Normative modelling of brain morphology trajectories across the menopause transition

Menopause is an endocrine transition characterized by a substantial decline in ovarian steroid hormone production. This transition occurs across several stages (Harlow et al., 2012). Premenopause refers to the reproductive phase, following menarche and before the menopause transition. Early perimenopause is marked by increased variability in menstrual cycle length and late perimenopause is characterized by amenorrhea lasting at least 60 days, accompanied by extreme hormonal fluctuations. Perimenopause typically lasts on average 4 years. Postmenopause begins following the final menstrual period, which is confirmed retrospectively after 12 consecutive months of amenorrhea. The menopause transition is associated with a wide range of symptoms spanning multiple organ systems (Barth et al., 2023), including metabolic and cardiovascular changes, musculoskeletal symptoms, and neurological effects such as vasomotor symptoms, sleep disturbances, anxiety, depression, and changes in cognitive performance (Monteleone et al., 2018). Vasomotor symptoms, sleep disturbance, and depressive symptoms are most prevalent during late perimenopause and the early years of postmenopause, with many women experiencing a reduction of symptoms as they move further into postmenopause (Thurston et al., 2025). At the neural level, neuroimaging studies have demonstrated differences in brain morphology across menopause stages. Postmenopausal and perimenopausal women show lower gray matter volumes and smaller cortical surface areas compared to premenopausal women across several brain regions including regions in the medial temporal lobe such as the hippocampus and entorhinal cortex (Kim et al., 2024; Lu et al., 2018; Mosconi et al., 2021; Schelbaum et al., 2021; Zuhlsdorff et al., 2026). Mosconi and colleagues (2021) observed lower gray matter volume in the precuneus during perimenopause relative to postmenopause, followed by a significant recovery in the postmenopausal stage. This trajectory was specific to women and not observed in age-matched men. Importantly, this regional recovery was also positively associated with preserved memory performance. This might suggest neural adaptation during postmenopause that could account for decreased symptoms over time (Mosconi et al., 2021; Thurston et al., 2025). A cross-sectional UK Biobank study found that menopausal status modifies the association between age and gray matter volume. Specifically, the rate of age-related gray matter decline was more pronounced in postmenopausal compared to pre/perimenopausal women (Than et al., 2021). Together, these findings highlight the importance of longitudinal studies to capture dynamic, individual-level changes in brain trajectories across the menopause transition. Several factors have been shown to modulate brain aging and symptom severity during the menopause transition. Surgical menopause, defined as bilateral oophorectomy performed prior to natural menopause, has been associated with faster decline in memory performance and processing speed (Georgakis et al., 2019) as well as more severe and frequent vasomotor symptoms (Crandall et al., 2023). Women undergoing surgical menopause also showed decreased gray matter volume and cortical thickness in the medial temporal lobe compared to those with natural menopause (Zeydan et al., 2019). The effect of menopausal hormone therapy (MHT) on the brain, however, is complex and inconsistent across studies. Some studies suggest that MHT use is associated with larger gray matter volumes in postmenopausal women (Kim et al., 2023; Pintzka & Håberg, 2015; Steventon et al., 2023), while a population-based UK Biobank study found lower gray matter volumes and greater mental health challenges in MHT users compared to non-users (Zuhlsdorff et al., 2026). MHT use also did not modify the interaction between age and menopause status on brain morphology (Than et al., 2021). Overall, the effects of MHT on the brain appear to depend on several factors including timing of initiation, duration of use, formulation, and individual genetic profile (Barth et al., 2023). Additionally, reproductive history in terms of reproductive lifespan and parity has been suggested as a protective factor for brain structure. A longer reproductive span has been associated with larger gray matter volumes in several brain regions (Schelbaum et al., 2021; Steventon et al., 2023). Nulliparous women who underwent surgical menopause showed significantly lower cortical volume in the temporal lobe and precuneus compared to parous women with surgical menopause. Parous women with surgical menopause, however, showed no significant differences in cortical volume compared to women with natural menopause (Fernández-Pena et al., 2024). Finally, the prevalence and severity of menopausal symptoms, including vasomotor symptoms, mood changes, and sleep disturbance, vary across ethnic groups and are further shaped by individual factors such as personal history of anxiety or depression as well as socioeconomic circumstances including household income, social support, and education (Monteleone et al., 2018). Most existing neuroimaging studies are cross-sectional, comparing postmenopausal women to premenopausal ones rather than tracking the same individuals across the transition. Furthermore, most studies relied on small samples and group-level statistical approaches. In the present study, we address these limitations, by using a large-scale population-based UK Biobank dataset with a longitudinal subsample and by focusing on individual-level variability to investigate how menopause affects brain morphology. We will apply normative modeling to quantify how far each individual woman's brain deviates from age-expected trajectories across different menopause stages. To achieve this, we will extract brain metrics including cortical surface area, cortical thickness, and subcortical gray matter volumes from T1-weighted structural MRI (see Indices section for more details) and calculate normative deviation scores for each brain metric and each participant. We take three complementary approaches that address different aspects of menopause-related brain changes and overcome the limitations of each individual approach. In the first approach, we will apply pre-trained lifespan normative models (CentileBrain) (Ge et al., 2024) to the full UK Biobank female imaging sample in a cross-sectional analysis, characterizing population-level deviation patterns across menopause stages. This approach provides well-powered estimates of group differences. In the second approach, we will use the same pre-trained normative models in a longitudinal subsample of women from the UK Biobank who were premenopausal at their first imaging session and postmenopausal at their second. This allows us to track changes in deviation scores within each individual transitioning from pre- to postmenopause rather than comparing group averages. In the third approach, we will train a custom normative model exclusively on premenopausal women and apply it to the same longitudinal subsample. This approach provides deviation scores that directly quantify menopausal brain changes relative to the pre-menopausal reference and overcomes mixed-menopausal-status reference of the pre-trained models. Because the UK Biobank does not include a formal assessment of menopause stages, we cannot directly distinguish between premenopausal and perimenopausal women. To address this, we will take an additional approach based on a previous publication (Crestol et al., 2024) to classify women as either pre- or perimenopausal based on available variables such as time since last menstrual period and menstrual cycle length. This enables a three-group (pre-, peri-, and postmenopause) cross-sectional comparison and two longitudinal groups: women transitioning from premenopause to postmenopause, and women transitioning from perimenopause to postmenopause. We additionally include women who were postmenopausal at both imaging sessions. Together, these three longitudinal groups allow us to examine whether menopause-related brain changes follow a transient pattern of neural reorganization and partial recovery (Mosconi et al., 2021), or represent a more permanent shift in the brain's aging trajectory. Additionally, in an exploratory analysis using the three longitudinal groups, we model the temporal dynamics of brain deviation as a function of time relative to menopause onset. Across all three primary approaches, we will examine whether deviation scores are moderated by menopause type (natural vs surgical), MHT use, reproductive lifespan, and parity. We will further investigate whether individual changes in brain deviation scores are associated with menopause-related cognitive and behavioral outcomes including memory performance, mood, and sleep.