September 9, 2024

expert reaction to study of adolescent brain structure over the COVID-19 pandemic

A study published in PNAS looks at the effect of COVID-19 lockdowns on the brain structure of adolescents. 

Dr Rebecca Sheriff, Consultant Psychiatrist and Senior Clinical Research Fellow, University of Oxford, said:

“This press release appears to overstate the findings of this current study. The study shows that post lockdown female brains overall demonstrated greater than expected cortical thinning but not that the lockdown measures actually caused the thinning (or as worded here ‘resulted in a deviation from the normal pattern of cortical thinning’).

“This paper discusses other studies that have found associations between various brain changes and mental health and behavioural disorders, but did not report on mental distress or behaviour changes of the participants, so we can’t tell if the people with the brain changes experienced emotional difficulties. In addition, it does not report on other sources of stress so we can’t tell if there are other possible reasons for the brain changes. Participants were excluded if they had ever been diagnosed with a developmental or psychiatric disorder, which, given that up to one in five adolescents have a probable mental disorder, seems quite an omission.

“What would be useful to know is how these brains are with a longer follow up, i.e. does the maturation slow down or reduce to normal or below normal levels, alongside information about other sources of stress and lifestyle factors and mental, emotional and behavioural status over time. Other studies have shown that problems relating to others is noticed by adolescent girls just prior to their descent into depression and connecting with others is recognised as a way of reducing emotional distress across diverse groups of adolescents. Supporting our adolescents and young adults to connect with others seems important, however the significance of gender differences in brain structure and its relationship with COVID restrictions is unclear.”

Dr Richard Bethlehem, Assistant Professor of Neuroinformatics, University of Cambridge, said:

“The authours should be congratulated on getting such an intricate sample together during such challenging circumstances of the pandemic. This is potentially an interesting proof of concept paper on how the brains of adolescents changed over the COVID lockdown but there are some key limitations to be aware of.

“Firstly, the samples are quite small, so we need to be cautious not to generalise these findings to all adolescents.

“Secondly, there is not a huge amount of information about these samples beyond the fact that they were collected at different times during the pandemic so we cannot assume it definitely is the lockdown which is the cause of these reported changes in the brain. For example, many other things may have happened during the pandemic period such as infection with covid or a number of infections. There are many factors that are not modelled or documented in this paper which could potentially explain these findings beyond the lockdowns themselves.

“Thirdly, there is also an interesting choice of methods to completely split the different samples and use the normative modelling approach without looking at individual level trajectories (i.e not using the inherent strengths of the longitudinal sample, but instead taking a pre and post lockdown split ). We know from much of the work on normative changes in these trajectories that there is considerably individual level variation that may not be captured in a cross-sectional analyses. As is clear from general growth chart models (and also from the test-retest approach in this paper), normative models will allow some variation over time that is still in the normal range  (and also in this paper the validation seems to range quite widely between around 0.013 and 0.78 which would indicate that the normative scores estimated on the post lockdown sample from those models probably has a very broad uncertainty around it). It is hard to determine whether the observed difference between pre and post lockdown falls within such normal levels of individual variation over time (or in fact within the ‘reliability’ range of the normative model itself). Increased certainty about how generalisable the normative model is would help, but since this is based on a small sample its again hard to say how strong these effects really are. This is all purely based on the methodology.

“The results need to be more firmly and generalisably established before we can conclude that adolescent brains changed beyond normal morphological changes during the pandemic. And even if these changes are established in further study, more work needs to be done to assess what factors during the pandemic are responsible for these changes, as there are many factors beyond lockdown which would also need to be considered and studied.”

Prof Derek Hill, Professor of Medical Imaging Science and Digital Health at UCL, said:

“It is important to rigorously study the impact of the COVID-19 pandemic in adolescent health and this study is a useful contribution. It is important, however, not to over interpret the results in this paper for the following reasons:

1. They are measuring the thickness of the grey matter at the edge of the brain from MRI scans. These scans don’t actually measure number of brain cells or their connections, nor link directly to any brain symptoms. So the observation that there is “accelerated cortical thinning” does not necessarily tell us about any actual long term harm to the adolescents. 
2. While the overall number of brain scans is reasonable (160 pre pandemic and 130 post pandemic) these are subdivided into subsets for training the model, validating the model, and then testing the model to measure effect. Furthermore the analysis is divided between male and female subjects. As a result there are only 29 male and 25 female subjects in the test group, divided into 3 age groups (e.g. only 10 male and 8 female ages 16).  This relatively small number of subjects reduces the confidence that we can have in the conclusions.
3. The authors separated their data into training and validation in order to build their model of normal adolescent brain development. However they did not use truly independent datasets to test and validate their model.  When a validation set is not truly independent, there is a risk that the validation results will exaggerate the model performance. 
4. It seems very likely that the adolescents studied had different experiences of lock down, and different post-pandemic impact on their health and lives. This study does not take account of individual isolation in lockdown nor of any subsequent psychological symptoms, so the proposed link between accelerated cortical thinning and potential harm to mental health is not derived from the data in this study, but by reference to other studies.
5. The study does not report individual subject cortical thinning based in an analysis of each subject’s pre- and post- pandemic brain scans. But rather uses a model to estimate accelerated brain aging.  This does make their results sensitive to any limitations of the model. 

“While the authors propose that there is a need for “ongoing monitoring and support to individuals who were adolescents during the COVID-19 pandemic”, their paper does not demonstrate that the sort of brain scans and model used in this study could be of clinical use in that monitoring, so this paper does not support scanning of brains of adolescents who experienced lockdown.”

Prof Iroise Dumontheil, Professor of Psychology and ARC Future Fellow, School of Psychological Sciences, University of Melbourne, said:

“Investigating the impact of the Covid pandemic on adolescent brain development is a worthy investigation. This study has however a number of limitations. Specifically, research on the adolescent brain have shown that longitudinal data (where the same child is followed over time) are needed to assess developmental trajectories, that there are individual differences related to pubertal development, and that females show earlier maturation (thinning) of cortical thickness. This PNAS study has a small sample size, developmental trajectories are assessed cross-sectionally, the age range is different for the model creation pre-Covid and in the post-Covid sample, and puberty is not taken into account. I think all these factors weaken the evidence for a specific link between the Covid pandemic and the observed accelerated thinning in female adolescents.”

‘COVID- 19 lockdown effects on adolescent brain structure suggest accelerated maturation that is more pronounced in females than in males’ by Neva M. Corrigana et al. was published in PNAS at 20:00 UK time on Monday 9^th September. 

DOI: https://doi.org/10.1073/pnas.2403200121

Declared interests

Dr Rebecca Sheriff: No conflicts

Dr Richard Bethlehem: I own stock and am a director at Centile Bioscience Inc.

Prof Derek Hill: Derek has no direct relationship with the authors or funders if this research. However he is co-founder and advisor to ixico plc which provides image analysis solutions to pharmaceutical companies. He is co-founder and employee of Panoramic Digital Health that provides services to pharmaceutical companies, including related to imaging. He is a consultant to Critical Path Institute that runs pre-competitive regulatory science consortia including about brain imaging. 

Prof Iroise Dumontheil: I do not have any conflict of interest to declare.