A total of 161 women with episodic migraine (mean age 39.3 ± 12.5, range 18–73 years; disease duration 15.6 ± 11.9, range 1–57 years; attack frequency/month 5.6 ± 4.5, range 0.2–20.0; 63 with aura; 52 with white matter lesions; Table 1) were prospectively enrolled in this study. As controls, 40 age-matched healthy female subjects were included (mean age 38.3 ± 10.0, range 19–66 years). Migraineurs had no other types of headaches. None of the included migraine patients' headaches or auras were unilaterally side-locked in nature, but the headaches showed right hemisphere dominance in 68.3% (right-sided or bilateral with right > left pain intensity). An MRI was performed during a headache-free period for each patient; the minimum headache-free period was 1 day after the postdrome headache phase. All migraineurs and controls reported right-handedness. All subjects were strictly investigated for medical comorbidities (e.g., past and current medical history, physical examination, routine and autoimmune blood tests, cardiovascular monitoring, BMI, sleep disease, hypertension, gynecological disease, thyroid gland, and cognitive disease), and only those were enrolled in this study who lacked any medical diseases. Brain MRI studies of healthy participants did not show any structural abnormalities.

All subjects were scanned on the same 3T MRI scanner (Magnetom TIM Trio, Siemens AG, Erlangen, Germany) using a 12-channel head matrix coil. A whole-brain T1-weighted three-dimensional axial magnetization-prepared rapid gradient-echo (3D MPRAGE) sequence was acquired using the following parameters: TR/TI/TE = 1,900/900/3.4 ms, BW = 179 Hz/px, flip angle = 9°, FOV = 210 × 240 mm², matrix size = 224 × 256, slice thickness = 0.94 mm, 176 axial slices, 0.94 × 0.94 × 0.94 mm³ isotropic voxels. Beyond the routine T1- and T2-weighted measurements, the scanning protocol also included fluid-attenuated inversion recovery (FLAIR) imaging (TR/TI/TE = 13,200/2,600/100 ms, BW = 401 Hz/px; echo trains = 14, FOV = 186 × 220 mm², matrix size = 162 × 192, slice thickness = 1.5 mm, 100 axial slices).

Left and right thalamus segmentation was performed on the high-resolution three-dimensional 3D MPRAGE images using the Freesurfer 5.3 image analysis suite (https://surfer.nmr.mgh.harvard.edu/fswiki). The technical details were described previously (13). Quality control was performed throughout the automatic processing stream. When the reconstruction was inaccurate, error correction was performed based on the recommended workflow (http://surfer.nmr.mgh.harvard.edu/fswiki/RecommendedReconstruction). The left and right hippocampus and their subfields (head, body, and tail) were segmented using the development version of FreeSurfer on 31 August 2017 (https://surfer.nmr.mgh.harvard.edu/fswiki/DownloadAndInstall). This new hippocampal subfield segmentation method is based on a novel statistical atlas built primarily upon ultra-high-resolution (~0.1 mm isotropic) ex vivo MRI data. This ex vivo atlas outperforms the in vivo counterpart that was distributed with Freesurfer 5.3 and yields subregion volumes that agree better with those from histological studies. The technical details of this new method were described earlier (14).

WMLs were considered if visible as hyperintensity on T2-weighted and FLAIR MRI but without a low-signal intensity lesion on T1-weighted MRI and larger than 3 mm, appearing in at least two consecutive slices (Figure 1). Supratentorial WMLs were delineated manually on the FLAIR images using the 3D Slicer software (http://www.slicer.org, Version 4.6.2, Figure 2). Total and lobar hemispheric WML numbers/volumes were calculated for all L+ subjects. The borders of the lobes were defined as previously described (6).

Statistical analyses were performed using the SPSS 20.0 software (IBM Corp., Armonk, NY, United States).

Differences in age between the whole migraine group and healthy controls were assessed using the Mann-Whitney U-test. Age differences among migraine subgroups (L+ and L–) and the healthy control group were assessed using the Kruskal–Wallis test followed by the Mann-Whitney U-test. The age difference between migraine patients without aura (A–) and those with aura (A+) was assessed using the Mann-Whitney U-test.

A Wilcoxon signed-rank test was used to identify whether there were statistically significant differences between the total volume of the left and right hippocampus, as well as between the left and right thalamus, separately in the healthy control group and the migraine group. Volume differences between the healthy control group and the migraine group in the left and right hippocampus, as well as the thalamus, were assessed by ANCOVA using age and total intracranial volume (ICV) as covariates.

Differences in continuous migraine-related variables (i.e., disease duration, migraine attack frequency) between migraine subgroups (L+ and L–) were compared using the Mann–Whitney U-test, while differences in the rate of aura (A+ or A–) between the same subgroups were assessed using Fisher's exact test. Within the migraine group with lesions (L+), a Wilcoxon signed-rank test was used to test for differences between left and right hemispheres in total and lobar hemispheric WML numbers/volumes. These differences were not tested for the temporal and occipital lobes because lesions were rare in these lobes, with a median lesion number of 0.

Differences in the thalamus, hippocampal, and hippocampal subfield volumes among the three groups (L+, L–, and control) were assessed using ANCOVA with age and ICV as covariates. In the case of significance, the least significant difference post-hoc analyses were conducted for pairwise comparison of the three groups.

Within the whole migraine group, the same volumes were compared between patients with aura (A+) and without aura (A–) using ANCOVA with age and ICV as covariates. The possible effects of other migraine characteristics (i.e., disease duration and migraine attack frequency) on the examined volumes were tested using stepwise multiple linear regression analyses. In these models, age, ICV, lesion, and aura group variables were also included as possible predictors.

Within the migraine group with lesions (L+), we assessed the potential relationships between left and right hemispheric brain structure volumes and the total and lobar lesion numbers/volumes in the corresponding hemisphere. Since the lesion number/volume data were not normally distributed (right-skewed with several extreme values), to perform powerful statistical analyses, the effects of lesion numbers/volumes were assessed by creating binarized subgroup variables based on the median split of them, e.g., the subgroup with a low number/volume (below or equal to the median) vs. the subgroup with a high number/volume of left hemispheric frontal lobe lesions. The effects of these subgroup variables on the examined volumes were tested using stepwise multiple linear regression analyses, including all these subgroup variables, age, and ICV as possible predictors. Since lesions were rare in the temporal and occipital lobes with a median lesion number of 0, the effects of lesion number/volume in these lobes were not assessed. Results were considered significant at p ≤ 0.05.

In our right-handed healthy female control group, significant volume differences were detected between the thalami (R < L) and hippocampi (R > L) of the two hemispheric sides. Since in the cited morphometric studies (15–17), the participants were not separated by sex, and handedness was not documented in one study, these findings are just partly comparable with our data. However, handedness may influence the hippocampus volume on the same side.

In a large retrospective study, the authors collected data from 546 patients with migraine aged between 16 and 65 years, reporting manual dominance according to the Edinburgh test. They included 466 right-handed and 80 left-handed subjects with migraine and registered 4,215 unilateral painful attacks. It was concluded that manual dominance influences the side of pain lateralization in migraine (18). Considering our 1-month-long database, the right-handed data are congruent.

In a study of 35 patients with migraine without aura and 40 healthy controls, the authors found thalamic nuclei with significantly increased and decreased volumes. The authors concluded that the result might contribute to the underlying pathogenesis of migraine (19).

In another study, significant volume reductions of three thalamic nuclei were observed in migraineurs: the central nuclear complex, the anterior nucleus, and the lateral dorsal nucleus (20). The thalamic nuclei with reduced volumes were connected to the limbic system.

In measurements of thalamic subregions in 77 migraineurs with episodic migraine and 30 controls, selective functional hypoconnectivity in the thalamic subregions was detected (21). The data provided neuroimaging evidence of thalamocortical pathway dysfunction in episodic migraines, including emotion and personality traits in migraineurs.

In a recent thalamic study, the volume was investigated by analyzing MRI data obtained from a large study, which specifically included women migraineurs with aura, unrelated migraine-free matched controls, and migraine aura-free co-twins. None of the analyses showed any between-group differences in the volume of the thalamus or of individual thalamic nuclei. These results indicated that the pathophysiology of migraine with aura did not involve alteration of thalamic volumes (22). According to our findings, thalamic volume abnormalities were not detected in migraine aura patients in the latter publication, but we found WML-associated smaller volume.

In migraine patients, after a long follow-up for 2 years, the right hippocampus volume was positively associated with a good migraine outcome after adjustment of headache frequency (OR 4.72, p = 0.024) (23).

In a review of hippocampal volume, a longitudinal study discovered decreased volume in newly diagnosed migraine patients after 1 year (24). A cross-sectional study also suggested an adaptive increase in volume at low headache frequency and a maladaptive decrease in volume at higher headache frequency (25). These results were interpreted as either initial adaptive plasticity of the hippocampus or a larger hippocampus (i.e., a pre-existing condition) in migraineurs that then decreases with increased attack frequency (26). Evidence for the first interpretation has come from animal models of neurogenesis in the structure that is known to persist into adulthood. Stress (e.g., migraine) may mediate adaptive structural plasticity through the remodeling of dendrites and synapses (27). With repeated stress involved in migraine attacks, including pain, glucocorticoids, cortical spreading depression, and gonadal hormones, elevated and prolonged levels of excitatory amino acids are likely released (28, 29). Moreover, the aging rat hippocampus displays elevated and prolonged levels of excitatory amino acids released during acute stress. It is probable that structural plasticity in response to repeated stress starts as an adaptive and protective response but ends up as damage if the imbalance in the regulation of the key mediators is not resolved. It is likely that morphological rearrangements in the hippocampus impair memory functions, and it is conceivable that these may also have a role in chronic pain perception (30). In another study, the authors concluded that migraineurs had smaller hippocampal volume and stronger hippocampal-cortico-limbic connectivity compared to healthy subjects. Hippocampal volumes and measures of hippocampal volume connectivity with other cortico-limbic network regions are associated with symptoms of allodynia (11). These studies show hippocampal volume changes and support the role of migraine frequency in volume abnormalities. Since we also found a hippocampal tail volume association with migraine frequency, these findings are consistent with our results.

The thalamic and hippocampal volume changes are likely the consequence of recurrent migraine headaches and accumulate with migraine years (5, 31). The WMLs are worsening with longer migraine disease duration, but they can be found at any age of migraineurs (5). A previous MRI study of supratentorial white matter hyperintensities in migraine patients demonstrated tissue damage with axonal loss, decreased glial cell density with impaired energy metabolism, enlarged extracellular space with an increased extracellular water fraction, and decreased blood flow and volume (31). In addition, reactive oxygen species (ROS) are the contributors to oxidative stress, which can cause vascular endothelial dysfunction in migraines and may lead to tissue damage in WMLs (32).

Both the thalamus and hippocampus are involved in migraine bout-related pain processing, which thereby causes stress. These include headache intensity, frequency, and the dominant hemisphere attack. The pain may have a role in the volume changes, especially in the hippocampi.

Migraine aura is usually unilateral and visual, but it can also be sensory, motor, brainstem, and retinal or cause speech/language disturbance. The cortical spreading depression (CSD) is a slowly propagated wave of depolarization of neurons and glial cells, followed by a subsequent sustained suppression of spontaneous neuronal activity, accompanied by complex and variable changes in vascular caliber, blood flow, and energy metabolism. Although spreading depression (SD) has been most extensively studied in the cortex, the phenomenon may occur in all neural tissues, including the hippocampus, cerebellum, and retina, among other regions (33, 34). Growing evidence suggests that an increased propensity to CSD could be a mechanism involved in the increased prevalence of migraine in women. Both estrogen and progesterone have been reported to increase the frequency of CSD (35).

Beyond the careful new migraine patient selection, the short- or long-term follow-up studies give hope for longitudinal remeasurements of the same patient group, to see further volume changes.