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


Effects of Thyroid Disorders on Cerebellar and Subcortical Brain Volumes

After exclusion, 18 825 individuals (age [mean ± SD] = 62.7 ± 7.45; male = 48%) remained (Table 1), of whom 538 (3.1%) had an ICD-10 thyroid-related disorder (n = 419 [2.2%] hypothyroidism, n = 33 [0.2%] hyperthyroidism, and n = 86 [0.5%] other thyroid disorder or a history of both hypo- and hyperthyroidism) (Table 1). Hypothyroidism cases were, on average, 2 years younger than controls (P = 5.09e−7), while hyperthyroidism cases showed no significant difference (P = 0.17) and both disorders were more common in females. Correcting for demographic and imaging covariates, head size (inverse of head size; see Methods) was not significantly altered in either hypo- or hyperthyroidism cases compared with controls (β [95% CI] = 0.023 [−0.052, 0.098], P = 0.55 and β [95% CI] = −0.023 [−0.493, 0.034], P = 0.09, respectively).

For our main analyses, we tested for differences in total cerebellar and subcortical volumes in cases of hypo- and hyperthyroidism compared to controls (Table 2; Figure 1). For hypothyroidism, we found significant (FDR < 0.05) reductions in bilateral total cerebellar and pallidum volumes (β [95% CI] = −0.14 [−0.23, −0.06], P = 0.001 and β [95% CI] = −0.12 [−0.20, −0.04], P = 0.004, respectively). There were no volume changes in hyperthyroidism cases (P > 0.05). Exploring the effects across the cerebellum (Supplementary Table 2;[10] Supplementary Figure 1[10]), we found significant reductions across most cerebellar lobules in hypothyroidism cases, aside from superior posterior vermal regions (VI-VIIIa vermis). Only one region, the Crus II hemispheric volume, indicated possible opposing effects between hypo- and hyperthyroidism (β [95% CI] = −0.11 [−0.20, −0.03], P = 0.02 and β [95% CI] = 0.40 [0.09, 0.71], P = 0.01, respectively).

Figure 1.

Effect on total cerebellar (CB) and subcortical volumes of a diagnosis of hypothyroidism and hyperthyroidism compared to normative controls. All values are following correction for demographic and imaging covariates. Standardized beta coefficients (β) and 95% CI are provided. *signifies results with FDR < 0.05.

Since BMI was higher in subjects with hypothyroidism (Table 1) we tested whether it mediated any of the significant relationships observed between hypothyroidism and brain morphometry. We confirmed this for total cerebellar (40% [25%, 82%] mediated, P < 2e−16) and pallidum volume (23% [12%, 61%] mediated, P < 2e−16) although significant average direct effects (ADE) remained (Supplementary Table 3[10]). The results, however, were sensitive to violations of sequential ignorability, and if relatively small residual correlation between BMI and brain volumes exist, then these would make this assumption invalid (rho > −0.2 and −0.1 for cerebellar and pallidal analyses).

Effects of Genetic Risk for Thyroid Disorders on Cerebellar and Subcortical Brain Volumes

For our genetic analyses, we removed ICD-10 or self-reported cases of any thyroid-related diagnoses, leaving N = 18 255 subjects (Supplementary Table 4[10]). We tested for the effect of increasing thyroid polygenic score for hypo- and hyperthyroidism, and TSH and fT4 levels on cerebellar and subcortical volumes, controlling the FDR for the number of volumes, thyroid traits, and SNP P value thresholds (PT-values) assessed. The number of GWAS-significant SNPs at each PT-value is provided in Supplementary Table 5.[10] We found no relationships across any thyroid trait at any PT-value reaching our significance threshold when controlling FDR < 0.05 (Figure 2, Supplementary Table 6[10]), although opposing effects in the pallidum of increasing polygenic scores for hyper- and hypothyroidism were close to this threshold (PT < 1e−7: β [95% CI] = 0.02 [0.01, 0.03], P = 0.0003 and PT < 1e−3: β [95% CI] = −0.02 [−0.03, −0.01], P = 0.0003, respectively). This included finding no evidence for total cerebellar volume alterations with increased polygenic score for any thyroid trait (P > 0.05). While this was also reflected at the cerebellar lobule level for most thyroid traits, there was a fairly consistent pattern of lobule volume increases at more stringent hyperthyroidism polygenic score PT-values, although none reached our significance threshold following multiple comparison correction (Supplementary Table 7;[10] Supplementary Figure 2[10]).

Figure 2.

Effect on total cerebellar (CB) and subcortical volumes of increasing polygenic score (PGS) for hypothyroidism, hyperthyroidism, free thyroxine (fT4) and thyroid-stimulating hormone (TSH) across the different SNP P value inclusion thresholds (P T). All values are following correction for demographic, genetic, and imaging covariates. Standardized beta coefficients (β) and 95% CI are provided. Of note, no results were below our FDR < 0.05 threshold.

Genetic Association Between Thyroid Disorders and ADHD

Finally, we investigated evidence for regional genetic pleiotropy between common genetic variants important for our thyroid disorders and those for ADHD disorder. Using GWAS-pairwise analyses, we found that neither hypothyroidism nor hyperthyroidism showed any evidence for regional genetic pleiotropy with ADHD; with no SNPs observing model-3 P model-3 < 5e−8, while also observing both P gwas-1 and P gwas-2 < 1e−5.