Effects of Thyroid Status on Regional Brain Volumes

A Diagnostic and Genetic Imaging Study in UK Biobank

Tom Chambers; Richard Anney; Peter N. Taylor; Alexander Teumer; Robin P. Peeters; Marco Medici; Xavier Caseras; D. Aled Rees


J Clin Endocrinol Metab. 2021;106(3):688-696. 

In This Article


This is the first study to have examined an effect of thyroid status on regional brain volumes in adults on a large scale. We used a combination of a recorded diagnosis and genetic risk scores to confirm an effect of thyroid status on gray matter volume, notably at the level of the cerebellum and pallidum.

A recorded diagnosis of hypothyroidism was associated with modest reductions in cerebellar and pallidum volume when controlling for overall head size. These findings are consistent with clinical observations of cerebellar dysfunction in hypothyroidism and Hashimoto's encephalopathy,[24] as well as rodent studies demonstrating high TH receptor expression in the cerebellum during development[25] and TH effects on cerebellar cell differentiation, arborization, migration, and death.[1] In addition to its well-established roles in motor coordination, the cerebellum also regulates various cognitive and affective processes.[26] These are mediated via widespread connections to cortical and subcortical regions, including direct connections with the pallidum[27] and other structures within the basal ganglia. The volume reductions in hypothyroidism were evident across cerebellar lobules, with the Crus II region, a nonmotor cerebellar region with high basal ganglia connectivity,[27] being the only lobule showing both hypothyroidism-related reductions and hyperthyroidism-related increases in volume. The findings of our study thus indicate that disruption to cerebellar-pallidal pathways might be a key neuronal feature of thyroid disorders.

We found no change in hypothyroidism cases in other subcortical regions or in head size derived from the scaling applied to T1 images. These findings differ from those of a recent study, in which 70 adults with elevated TSH had significantly lower total brain volume than euthyroid controls, while low TSH was not associated with change in brain volume.[28] In both cases, total gray matter volume was unaffected by TSH status. A reduction in hippocampal volume in subjects with elevated TSH was also apparent in this and another study.[29] Since we did not observe any alterations in hippocampal or total volume, these differences may suggest that hippocampal effects occur only in active disease and resolve with appropriate treatment. Alternatively, the differences may relate to our larger sample size, different populations, or the diagnoses used. We did not find any significant alterations in subjects with a diagnosis of hyperthyroidism, although the volume effects were broadly opposite to hypothyroidism and we were likely underpowered in view of the much smaller subject numbers.

Since hypothyroidism is associated with weight gain, we undertook mediation analysis to examine the influence of BMI, finding that the effects of hypothyroidism on gray matter volume reduction were mediated in part by BMI. To our knowledge, this finding has not been reported previously, although increased BMI is recognized as a risk factor for gray matter volume loss in in older age.[30] Our findings may thus have important implications for physicians managing patients with hypothyroidism, placing an emphasis on minimization of weight gain to protect against gray matter volume loss. However, since our analysis was retrospective, further studies assessing causality should be performed. For instance, in addition to BMI, an effect of elevated weight on reducing MRI signal-to-noise ratio, including via increased head motion,[18,31] cannot entirely be excluded and should be explored in future studies, in addition to the effect of smoking status, which is known to influence thyroid function[32] as well as affecting brain volume.[33] Since direct effects on volume loss still remained when controlling for BMI, other mechanisms may also be in operation, including reduced T3 signaling, as patients established on levothyroxine replacement for hypothyroidism display reduced serum T3 levels despite normalization of TSH.[34]

We undertook our genetic substudy with an aim of replicating findings from recorded diagnosis. We found little evidence for an effect of polygenic scores on gray matter volume in most regions of interest, with the exception of the pallidum in which opposing influences of polygenic scores for hypo- and hyperthyroidism were found, albeit of borderline significance. However, it should be recognized that only a relatively small proportion of the variance in thyroid function is explained by the latest GWAS, with reduced power especially for increased or decreased TSH, since subjects with known thyroid disease were excluded. Nevertheless, the effects on gray matter volume in the pallidum raise the possibility of a shared influence of genetic risk for hypo- and hyperthyroidism with gray matter volume in this subcortical region. Large-scale genetic imaging analysis has led to the discovery of several genetic variants important for subcortical development[35] and our results indicate that exploration of genomic loci showing pleiotropy for both pallidal development and thyroid disorder might prove fruitful.

Finally, in light of several studies suggesting a link between thyroid status and ADHD risk, and common morphometric alterations at the level of the cerebellum,[5] we sought to establish if there was any genetic evidence for pleiotropy between thyroid disorders and ADHD using GWAS-pairwise analyses. Since we found no evidence for pleiotropy, this may imply that any excess risk of ADHD related to disturbed thyroid status relates to environmental influences, such as altered maternal thyroid function, rather than shared genetic etiology. Our recent observations of an increased risk of behavioral disturbances in children born to mothers exposed to excess thyroxine replacement in pregnancy[4] supports the view that the in utero environment is a critical window. This risk may be mediated, at least in part, through excess TH receptor alpha signaling in view of the high prevalence of ADHD reported in children with resistance to thyroid hormone beta.[36]

Our study has a number of strengths, including the large sample size and integrated genetic and imaging analysis. This provides a template by which future studies could be undertaken, for example when the imaging dataset in UK Biobank extends to its intended target of over 100 000 participants. However, our study also has several limitations, including a lack of thyroid function tests, which necessitated our use of recorded diagnosis and polygenic scores as markers of thyroid status. The prevalence of hypothyroidism was also lower than the most recent UK data. We suspect that this may relate to differences in sociodemographic and health-related characteristics of participants in UK Biobank, who are more likely to be older, live in less socioeconomically deprived areas, and have fewer self-reported health conditions.[37] We also confined our analyses to gray matter volume; since TH also affects myelination, future studies might also include an assessment of white matter volume, especially in childhood and adolescence when the expansion in myelin deposition is profound. In addition, as with almost all genetic imaging analyses, we found that the variance explained in our volumes by genetic risk scores was relatively low. Methods to improve this in future studies, including larger samples, more complex genetic analyses (eg, gene × gene and gene × environment interactions) and use of cerebellar-specific registration tools, would be invaluable. Finally, anatomical lobules are unlikely to offer the best separation of cerebellar function, with atlases defined by resting and task-based functional MRI scans available.[26,38]

To conclude, our study provides evidence for an effect of thyroid status on gray matter volume in adults, most notably at the level of the cerebellum and pallidum. These observations extend our understanding of the influence of TH on neuronal structure in the human brain. However, further studies are needed to replicate and extend our findings, including a focus on imaging datasets in childhood and adolescence, where any effects of altered thyroid exposure on neurodevelopment might be expected to be more profound.