Development of subcortical volumes across adolescence in males and females: A multisample study of longitudinal changes

Megan M. Herting, Cory Johnson, Kathryn L. Mills, Nandita Vijayakumar, Meg Dennison, Chang Liu, Anne‐Lise Goddings, Ronald E. Dahl, Elizabeth R. Sowell, Sarah Whittle, Nicholas B. Allen, Christian K. Tamnes

TL;DR

This study tackles inconsistencies in previous research on how subcortical brain volumes change during adolescence in males and females. By analyzing data from three independent longitudinal studies, the authors found that while males generally have larger volumes, the specific developmental trajectories and sex differences can vary significantly across samples, highlighting the impact of factors like data acquisition and sample composition.

What Problem Does This Paper Address?

Understanding how the brain develops during adolescence is crucial, especially concerning sex differences, as this period is associated with the emergence of many mental health disorders that also show sex-specific patterns. Subcortical brain structures, in particular, have been linked to these mental health issues. However, previous research on the developmental trajectories of these subcortical volumes during adolescence has yielded inconsistent findings regarding sex differences. This inconsistency makes it difficult to draw firm conclusions about how male and female brains mature differently and how these differences might relate to health and behavior. The authors aim to resolve these discrepancies by using a multisample, longitudinal approach.

Methods

The study employed advanced statistical modeling techniques to analyze longitudinal neuroimaging data. Key methods include:

  • Mixed-effects modeling: This is a standard approach for analyzing longitudinal data, allowing researchers to account for both within-individual changes over time and between-individual differences.
  • Generalized Additive Models (GAMM): This technique offers a more flexible way to model non-linear relationships between age and brain volume compared to traditional linear models. GAMM can capture complex developmental curves without pre-specifying their exact shape.
  • General Linear Mixed Effects Modeling (LME): This is a more common, but potentially less flexible, approach for longitudinal data, often used for comparison with previous studies.
  • Identical Preprocessing Pipeline: A crucial aspect of this study was the use of a consistent image processing pipeline (FreeSurfer's longitudinal pipeline) and quality control (QC) procedures across all three datasets. This standardization aimed to minimize technical variability that could contribute to inconsistent results.

Datasets

The study utilized data from three independent longitudinal neuroimaging cohorts:

  • Pittsburgh (PIT): University of Pittsburgh, USA
  • NeuroCognitive Development (NCD): University of Oslo, Norway
  • Orygen Adolescent Development Study (OADS): The University of Melbourne, Australia

These datasets collectively included 216 participants (110 females) with a total of 467 scans, covering an age range of 8 to 22 years. The study design ensured that participants had at least two scans to allow for longitudinal analysis.

Key Insights and Ideas

  • Replication is Hard: Even with standardized methods, replicating findings across different datasets is challenging, suggesting that subtle differences in data acquisition, sample characteristics, or even statistical nuances can lead to divergent results.
  • Sex Differences are Real, but Nuanced: While males generally show larger subcortical volumes than females, the way these volumes change over adolescence (the developmental trajectories) and the magnitude of sex differences can vary significantly by brain region and by the specific sample studied.
  • Consistency Varies by Region: Some subcortical regions, like the amygdala, putamen, and nucleus accumbens, showed more consistent developmental patterns and sex differences across the three samples compared to others, such as the thalamus and pallidum.
  • GAMM's Flexibility: The use of GAMM proved valuable in capturing non-linear developmental trajectories, offering a more nuanced understanding of brain maturation than traditional linear models, especially when examining sex differences.

Key Results

  • Overall Sex Differences: Across all participants, males consistently showed larger volumes in all examined subcortical regions compared to females.
  • Age-Related Changes: Significant non-linear changes with age were observed in several regions, including curvilinear growth in the pallidum and amygdala, and decreases in the caudate, putamen, and nucleus accumbens.
  • Sample-Specific Variations: Conspicuous differences were found in developmental trajectories between the samples for regions like the thalamus, pallidum, caudate, and hippocampus. This suggests that findings from one study may not generalize directly to another without considering these sample-specific effects.
  • Region-Specific Consistency: The trajectories for the amygdala, putamen, and nucleus accumbens were the most consistent across the three samples.

Limitations and Future Work

  • Sample Composition and Acquisition: The authors acknowledge that factors such as population differences, sampling strategies, scanning protocols, and age ranges likely contribute to the observed discrepancies between samples.
  • Statistical Power: Examining each sample separately (with smaller Ns) reduced statistical power compared to analyzing the combined dataset.
  • Puberty Metrics: Pubertal status was not consistently measured across all cohorts, preventing the authors from investigating its specific role in amygdala and nucleus accumbens development. Future research should aim to include standardized pubertal metrics.
  • FreeSurfer Version: The study used FreeSurfer 5.3, but FreeSurfer 6.0 was released during the project. Future studies should consider the impact of newer versions, especially for regions like the putamen.
  • Motion Artifacts: While efforts were made to control for motion, it remains an important confound in developmental studies. Future work could benefit from standardized QC and motion correction procedures.
  • Generalizability: The authors suggest that focusing on patterns of change (periods of stability/change) rather than strict model terms might be more useful for understanding developmental trajectories across different studies. They also advocate for sharing statistical code to facilitate the testing of published models on new datasets.
  • Hormonal Influences: Future research is warranted to explore the contributions of hormones to sex differences in the neurodevelopmental trajectories of the amygdala and nucleus accumbens.

Summary updated

Abstract

Abstract The developmental patterns of subcortical brain volumes in males and females observed in previous studies have been inconsistent. To help resolve these discrepancies, we examined developmental trajectories using three independent longitudinal samples of participants in the age-span of 8-22 years (total 216 participants and 467 scans). These datasets, including Pittsburgh (PIT; University of Pittsburgh, USA), NeuroCognitive Development (NCD; University of Oslo, Norway), and Orygen Adolescent Development Study (OADS; The University of Melbourne, Australia), span three countries and were analyzed together and in parallel using mixed-effects modeling with both generalized additive models and general linear models. For all regions and across all samples, males were found to have significantly larger volumes as compared to females, and significant sex differences were seen in age trajectories over time. However, direct comparison of sample trajectories and sex differences identified within samples were not consistent. The trajectories for the amygdala, putamen, and nucleus accumbens were most consistent between the three samples. Our results suggest that even after using similar preprocessing and analytic techniques, additional factors, such as image acquisition or sample composition may contribute to some of the discrepancies in sex specific patterns in subcortical brain changes across adolescence, and highlight region-specific variations in congruency of developmental trajectories.
BibTeX
@article{herting2018development,
  author = {Megan M. Herting and Cory Johnson and Kathryn L. Mills and Nandita Vijayakumar and Meg Dennison and Chang Liu and Anne‐Lise Goddings and Ronald E. Dahl and Elizabeth R. Sowell and Sarah Whittle and Nicholas B. Allen and Christian K. Tamnes},
  title = {{Development of subcortical volumes across adolescence in males and females: A multisample study of longitudinal changes}},
  journal = {NeuroImage},
  year = {2018},
  doi = {10.1016/j.neuroimage.2018.01.020},
}
Figure