Assessing volumetric brain differences in migraine and depression patients: a UK Biobank study

Migraine is a debilitating neurological disorder characterized by severe and recurrent headaches, and one of the leading causes of disability worldwide [1]. Migraine is also characterized by high degree of comorbidity with a variety of psychiatric, neurologic, vascular, and cardiac conditions [2, 3]. This interconnection is of relevance in the clinical practice, as it might influence the efficacy of treatments. Regarding psychiatric conditions, epidemiological studies have shown that migraineurs have an increased risk of developing depression, anxiety and suicidal behavior when compared to non-migraineurs [2]. More specifically, major depressive disorder (MDD) has shown to be the most frequent psychiatric diagnosis among patients with migraine, especially in chronic or with aura subtypes [3‐5]. The relation between the two disorders is bidirectional, i.e. having one disorder significantly increases the risk of manifesting the other one as well [6, 7]. This epidemiological evidence of a relationship between the two health conditions is not yet fully understood but might be the result of a partly overlapping pathophysiology that connects pathological pathways of migraine and depression.

With regard to clinical presentation, migraine and depression affect mostly women and are characterized by heterogeneity and partial overlap of symptoms [1, 8]. Their symptomatology is also shared with similar disorders, e.g. generalized anxiety disorder, and this may complicate the diagnosis and treatment selection [9]. The overlapping symptomatology and the high degree of comorbidity of migraine and depression supports the hypothesis that they might be the results of a common pathophysiological pathway. In this sense, increasing attention has been devoted in recent years to structural and functional brain imaging techniques [10, 11]. In the case of MDD, structural magnetic resonance imaging (MRI) studies have shown lower brain volumes in regions involved in emotional processing including amygdala and cingulate cortex, but also in other areas such as frontal cortex, orbitofrontal cortex, hippocampus, striatum and cerebellum [10, 12‐14]. In the case of migraine, the structural alterations of the brain are related to regions implicated in pain experience and in visual and motion processing [11, 15]. Studies have shown that migraineurs manifested lower volumes in the bilateral insula, frontal/prefrontal, temporal, parietal and occipital cortices, as well as the anterior cingulate cortex, basal ganglia and cerebellum [16, 17]. These studies showed that subjects with migraine and depression manifest overall different volumetric patterns in the brain regions analyzed (with few exceptions, such the anterior cingulate cortex and the amygdala). This could be due to the different choice of parameters used for the MRI scan. Moreover, the aforementioned studies are generally characterized by small sample sizes. To overcome these limitations, we used large-scale UK Biobank data to perform study with a larger sample size. Another important feature of the UK Biobank cohort is that for the MRI scans it was adopted a unique methodology in all the Assessment Centres. The comprehensive dataset of this cohort grants the possibility of a more in-depth analysis of the structural changes in the various brain regions and to reduce the confounding level adjusting our statistical model using many important variables.

The purpose of the current study is to provide a more comprehensive and in-depth set of measurements of subcortical volumetric changes in gray and white matter in subjects with migraine or depression in comparison to healthy controls to elucidate common and different morphological features of these two disorders.

After the exclusion of the participants without MRI brain scan, we obtained a final sample that comprises 712 individuals with migraine, 1 853 with depression and 43 930 controls. In Table 1 are summarized main sociodemographic features of the sample.

Tables 2 and 3 display the main descriptive statistics of the volumes of the different subcortical regions. In Table 2 we reported the overall volumes (gray + white matter) statistics, based on the FIRST segmentation tool, while in Table 3 we reported the descriptive statistics of the gray matter volumes of the same subcortical regions (except nucleus accumbens), based on the FAST segmentation method.

We can observe from the descriptive statistics displayed in Tables 2 and 3 some general features. In all cases, mean and median are close to each other, therefore the distributions of the volumes of each region and each group are symmetric. The standard deviations (SD) are generally quite large, as the IQRs. Summarizing these facts, we can conclude that the volumetric distributions of each region between cases and control are extensively overlapping.

To the best of our knowledge, this is the first study that addresses associations between migraine and depression diagnosis and subcortical volumetric differences using data from a large population-based cohort. Notably, participants with migraine manifested larger overall volume of the caudate (mean difference: 66 mm³, 95% CI [-3, 135]) than healthy controls. The nucleus caudate is known to have important functional connections with other brain regions which are likely to be integral in migraine pathophysiology [39]. The caudate has also been shown to manifest anti-nociceptive functions and to play a role in pain modulation in connection with the periaqueductal gray matter [40]. Therefore, abnormal activity at the level of the nucleus caudate might imply pain regulation dysfunction which could in turn increase migraine susceptibility. It has been also shown that the cortical spreading depression, a depolarization wave that is associated with migraine, inhibits the neuronal activity in the caudate [39, 41]. This reduced activity might imply disruption of pain regulation and therefore lead to migraine pain. Moreover, we found larger overall volume of the thalamus (mean difference: 103 mm³, 95% CI [-2, 208]) than controls. The thalamus is known to play an important role in migraine pathophysiology and therefore the larger thalamic volume could be due to the increased activity of this region in subjects with migraine [1, 42].

We also observed that subjects with depressive symptoms manifested a larger overall amygdala volume (mean difference: 17 mm³, 95% CI [-7, 40]) than healthy controls. This phenotype might reflect increased activity of the neurons in this brain region. Other studies support the hypothesis that the hyper-activity of the amygdala increases the risk of developing depressive symptoms and related comorbidities [43]. Increased amygdala activity has been observed in people diagnosed with general internalizing disorders [44, 45]. Moreover, it has been shown reduced amygdalar reactivity after administration of effective treatment for depression and anxiety disorders [46‐48]. These observations supports the hypothesis that an increased activity of the amygdala is associated with negative disposition, anxiety and internalizing symptoms and therefore to future development of depressive symptoms [43]. We also found lower gray matter amygdala volume (mean difference: -21 mm³, 95% CI [-44, 2]). The discrepancy between the overall and gray matter level could be explained by a compensation mechanism. Individuals with depression might have lower gray matter neurons and therefore the brain increases the white matter to compensate. The larger overall volume could be due over-compensation.

Individuals with depression appear to have also larger overall (mean difference: 47 mm³, 95% CI [-7, 100]) and gray matter (mean difference: 49 mm³, 95% CI [2, 92]) putamen volumes. The role of the putamen in depressive disorders has not been fully elucidated, and its pathophysiological involvement is currently under thorough investigation [49]. In particular, the putamen is known for playing an important role in motor control and movement disorders and for being an integral part of reward and learning circuits, which in turn play an important role in depressive conditions [50]. Previous literature has showed association between depressive symptoms and lower putamen volumes, in contrast with our findings. This could be due to different MRI and statistical analyses.

In general, we can observe that the volumetric differences were small. Moreover, even though in some cases the standard error was equal or larger than the effect sizes, the confidence intervals are also generally compatible with small effect sizes, in most of the cases. Our estimates were generally characterized by small precision, and therefore our findings should be generally treated with caution.

A strength of this study is that we provided a set of estimations of a wide variety of subcortical brain regions using a large cohort. To decrease the level of bias due to confounding we included in our model a wide set of important biological and sociodemographic predictors. Another important feature of this study is the analysis of the differences in gray and white matter volumes. Data from UK Biobank allowed us to assess not only brain volume differences among several subcortical regions, but also to address differences between white and gray matter volumes. Our study also provides many estimations that can be used in meta-analytical research to assess on stronger basis the actual direction of the associations for all the specific subcortical regions. A limitation of our study is related to the average age of the UK Biobank cohort. This sample comprises mainly older participants (mean age approximately 60 years), while migraine is known to be mostly prevalent in younger people [1]. Age does not only influence whether a participant is more likely to manifest a condition, but it has also an impact on brain structure and its functional connectivity [51]. These features pose limits on the generalization of our results. Another limitation is the cross-sectional nature of this study. We were not able to provide any causal trajectory between differences in subcortical structures and the presence of certain conditions. However, the patterns of association we found can be used to generate hypotheses that can lead to design studies assessing causal relationships. Another limitation from the UK Biobank is the sampling bias. The UK Biobank has an uncommonly low acceptance rate (around 6%) and this poses an important limitation on our research, given the presence of such a bias [52].

This study provides estimations of subcortical volumes in a broad variety of brain regions as well as mean volumetric differences between subjects with migraine, depression and healthy controls. Migraineurs manifested larger overall volumes at the level of the nucleus caudate and of the thalamus. Abnormal activity in the caudate and the thalamus might imply abnormal pain modulation and increased migraine susceptibility. Subjects with depressive symptoms manifested larger amygdala and putamen overall volumes than controls, which might be due to increased activity in these regions. Migraine and depression are not likely to manifest similar patterns at the gross anatomy level in the main sub-cortical regions. Considering the large prevalence of migraine and depression in today’s society, mapping the neural signatures of the disorders will be critical for clarifying their causes.