Associations between multi-method latent factors of puberty and brain structure in adolescent girls

Michelle L. Byrne, Nandita Vijayakumar, Samantha Chavez, John C. Flournoy, Theresa W Cheng, Kathryn L. Mills, Marjolein E.a. Barendse, Arian Mobasser, Jessica Flannery, Benjamin W. Nelson, Wen Wang, Elizabeth A. Shirtcliff, Nicholas B. Allen, Jennifer H. Pfeifer

TL;DR

This study reveals that how we measure puberty significantly impacts our understanding of its link to brain development. By combining hormone levels and self-reported physical changes in adolescent girls, researchers found that advanced puberty is associated with thinner posterior cortical regions, but this association varies depending on whether we focus on adrenarche or gonadarche, and the measurement method used.

What Problem Does This Paper Address?

Puberty is a critical period of brain development, heavily influenced by hormonal and physical changes. However, research on how these pubertal changes relate to brain structure has yielded inconsistent results. This inconsistency is likely due to the varied ways researchers measure puberty. Some studies rely solely on self-reported physical characteristics (like breast development or pubic hair growth), while others use hormone levels. This paper addresses the challenge of integrating these different measurement approaches to get a more comprehensive picture of pubertal development and its impact on the adolescent brain. Specifically, it aims to disentangle the distinct processes of adrenarche (driven by adrenal glands) and gonadarche (driven by the pituitary-gonadal axis) and understand how they, individually and combined, relate to cortical thickness.

Methods

The researchers employed a multi-method approach to measure puberty, combining two main categories:

  • Hormone Levels: Saliva samples were collected weekly from participants, and hair samples were also collected. These were assayed for dehydroepiandrosterone (DHEA), testosterone (T), and estradiol (E2).
  • Self-Reported Physical Characteristics: Participants completed the Pubertal Development Scale (PDS) and used Morris & Udry Line Drawings (LD) to indicate their stage of development for various physical traits.

These diverse measures were then integrated using structural equation modeling (SEM). The core of their methodological innovation lies in developing latent factors that represent underlying pubertal processes. They tested both a single "overall puberty" factor and separate "adrenarche" and "gonadarche" factors. These latent factors were then correlated with cortical thickness estimates derived from high-resolution T1-weighted MRI scans. A key aspect of their analysis was to explore effect sizes rather than relying solely on p-values, aiming to provide a more robust understanding of the magnitude of associations. They also conducted post-hoc analyses to examine the influence of chronological age on these relationships.

Datasets

The study utilized data from the Transitions in Adolescent Girls (TAG) Study. This cross-sectional dataset included 174 early adolescent girls aged 10-13 years. MRI data was available for a subsample of 112 participants. No new benchmarks were introduced; the study focused on integrating existing measurement techniques within a novel modeling framework.

Key Insights and Ideas

  • Puberty is not monolithic: The study strongly suggests that treating puberty as a single, uniform process can obscure important nuances. Differentiating between adrenarche and gonadarche, and considering how different measurement methods capture these, is crucial.
  • Measurement matters: The strength and even direction of associations between pubertal development and brain structure can change dramatically depending on whether you use self-reports, hormones, or a combination. This highlights the need for careful consideration of measurement choices in future research.
  • Posterior cortex is sensitive: Advanced pubertal development, particularly when measured holistically, is consistently linked to thinner cortical regions in the back of the brain (occipital and parietal lobes). This aligns with theories suggesting earlier maturation of sensory areas.
  • Adrenarche is complex: The link between adrenarche and brain structure appears weaker and more nuanced when examined with a multi-method approach compared to gonadarche. This suggests that self-reports and hormones might capture different aspects of adrenarcheal development, and their combined effect on the brain is less straightforward.
  • Age is a confounder, but also informative: While controlling for age can weaken some associations, it also reveals that pubertal timing (when puberty occurs relative to age) might have different effects on the brain than pubertal stage itself.

Key Results

  • A one-factor model (overall puberty) and a two-factor model (adrenarche and gonadarche) both provided good fit for the multi-method pubertal data, suggesting that both approaches are viable depending on the research question.
  • The overall puberty factor showed negative associations with cortical thickness in the posterior cortex (occipital and parietal lobes).
  • When examining adrenarche and gonadarche separately, the multi-method approach revealed weaker cortical associations for adrenarche compared to gonadarche.
  • Hormone-only models generally showed weaker correlations with cortical thickness compared to self-report models.
  • Controlling for chronological age weakened some of the negative associations between pubertal stage and posterior cortical thickness, particularly for self-report measures.

Limitations and Future Work

  • Sample Diversity: The study's sample was limited in racial and ethnic diversity, which is a significant limitation given that pubertal processes can vary across different groups. The authors emphasize the need for more diverse samples in future research.
  • Cross-Sectional Design: The study is cross-sectional, meaning it captures a snapshot in time. It cannot establish developmental trajectories or causal relationships. Longitudinal studies are needed to track these changes over time.
  • Subcortical Structures: The study focused on cortical thickness, excluding subcortical structures which often exhibit more complex, non-linear developmental patterns.
  • Covariates: Factors like body mass index and early life stress were not included as covariates, which could influence pubertal and brain development.
  • Measurement Nuances: The authors acknowledge that the duration of saliva collection (over a month) differed from hair hormone collection (aggregated over five months), which could impact comparisons. They also note that physical exams, which can capture earlier adrenarcheal changes missed by self-reports, were not included.

Future work should focus on replicating these findings in larger, more diverse, and longitudinal samples. The authors suggest exploring the growth hormone/IGF-1 axis more thoroughly, as well as investigating the distinct roles of pubertal timing versus pubertal stage in brain development. Further research is also needed to understand the specific mechanisms by which adrenarche and gonadarche influence brain structure, and how different measurement methods contribute to our understanding of these processes.

Summary updated

Abstract

Pubertal processes are associated with structural brain development, but studies have produced inconsistent findings that may relate to different measurements of puberty. Measuring both hormones and physical characteristics is important for capturing variation in neurobiological development. The current study explored associations between cortical thickness and latent factors from multi-method pubertal data in 174 early adolescent girls aged 10-13 years in the Transitions in Adolescent Girls (TAG) Study. Our multi-method approach used self-reported physical characteristics and hormone levels (dehydroepiandrosterone (DHEA), testosterone (T), and estradiol (E2) from saliva) to estimate an overall pubertal factor and for each process of adrenarche and gonadarche. There were negative associations between the overall puberty factor representing later stage and thickness in the posterior cortex, including the occipital cortices and extending laterally to the parietal lobe. However, the multi-method latent factor had weaker cortical associations when examining the adnearcheal process alone, suggesting physical characteristics and hormones capture different aspects of neurobiological development during adrenarche. Controlling for age weakened some of these associations. These findings show that associations between pubertal stage and cortical thickness differ depending on the measurement method and the pubertal process, and both should be considered in future confirmatory studies on the developing brain.
BibTeX
@inproceedings{byrne2023associations,
  author = {Michelle L. Byrne and Nandita Vijayakumar and Samantha Chavez and John C. Flournoy and Theresa W Cheng and Kathryn L. Mills and Marjolein E.A. Barendse and Arian Mobasser and Jessica Flannery and Benjamin W. Nelson and Wen Wang and Elizabeth A. Shirtcliff and Nicholas B. Allen and Jennifer H. Pfeifer},
  title = {{Associations between multi-method latent factors of puberty and brain structure in adolescent girls}},
  booktitle = {Developmental},
  year = {2023},
  doi = {10.1016/j.dcn.2023.101228},
}
Figure