CI = −0.78, −0.68). There was no significant difference in age between the stable premenopausal and pre-to-post groups (Z = −1.55, p = 0.183, r = −0.23, % CI = −0.40, −0.04). Groups also differed significantly in time between visits (H(2) = 28.00, p < .001, η²= 0.03, % CI = 0.01, 0.07). Dunn post-hoc tests with Bonferroni correction indicated that the pre-to-post group had significantly more days between visits than the stable premenopausal group (Z = −3.30, p = 0.001, r = −0.32, % CI = −0.48, −0.13) and significantly fewer days between visits than the stable postmenopausal group (Z = 5.23, p < 0.001, r = 0.30, % CI = 0.19, 0.40), while the stable groups did not differ significantly (Z = −0.28, p = 1.00, r = −0.03, % CI = −0.19, 0.14).

We tested whether HRT prevalence differed across menopause groups at each timepoint using Pearson’s chi-square tests with Cramér’s V as the effect size measure. HRT prevalence differed significantly across menopause groups at baseline (χ²(2) = 20.52, p < 0.001, Cramér’s V = 0.15, % CI = 0.07, 0.22), with 6.1% of the stable premenopausal participants, 7.5% of the pre-to-post participants, and 22.5% of stable postmenopausal participants using HRT. By follow-up, these proportions increased to 14.3%, 23.5%, and 24.4% respectively, and group differences were no longer significant (χ²(2) = 2.59, p = 0.274, Cramér’s V = 0.03, % CI = 0.00, 0.10). In addition, parity was assessed using self-reported number of live births. Groups did not differ significantly in parity (H(2) = 3.90, p = 0.142, η² = 0.005, % CI = 0.00, 0.02). Dunn post-hoc tests with Bonferroni correction indicated no significant pairwise differences: stable premenopausal versus pre-to-post (Z = 1.97, p = 0.073, r = 0.18, % CI = −0.01, 0.36), stable premenopausal versus stable postmenopausal (Z = 1.65, p = .149, r = 0.13, % CI = -0.03, 0.29), and pre-to-post versus stable postmenopausal (Z = −0.92, p = 0.538, r = −0.05, % CI = −0.16, 0.06). The median number of live births was 2 across all three groups (stable premenopausal: mean = 1.78, SD = 1.14, range 0 to 4; pre-to-post: mean = 1.38, SD = 1.13, range 0 to 4; stable postmenopausal: mean = 1.50, SD = 1.15, range 0 to 6), see Supplementary Table 3.

Brain MRI acquisition and preprocessing

Puberty cohort

For all three cohorts, all MRI scans were acquired on a Philips 3 T MRI system. T1-weighted anatomical scans were obtained at each time point. BrainLinks and BrainTime were acquired on the same scanner and all three cohorts had a similar protocol.

For BrainLinks, T1-weighted images were acquired with the following acquisition parameters: repetition time (TR) = 7.9 ms, echo time (TE) = 3.5 ms, flip angle = 8°, field of view (FOV) = 250 × 196 × 170 mm, voxel size = 1.1 × 1.1 × 1.1 mm, 228 slices. For the BrainTime study, T1-weighted images were acquired with the following acquisition parameters: TR = 9.8 ms, TE = 4.6 ms, flip angle = 8°, FOV = 224 × 177 × 168 mm, voxel size = 0.875 × 0.875 × 1.2 mm, 140 slices. For the Self-Cohort subsample, T1-weighted images were acquired with the following acquisition parameters: TR = 9.751 ms, TE = 4.59 ms, flip angle = 8°, FOV = 224 × 179 × 168, voxel size = 1.17 × 1.17 × 1.20 mm, 140 slices.

All MRI scans were visually checked for quality control, and two participants had to be excluded. In addition, one participant was reassigned from T1-T2 to T2-T3 due to a complete processing failure at T1 but acceptable quality at T2 and T3. After applying these exclusions, the final sample comprised 213 participants.

Pregnancy cohort

All MRI scans were acquired on a Philips 3 T MRI system. T1-weighted images were acquired with the following acquisition parameters: TR = 9.8 ms, TE = 4.6 ms, flip angle = 8°, FOV = 178 × 224 × 168 mm, voxel size = 0.875 × 0.875 × 1 mm, 140 slices. All MRI scans were visually checked for quality control, and no scans had to be excluded.

Menopause cohort

All MRI scans were acquired on a Siemens 3.0 Tesla Skyra MRI system. T1-weighted images were acquired with the following acquisition parameters: TR = 2000 ms, TE = 2.01 ms, flip angle = 8°, FOV = 256 × 256 × 208 mm, voxel size = 1.0 × 1.0 × 1.0 mm, 208 slices. All MRI scans were visually checked for quality control, and no scans had to be excluded.

Brain analysis

The anatomical MRI images were processed in Freesurfer (version 7.4.1; version 7.2.0 for the pregnancy cohort; version 7.3.2 for the menopause cohort), using the longitudinal recon-all stream³⁶. We used the aseg (automated segmentation) and a2009s parcellation schemes, which perform automated labeling of neuroanatomical structures in the human brain^37,38.

All analyses proceeded at two levels. First, within each cohort, we compared monthly rates of gray matter volume change between groups to characterize the magnitude and direction of pre-to-post specific change. These analyses were conducted globally (total, cortical, and subcortical gray matter) and followed up with region-specific analyses where global group differences emerged, using PCA for the cortex and equivalent tests for subcortical structures. Second, to directly compare the spatial patterning of brain change across the three cohorts, we conducted cross-cohort comparisons based on control-subtracted regional rates, applying simultaneous omnibus tests across all three pre-to-post groups followed by pairwise post-hoc tests. Full details of each analytical step are provided in the sections below.

First, we wanted to compare the monthly rate of gray matter volume change across groups, to explore the magnitude of change. This metric represents the average rate of change over the observed interval and does not assume that change occurred linearly within that interval. All volumes were adjusted for total intracranial volume using the formula: adjusted volume = (raw volume/total intracranial volume) × 1000, to control for individual differences in head size. We used a similar method to Carmona et al.¹⁸, where we calculated for each subject the longitudinal change by means of the formula: percentage of change = (volume T2 − volume T1)/ volume T1 × 100. Because there were differences between groups in time between sessions (see Supplementary Table 1–3), we calculated the monthly rate of change by dividing the percentage of change/months between sessions per subject. We did this for the total gray matter volume, and to see where the change is concentrated, we also did this for the total cortical gray matter volume and the total subcortical gray matter volume. One-way ANOVA was used to test for overall group differences, followed by Tukey’s HSD post-hoc tests to examine pairwise comparisons between groups, in RStudio (version 2025.09.1). All tests were two-sided. Effect sizes for pairwise comparisons were calculated as Cohen’s d using pooled standard deviations, derived from the eff_size function in the emmeans package.

One-sample t-tests were used to assess if the groups’ monthly rate percentage of change was significantly different from zero (two-sided). Per cohort, we had three groups and three separate measures (i.e., total gray matter, cortical, and subcortical) and we therefore applied a Bonferroni corrected threshold of p = 0.0055. As there was a significant difference in HRT use at timepoint 1 in the menopause cohort between groups, we also executed the analysis only including participants that did not use HRT at timepoint 1. In addition, we redid the analysis, excluding women who had undergone a bilateral oophorectomy or hysterectomy for the menopause cohort.

As there were differences between groups in all three cohorts in age, we examined the relationship between baseline age and monthly change rates within each group. Normality assumptions were assessed using Shapiro-Wilk tests for both age and monthly rate variables. Where normality was violated (p < 0.05), Spearman’s rank correlation was used; otherwise, Pearson correlation was applied (both two-sided). Correlations were tested separately for each group to account for potential group-specific age effects. Here, we also used the Bonferroni corrected threshold of p = 0.0055.

Second, following significant group differences in global brain volume measures, we conducted region-specific analyses tailored to the anatomical level and cohort-specific findings.

For cortical gray matter, groups within each of the three cohorts exhibited differences in monthly rate changes. Given the high dimensionality of cortical data, we applied Principal Component Analysis (PCA) to identify coordinated patterns of regional change while reducing the multiple comparison burden. For this cortical PCA, we used a2009s parcellation (with the Destrieux atlas) which yielding 74 cortical regions (gyri and sulci) per hemisphere. For each cortical region, left and right hemisphere volumes were averaged to create bilateral measures.

All regional volumes were adjusted for total intracranial volume using the formula: adjusted volume = (raw volume / total intracranial volume) × 1000. For each region, longitudinal change scores were calculated by subtracting baseline from follow-up values (Δ = post − pre), resulting in 74 regional change scores per participant. Rather than expressing change as a monthly rate, absolute change scores were used as input for the PCA, as this PCA analysis aims to identify coordinated patterns of regional co-variation rather than to quantify the magnitude of change per unit time; differences in follow-up duration between participants were accounted for by including scan interval as a covariate in subsequent analyses. PCA was conducted using the prcomp function (base R stats package, RStudio version 2025.09.1), with variables centered and scaled prior to decomposition (center = TRUE, scale. = TRUE) to give equal weight to each region regardless of absolute volume.

PCA was performed on the combined dataset per cohort, with all three groups pooled, to extract data-driven patterns of brain change without a priori assumptions about group differences. The number of principal components to retain was determined by examining the scree plot and cumulative variance explained. Components accounting for substantial variance were retained for further analysis. Based on this, for all further analyses, the first component (principal component 1; PC1) was further examined. We examined the loadings from PC1 to identify which brain regions contributed most strongly to that component. Regions with absolute loadings >0.15 were considered strong contributors to the component and were visualized on the cortical surface using the Destrieux atlas parcellation as implemented in nilearn³⁹. To test for group differences in brain change patterns, we extracted individual participant scores on each retained principal component. ANOVA was used to test for overall group differences, followed by Tukey’s HSD post-hoc tests to examine pairwise comparisons between groups.

To ensure that group differences in PC1 scores were not confounded by baseline characteristics, we tested potential covariates using linear models with PC1 as the outcome, group as the predictor, and covariates of interest. We also tested group × covariate interactions to determine whether covariate effects differed by group. Age at baseline was included as a covariate across all three cohorts (puberty, pregnancy, and menopause), as there were significant age differences between groups in each cohort (see above). Follow-up duration (i.e., interval between scans) was tested and included as a covariate for the puberty and menopause cohorts. For the menopause cohort specifically, baseline HRT status was additionally included as a covariate, as there was a significant difference between groups in HRT use (see above). In addition, number of children was included as a covariate.

For subcortical structures, only the pubertal cohort showed significant group differences in total subcortical gray matter monthly rates of change. We therefore examined monthly rates across 8 bilateral subcortical regions (left and right hemispheres averaged from FreeSurfer aseg output) in this cohort only. We used the same method as for the total, cortical, subcortical changes mentioned above, including the correlation with age, but because we had eight separate measures here, we used a Bonferroni correction of p < 0.0021.

Third, to examine whether the same brain regions undergo structural changes across different female hormonal life events, we conducted a cross-cohort comparison of regional monthly rates of change. For all 74 bilateral Destrieux cortical regions, experimental group’s monthly rates of change were computed by subtracting the mean control-group rate from each pre-to-post individual’s rate within their cohort. This approach isolated neuroplastic changes specifically associated with the hormonal life event, controlling for normal age-related or developmental changes occurring in parallel. Control groups were defined as: stable premenarchal (puberty cohort), nulliparous women (pregnancy cohort), and stable premenopausal women (menopause cohort), that were all age-matched with their respective experimental group. This within-cohort subtraction approach also reduces the influence of between-cohort differences in scanner hardware and FreeSurfer version on cross-cohort comparisons, as systematic measurement offsets are partially absorbed by the control-group subtraction. Nevertheless, scanner-related differences across cohorts cannot be fully eliminated by this approach, and cross-cohort comparisons should therefore be interpreted with appropriate caution.

For each of the 74 cortical regions, the choice of omnibus test was determined empirically by first assessing two statistical assumptions: normality of the pooled residuals using a Shapiro-Wilk test, and homogeneity of variance across cohorts using Levene’s test. If either assumption was violated (p < 0.05), Kruskal-Wallis was used; if both were met, one-way ANOVA was applied instead. The majority of regions violated at least one assumption, and Kruskal-Wallis was therefore used for most regions. FDR correction (Benjamini-Hochberg) was applied across all 74 omnibus tests. For the overall cross-cohort tests, effect size was quantified as eta-squared (η²), computed as the ratio of between-group sum of squares to total sum of squares for parametric tests and via the rank-based formula implemented in the kruskal_effsize function of the rstatix package for non-parametric tests. For regions showing significant omnibus effects (FDR-corrected p <0.05), we conducted post-hoc pairwise comparisons using pairwise Wilcoxon tests following Kruskal-Wallis, and Tukey HSD following ANOVA, all two-sided, to identify which specific cohort pairs differed significantly.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information^{(1.1MB, pdf)}

Reporting Summary^{(4.5MB, pdf)}

Transparent Peer Review file^{(446.6KB, pdf)}

Source data

Source data^{(317.5KB, xlsx)}

Acknowledgements

We acknowledge the participants for their contribution to this study. This research used data assets made available by National Safe Haven as part of the Data and Connectivity National Core Study, led by Health Data Research UK in partnership with the Office for National Statistics and funded by UK Research and Innovation.

Author contributions

S.H. designed the study, analyzed data and wrote the paper, M.B. contributed to processing structural puberty cohort data, M.S. contributed to processing structural cohort data, E.C. provided structural puberty cohort data, E.H. provided structural pregnancy cohort data, supervised design and analysis of study. All authors evaluated the manuscript.

Peer review

Peer review information

Nature Communications thanks Elvisha Dhamala and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This project was supported by an Innovational Research Incentives Scheme grant (Veni, 451-14-036) by the Netherlands Organization for Scientific Research (NWO), a grant of the Leiden University Fund / Elise Mathilde Fund (CWB 740 s / 2t-03-2017 /EM), and a NARSAD grant from the Brain and Behaviour Research Foundation, U.S.A. (grant number 25312) awarded to E. Hoekzema. E. Hoekzema is currently supported by an ERC Starting Grant (948031) provided by the European Research Council.

Data availability

The menopause data used in this study are available in the UK Biobank database [https://www.ukbiobank.ac.uk]. The pregnancy data generated in this study have been deposited in the Open Science Framework repository [https://doi.org/10.17605/OSF.IO/5MT8Z]. The pubertal data generated in this study have been deposited in the Erasmus University Rotterdam Data Repository [https://dataverse.nl/dataset.xhtml?persistentId=doi:10.34894/CPARII]. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76755-2.

References

1.Mills, K. L. et al. Structural brain development between childhood and adulthood: convergence across four longitudinal samples. Neuroimage141, 273–281 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Tamnes, C. K. et al. Development of the cerebral cortex across adolescence: a multisample study of inter-related longitudinal changes in cortical volume, surface area, and thickness. J. Neurosci.37, 3402–3412 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Vijayakumar, N. et al. Brain development during adolescence: a mixed-longitudinal investigation of cortical thickness, surface area, and volume. Hum. Brain Mapp.37, 2027–2038 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Beck, D. et al. Puberty differentially predicts brain maturation in male and female youth: a longitudinal ABCD study. Dev. Cogn. Neurosci.61, 101261 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Christova, P. & Georgopoulos, A. P. Changes of cortical gray matter volume during development: a Human Connectome Project study. J. Neurophysiol.130, 117–122 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Wierenga, L. M. et al. Unraveling age, puberty and testosterone effects on subcortical brain development across adolescence. Psychoneuroendocrinology91, 105–114 (2018). [DOI] [PubMed] [Google Scholar]
7.MacSweeney, N. et al. The role of brain structure in the association between pubertal timing and depression risk in an early adolescent sample (the ABCD Study): a registered report. Dev. Cogn. Neurosci.60, 101223 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Gottschewsky, N., Kraft, D. & Kaufmann, T. Menarche, pubertal timing and the brain: female-specific patterns of brain maturation beyond age-related development. Biol. Sex. Differ.15, 1–11 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Hoekzema, E. et al. Pregnancy leads to long-lasting changes in human brain structure. Nat. Neurosci.20, 287–296 (2017). [DOI] [PubMed] [Google Scholar]
10.Hoekzema, E. et al. Mapping the effects of pregnancy on resting state brain activity, white matter microstructure, neural metabolite concentrations and grey matter architecture. Nat. Commun.13, 6931 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Straathof, M., Halmans, S., Pouwels, P. J. W., Crone, E. A. & Hoekzema, E. The effects of a second pregnancy on women’s brain structure and function. Nat. Commun.17, 1495 (2026). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Pritschet, L. et al. Neuroanatomical changes observed over the course of a human pregnancy. Nat. Neurosci.27, 2253–2260 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Servin-Barthet, C. et al. Pregnancy entails a U-shaped trajectory in human brain structure linked to hormones and maternal attachment. Nat. Commun.16, 730 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Paternina-Die, M. et al. Women’s neuroplasticity during gestation, childbirth and postpartum. Nat. Neurosci.27, 319–327 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Martínez-García, M. et al. Do pregnancy-induced brain changes reverse? The brain of a mother six years after parturition. Brain Sci.11, 1–14 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Pawluski, J. L., Hoekzema, E., Leuner, B. & Lonstein, J. S. Less can be more: fine tuning the maternal brain. Neurosci. Biobehav Rev.133, 104475 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Anderson, M. V. & Rutherford, M. D. Cognitive reorganization during pregnancy and the postpartum period: An evolutionary perspective. Evol. Psychol.10, 659–687 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Carmona, S. et al. Pregnancy and adolescence entail similar neuroanatomical adaptations: a comparative analysis of cerebral morphometric changes. Hum. Brain Mapp.40, 2143–2152 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Mosconi, L. et al. Increased Alzheimer’s risk during the menopause transition: a 3-year longitudinal brain imaging study. PLoS One13, 1–13 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Ramli, N. Z. et al. Brain volumetric changes in menopausal women and its association with cognitive function: a structured review. Front Aging Neurosci.15, 1158001 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Barth, C., Crestol, A., de Lange, A.-M. G. & Galea, L. A. M. Sex steroids and the female brain across the lifespan: insights into risk of depression and Alzheimer’s disease. Lancet Diab. Endocrinol.11, 926–941 (2023). [DOI] [PubMed] [Google Scholar]
22.Peper, J. S., Hulshoff Pol, H. E., Crone, E. A. & van Honk, J. Sex steroids and brain structure in pubertal boys and girls: A mini-review of neuroimaging studies. Neuroscience191, 28–37 (2011). [DOI] [PubMed] [Google Scholar]
23.Herting, M. M. et al. The role of testosterone and estradiol in brain volume changes across adolescence: a longitudinal structural MRI study. Hum. Brain Mapp.35, 5633–5645 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Zeydan, B. et al. Association of bilateral salpingo oophorectomy before menopause onset with medial temporal lobe neurodegeneration. JAMA Neurol. 95–95 (2019). [DOI] [PMC free article] [PubMed]
25.Spalek, K. et al. Pregnancy renders anatomical changes in hypothalamic substructures of the human brain that relate to aspects of maternal behavior. Psychoneuroendocrinology164, 107021 (2024). [DOI] [PubMed] [Google Scholar]
26.Vijayakumar, N. et al. The effects of puberty and its hormones on subcortical brain development. Compr. Psychoneuroendocrinology7, 100074 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Goddings, A.-L. et al. The influence of puberty on subcortical brain development. Neuroimage88, 242–251 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Wierenga, L. M., Langen, M., Oranje, B. & Durston, S. Unique developmental trajectories of cortical thickness and surface area. Neuroimage87, 120–126 (2014). [DOI] [PubMed] [Google Scholar]
29.Raznahan, A. et al. Longitudinal four-dimensional mapping of subcortical anatomy in human development. Proc. Natl. Acad. Sci. USA111, 38 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Rechlin, R. K., Splinter, T. F. L., Hodges, T. E., Albert, A. Y. & Galea, L. A. M. An analysis of neuroscience and psychiatry papers published from 2009 and 2019 outlines opportunities for increasing discovery of sex differences. Nat. Commun.13, 2137 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
31.de Lange, A.-M. G., Jacobs, E. G. & Galea, L. A. M. The scientific body of knowledge: Whose body does it serve? A spotlight on women’s brain health. Front Neuroendocrinol.60, 100898 (2021). [DOI] [PubMed] [Google Scholar]
32.Than, S. et al. Interactions between Age, Sex, Menopause, and Brain Structure at Midlife: A UK Biobank Study. J. Clin. Endocrinol. Metab.106, 410–420 (2021). [DOI] [PubMed] [Google Scholar]
33.Bos, M. G. N., Peters, S., Van De Kamp, F. C., Crone, E. A. & Tamnes, C. K. Emerging depression in adolescence coincides with accelerated frontal cortical thinning. J. Child Psychol. Psychiatry Allied Discip.59, 994–1002 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Achterberg, M. et al. Longitudinal associations between social media use, mental well-being and structural brain development across adolescence. Dev. Cogn. Neurosci.54, 101088 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Harlow, S. D. et al. Executive summary of the Stages of Reproductive Aging Workshop+10: addressing the unfinished agenda of staging reproductive aging. Climacteric15, 105–114 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Reuter, M., Schmansky, N. J., Rosas, H. D. & Fischl, B. Within-subject template estimation for unbiased longitudinal image analysis. Neuroimage61, 1402–1418 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Fischl, B. et al. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron33, 341–355 (2002). [DOI] [PubMed] [Google Scholar]
38.Destrieux, C., Fischl, B., Dale, A. & Halgren, E. Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature. Neuroimage53, 1–15 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Abraham, A. Machine learning for neuroimaging with scikit-learn. Front Neuroinform.8, 971 (2014). [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

Supplementary Information^{(1.1MB, pdf)}

Reporting Summary^{(4.5MB, pdf)}

Transparent Peer Review file^{(446.6KB, pdf)}

Source data^{(317.5KB, xlsx)}

Data Availability Statement

[CR1] 1.Mills, K. L. et al. Structural brain development between childhood and adulthood: convergence across four longitudinal samples. Neuroimage141, 273–281 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR2] 2.Tamnes, C. K. et al. Development of the cerebral cortex across adolescence: a multisample study of inter-related longitudinal changes in cortical volume, surface area, and thickness. J. Neurosci.37, 3402–3412 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR3] 3.Vijayakumar, N. et al. Brain development during adolescence: a mixed-longitudinal investigation of cortical thickness, surface area, and volume. Hum. Brain Mapp.37, 2027–2038 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR4] 4.Beck, D. et al. Puberty differentially predicts brain maturation in male and female youth: a longitudinal ABCD study. Dev. Cogn. Neurosci.61, 101261 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR5] 5.Christova, P. & Georgopoulos, A. P. Changes of cortical gray matter volume during development: a Human Connectome Project study. J. Neurophysiol.130, 117–122 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR6] 6.Wierenga, L. M. et al. Unraveling age, puberty and testosterone effects on subcortical brain development across adolescence. Psychoneuroendocrinology91, 105–114 (2018). [DOI] [PubMed] [Google Scholar]

[CR7] 7.MacSweeney, N. et al. The role of brain structure in the association between pubertal timing and depression risk in an early adolescent sample (the ABCD Study): a registered report. Dev. Cogn. Neurosci.60, 101223 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR8] 8.Gottschewsky, N., Kraft, D. & Kaufmann, T. Menarche, pubertal timing and the brain: female-specific patterns of brain maturation beyond age-related development. Biol. Sex. Differ.15, 1–11 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR9] 9.Hoekzema, E. et al. Pregnancy leads to long-lasting changes in human brain structure. Nat. Neurosci.20, 287–296 (2017). [DOI] [PubMed] [Google Scholar]

[CR10] 10.Hoekzema, E. et al. Mapping the effects of pregnancy on resting state brain activity, white matter microstructure, neural metabolite concentrations and grey matter architecture. Nat. Commun.13, 6931 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR11] 11.Straathof, M., Halmans, S., Pouwels, P. J. W., Crone, E. A. & Hoekzema, E. The effects of a second pregnancy on women’s brain structure and function. Nat. Commun.17, 1495 (2026). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.Pritschet, L. et al. Neuroanatomical changes observed over the course of a human pregnancy. Nat. Neurosci.27, 2253–2260 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR13] 13.Servin-Barthet, C. et al. Pregnancy entails a U-shaped trajectory in human brain structure linked to hormones and maternal attachment. Nat. Commun.16, 730 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR14] 14.Paternina-Die, M. et al. Women’s neuroplasticity during gestation, childbirth and postpartum. Nat. Neurosci.27, 319–327 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR15] 15.Martínez-García, M. et al. Do pregnancy-induced brain changes reverse? The brain of a mother six years after parturition. Brain Sci.11, 1–14 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR16] 16.Pawluski, J. L., Hoekzema, E., Leuner, B. & Lonstein, J. S. Less can be more: fine tuning the maternal brain. Neurosci. Biobehav Rev.133, 104475 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR17] 17.Anderson, M. V. & Rutherford, M. D. Cognitive reorganization during pregnancy and the postpartum period: An evolutionary perspective. Evol. Psychol.10, 659–687 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.Carmona, S. et al. Pregnancy and adolescence entail similar neuroanatomical adaptations: a comparative analysis of cerebral morphometric changes. Hum. Brain Mapp.40, 2143–2152 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR19] 19.Mosconi, L. et al. Increased Alzheimer’s risk during the menopause transition: a 3-year longitudinal brain imaging study. PLoS One13, 1–13 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR20] 20.Ramli, N. Z. et al. Brain volumetric changes in menopausal women and its association with cognitive function: a structured review. Front Aging Neurosci.15, 1158001 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR21] 21.Barth, C., Crestol, A., de Lange, A.-M. G. & Galea, L. A. M. Sex steroids and the female brain across the lifespan: insights into risk of depression and Alzheimer’s disease. Lancet Diab. Endocrinol.11, 926–941 (2023). [DOI] [PubMed] [Google Scholar]

[CR22] 22.Peper, J. S., Hulshoff Pol, H. E., Crone, E. A. & van Honk, J. Sex steroids and brain structure in pubertal boys and girls: A mini-review of neuroimaging studies. Neuroscience191, 28–37 (2011). [DOI] [PubMed] [Google Scholar]

[CR23] 23.Herting, M. M. et al. The role of testosterone and estradiol in brain volume changes across adolescence: a longitudinal structural MRI study. Hum. Brain Mapp.35, 5633–5645 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR24] 24.Zeydan, B. et al. Association of bilateral salpingo oophorectomy before menopause onset with medial temporal lobe neurodegeneration. JAMA Neurol. 95–95 (2019). [DOI] [PMC free article] [PubMed]

[CR25] 25.Spalek, K. et al. Pregnancy renders anatomical changes in hypothalamic substructures of the human brain that relate to aspects of maternal behavior. Psychoneuroendocrinology164, 107021 (2024). [DOI] [PubMed] [Google Scholar]

[CR26] 26.Vijayakumar, N. et al. The effects of puberty and its hormones on subcortical brain development. Compr. Psychoneuroendocrinology7, 100074 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR27] 27.Goddings, A.-L. et al. The influence of puberty on subcortical brain development. Neuroimage88, 242–251 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Wierenga, L. M., Langen, M., Oranje, B. & Durston, S. Unique developmental trajectories of cortical thickness and surface area. Neuroimage87, 120–126 (2014). [DOI] [PubMed] [Google Scholar]

[CR29] 29.Raznahan, A. et al. Longitudinal four-dimensional mapping of subcortical anatomy in human development. Proc. Natl. Acad. Sci. USA111, 38 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR30] 30.Rechlin, R. K., Splinter, T. F. L., Hodges, T. E., Albert, A. Y. & Galea, L. A. M. An analysis of neuroscience and psychiatry papers published from 2009 and 2019 outlines opportunities for increasing discovery of sex differences. Nat. Commun.13, 2137 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR31] 31.de Lange, A.-M. G., Jacobs, E. G. & Galea, L. A. M. The scientific body of knowledge: Whose body does it serve? A spotlight on women’s brain health. Front Neuroendocrinol.60, 100898 (2021). [DOI] [PubMed] [Google Scholar]

[CR32] 32.Than, S. et al. Interactions between Age, Sex, Menopause, and Brain Structure at Midlife: A UK Biobank Study. J. Clin. Endocrinol. Metab.106, 410–420 (2021). [DOI] [PubMed] [Google Scholar]

[CR33] 33.Bos, M. G. N., Peters, S., Van De Kamp, F. C., Crone, E. A. & Tamnes, C. K. Emerging depression in adolescence coincides with accelerated frontal cortical thinning. J. Child Psychol. Psychiatry Allied Discip.59, 994–1002 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR34] 34.Achterberg, M. et al. Longitudinal associations between social media use, mental well-being and structural brain development across adolescence. Dev. Cogn. Neurosci.54, 101088 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.Harlow, S. D. et al. Executive summary of the Stages of Reproductive Aging Workshop+10: addressing the unfinished agenda of staging reproductive aging. Climacteric15, 105–114 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR36] 36.Reuter, M., Schmansky, N. J., Rosas, H. D. & Fischl, B. Within-subject template estimation for unbiased longitudinal image analysis. Neuroimage61, 1402–1418 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR37] 37.Fischl, B. et al. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron33, 341–355 (2002). [DOI] [PubMed] [Google Scholar]

[CR38] 38.Destrieux, C., Fischl, B., Dale, A. & Halgren, E. Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature. Neuroimage53, 1–15 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR39] 39.Abraham, A. Machine learning for neuroimaging with scikit-learn. Front Neuroinform.8, 971 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

Puberty, pregnancy, and menopause show shared and distinct structural changes across the lifespan

Sophie R van ’t Hof

Marieke G N Bos

Milou Straathof

Eveline A Crone

Elseline A Hoekzema

Abstract

Introduction

Results

Changes in total, cortical and subcortical gray matter volume

Pubertal cohort

Fig. 1. Monthly rates of gray matter volume change across three female hormonal transitions.

Pregnancy cohort

Menopause cohort

Region-specific analyses

Pubertal cohort

Fig. 2. Principal component loading patterns and PC1 score distributions across the puberty, pregnancy, and menopause cohorts.

Pregnancy cohort

Menopause cohort

Comparison of regional localization across cohorts

Menarche & pregnancy > menopause

Fig. 3. Convergence between puberty and pregnancy cohort.

Menarche > pregnancy > menopause

Fig. 4. Divergence between puberty and pregnancy cohort.

Menarche > pregnancy & menopause

Discussion

Methods

Participants

Table 1.

Puberty cohort

Pregnancy cohort

Brain MRI acquisition and preprocessing

Puberty cohort

Pregnancy cohort

Menopause cohort

Brain analysis

Reporting summary

Supplementary information

Source data

Acknowledgements

Author contributions

Peer review

Peer review information

Funding

Data availability

Competing interests

Footnotes

Supplementary information

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases