Abstract
Objectives
To investigate the structural changes of hippocampus and amygdala and their relationships with migraine frequency and prognosis.
Methods
Hippocampus and amygdala volumes were measured by 3-T brain magnetic resonance imaging (MRI) in 31 controls and 122 migraine patients who were categorized into eight groups by headache frequency: group 1 (1–2 headache days/month), 2 (3–4), 3 (5–7), 4 (8–10), 5 (11–14), 6 (15–19), 7 (20–24), and 8 (25–30). Headache frequency was reassessed 2 years later and a frequency reduction ≥50% was regarded a good outcome.
Results
Hippocampus and amygdala volumes fluctuated in patient groups but did not differ from the controls. In migraine patients, the bilateral hippocampus volumes peaked in group 3. The volumes and headache frequencies correlated positively in groups 2–3 on bilateral sides (L: r = 0.44, p = 0.007; R: r = 0.35, p = 0.037), and negatively in groups 3–7 on the left side (5–24 days/month; L: r = −0.31, p = 0.004) and groups 3–8 on the right side (r = −0.31, p = 0.002). The left amygdala volume also peaked in group 3, and correlated with headache frequency in groups 1–3 (r = 0.34, p = 0.020) and groups 3–6 (r = −0.30, p = 0.012). The volumetric changes of the right amygdala with headache frequency did not reach statistical significance. At 2-year follow-up, the right hippocampus volume was positively associated with a good migraine outcome after adjustment of headache frequency (OR 4.72, p = 0.024).
Conclusions
Hippocampus and amygdala display a structural plasticity linked to both headache frequency and clinical outcome of migraine.
Introduction
Episodic migraine (<15 monthly headache days) may evolve to chronic migraine (≥15 days) with an unknown mechanism. On the one hand, migraine chronification with an increase of attack frequency brings about more psychiatric comorbidity and higher suicide risk (1). Furthermore, depression and stressful life events are well-known risk factors for migraine chronification (2). This mutual relationship implicates a link between the limbic system and migraine frequency, as the limbic system is involved in pain processing, emotion modulation and stress adaptation (3,4).
A maladaptive stress response has recently been proposed as one of the mechanisms underlying migraine chronification (5,6). Two subcortical limbic structures that are closely related to stress adaptation, hippocampus and amygdala, change structurally and functionally under stressful and chronic pain conditions (7). Moreover, smaller hippocampus volume may contribute to the persistent pain state (8). Among the few studies in migraine, one showed a larger hippocampus volume in patients with 1–2 headache days/month compared with those with 8–14 headache days/month and the controls (9). Another study reported a smaller amygdala volume in chronic migraine than in episodic migraine (10). It is unknown whether these structural changes are longitudinally linked to migraine prognosis.
The present study investigated hippocampus and amygdala volumes in migraine patients with a full spectrum of headache frequencies, and examined the association between these limbic volumes at baseline and the 2-year migraine outcome. It is hypothesized that limbic structures may not only reflect current headache frequency but also associate with future outcome of migraine.
Methods
Study subjects
Patients aged 20–60 years old diagnosed with migraine were prospectively surveyed by headache specialists at the Headache Clinic of the Taipei Veterans General Hospital from May 2011 to July 2013. The diagnosis of episodic migraine followed the second version of International Classification of Headache Disorders (ICHD-2) (11), whereas the diagnosis of chronic migraine was based on the 2006 revised criteria (12). Medication overuse was defined by the same 2006 criteria. Patients who received medications other than acute abortive therapy on a regular basis were excluded. Volunteers who did not have past or family histories of migraine, nor any headache attack during the past year served as healthy controls (HC). All participants were right-handed, denied any history of systemic or major neurological diseases, and presented with normal neurological examinations.
Information that would potentially expose individual identity was encrypted. Informed consent forms were completed by all the participants after receiving an explanation of the study. The whole study protocol was approved by the Institutional Review Board of the same hospital.
Study design
A questionnaire was filled out by all the participants at their first visit to obtain demographic information and evaluate the severity of depression and anxiety using the Hospital Anxiety and Depression Scale (HADS) (13). Patients with migraine also completed a semi-structured questionnaire to report their headache profile. Headache frequency was defined as the average number of headache days/month in the recent 3 months. Each subject underwent scheduled magnetic resonance imaging (MRI) during an interictal period, which was defined as absence of acute migraine attack within 2 days prior and subsequent to the date of image acquisition. The scanning was re-scheduled if there was an acute migraine during this period or use of analgesics, triptans, or ergots for any reason within the last 48 hours prior to scanning.
As dividing the patients into two groups (such as low versus high headache frequency, or episodic versus chronic migraine) may obscure the potential limbic plasticity in relation to headache frequency, the present study divided the patients to eight groups, based on the number of headache days/month: group 1 (1–2 headache days/month), group 2 (3–4 days/month), group 3 (5–7 days/month), group 4 (8–10 days/month), group 5 (11–14 days/month), group 6 (15–19 days/month), group 7 (20–24 days/month), and group 8 (25–30 days/month). This arbitrary categorization was based on earlier studies showing the associations of headache frequency with clinical profile or neuroimaging measures (9,14–16), and to keep at least 10 subjects in each frequency group.
MRI data acquisition
MRI data were acquired on the same 3.0 T GE Discovery MR750 scanner (General Electric Healthcare, Milwaukee, WI) using a standard eight-channel phase array head coil at Taipei VGH. Anatomical scans were acquired using a T1-weighted three-dimensional inversion recovery prepared fast-spoiled gradient-recalled echo sequence (IR-FSPGR) with the following parameters: repetition time/echo time/inversion time = 9.2/3.7/450 ms; flip angle = 12°; matrix size = 256 × 256; FOV = 256 × 256 mm2; slice thickness = 1 mm without inter-slice gap; number of excitations = 1, and 168 axial contiguous slices. All imaging data were acquired parallel to the anterior commissure–posterior commissure line.
Volumetric estimation of global brain tissues
In the native space of each participant, global gray matter volume (GMV), white matter volume (WMV), cerebrospinal fluid volume (CSFV), and total intracranial volume (TIV) were calculated using the VBM8 toolbox (http://dbm.neuro.uni-jena.de/vbm/) and the default settings in Statistical Parametric Mapping software (SPM8, http://www.fil.ion.ucl.ac.uk/spm/). The detailed method of using a pipeline algorithm to assess global brain tissue volume has been previously published (17,18).
Subcortical volumetric analysis of structural brain MRI
The volumes of the bilateral hippocampus and amygdala were obtained from the T1-weighted anatomical scan using the FreeSurfer (http://surfer.nmr.mgh.harvard.edu/) with default analysis settings (19). The details of this image processing pipeline are described in previous publications (20,21). After processing, each voxel of the MRI volume was automatically assigned a neuroanatomical label, and the bilateral hippocampus and amygdala were extracted for further statistical analysis. An experienced neuroradiologist who was blind to participants’ status visually checked errors for all image processing steps, including initial skull-striping, spatial registration, and subcortical segmentation according to FreeSurfer guidelines.
Follow-up
Two years after the MRI scan, each patient was contacted by telephone by the same physician (HYL) to assess their migraine status, including their headache frequency within the last 3 months. Based on the comparison of this frequency with that at their first visit, good outcome was defined as ≥50% reduction in headache frequency, whereas poor outcome was defined as <50% reduction in headache frequency or having an even higher frequency. Patients who could not be contacted due to a change in phone number, unanswered phone calls, or erroneously recorded phone numbers were treated as patients with missing data.
Statistical analysis
The descriptive data in demographics and clinical profiles are presented as mean ± standard deviation or percentage. The chi-square test was used to test the difference in categorical data. Student’s t-test and analysis of variance (ANOVA) were used to compare the means of continuous variables.
The comparisons of group difference (migraine vs. HC, aura, and medication overuse) in global tissues volumes, the left, and the right hippocampus and amygdala volumes were conducted by ANCOVA. Age and sex were regressed out as covariates of no interest for comparing TIV, whereas age, sex, and TIV were regressed out as covariates of no interest for comparing the other volumes. The associations of GMV and the two limbic structural volumes with disease duration were tested by Pearson’s correlation, controlled by age, sex, and TIV.
The adjusted volumes (adjusted volume 1) of all participants’ hippocampi and amygdala were obtained using a General Linear Model, with age, sex, HADS score and TIV entered as covariates of no interest. The Shapiro–Wilk test was used to examine the data distribution of the adjusted limbic volumes in eight subgroups of the patients and HC. The comparisons of the adjusted limbic structural volumes between specific patient subgroups were according to the exploratory changes of limbic volumes with headache frequency (Figures 1 and 2). Student’s t-test was used to compare the group with mean volume at peak and the two groups at valley on the bilateral sides, respectively. The Bonferroni method was used to conduct multiple comparisons. The associations between adjusted volumes and headache frequency (monthly headache days) were tested with Pearson’s correlation, and the multiple comparisons were conducted by Hochberg’s step-up method.
Volumetric changes of the bilateral hippocampi with headache frequency. The Y-axis indicates the adjusted volume of hippocampus (means ± standard error). The dotted line indicates the mean adjusted hippocampus volume of healthy controls. The bilateral hippocampus volumes fluctuated with headache frequency and peaked at 5–7 monthly headache days. The correlations between hippocampus volume and monthly headache days were positive within 3–7 headache days on both sides, and negative within 5–24 headache days on the left side and 5–30 headache days on the right side, as denoted by *. Volumetric changes of the bilateral amygdala with headache frequency. The Y-axis indicates the adjusted volume of amygdala (means ± standard error). The dotted line indicates the mean adjusted amygdala volume of healthy controls. The relationship between amygdala volume and headache frequency differed between both sides. On the left side, the volume fluctuated across headache frequency and peaked at 5–7 monthly headache days, and the volume–headache days correlation was positive within 1–7 headache days and negative within 5–19 headache days, as denoted by *. On the right side, the volume plotted as a function of headache frequency rose gradually from 3–4 headache days and peaked at 20–24 headache days, but the volumetric difference between them was not significant.
A logistic regression model was used to test the association between structural volumes and 2-year migraine outcome. Covariates including age, sex, headache frequency (headache days/month), and HADS score were controlled by the enter method, whereas the adjusted volume (adjusted volume 2) of each structure was selected by forward LR method. The frequency of acute abortive medications usage was not controlled, as it is highly correlated with headache frequency (r = 0.57, p < 0.001). Adjusted volume 2 was obtained using a General Linear Model in patients only, with TIV entered as a covariate of no interest.
The threshold for statistical significance was taken to be p-value less than 0.05 (two-tailed) for all the above analyses.
Results
Demographics, clinical profiles and distribution of headache frequencies
Clinical profiles and imaging data in patients with migraine and healthy control (HC) subjects.
HADS: Hospital Anxiety and Depression Scale. p < 0.05 vs. migraine.
Volumetric data
The volumes of global brain structures, including GMV, WMV, CSFV, and TIV, are listed in Table 1. The TIV and GMV were smaller in patients with migraine than in HC (p = 0.030 and 0.002, respectively), but the WMV did not differ between the two groups. The GMV was not correlated to disease duration in patients with migraine. Moreover, the GMV was not influenced by the presence of aura (vs. without aura) or medication overuse (vs. no medication overuse) in patients with migraine.
The volumes of hippocampus and amygdala were no different between patients with migraine and HC, and were not correlated with disease duration either. Again, the volumes of hippocampus and amygdala were not influenced by the presence of aura or medication overuse.
Volumetric changes of hippocampus in relation to headache frequency
In both the right and the left hippocampi, the mean adjusted volume varied as a function of headache frequency and peaked in group 3 (Figure 1; unadjusted raw volumes shown in Supplementary Figure 1). The left hippocampus volume was larger in group 3 than in group 2 (p = 0.010) and group 7 (p = 0.020). The right hippocampus volume was larger in group 3 than in group 2 (p = 0.022) and group 8 (p = 0.026). The volume–headache days correlation was positive in groups 2–3 (3–7 headache days/month; L: r = 0.44, p = 0.007; R: r = 0.35, p = 0.037) on bilateral sides, and negative in groups 3–7 on the left side (5–24 days/month; L: r = −0.31, p = 0.004) and in groups 3–8 on the right side (5–30 days/month; r = −0.31, p = 0.002). The headache days in group 3 (5–7 days/month) were not correlated with bilateral hippocampal volumes. The hippocampus volumes were comparable between each patient group and the HC.
Volumetric changes of amygdala in relation to headache frequency
The mean adjusted volume of the left amygdala plotted as a function of headache frequency fluctuated and peaked at group 3 (Figure 2; unadjusted raw volumes shown in Supplementary Figure 2). The left amygdala volume was larger in group 3 (peak group) than in group 1 (p = 0.034) and group 6 (p = 0.014). The correlation of volume–headache days was positive in groups 1–3 (1–7 days/month, r = 0.34, p = 0.020) and negative in groups 3–6 (5–19 days/month, r = −0.30, p = 0.012).
The mean adjusted right amygdala volume plotted as a function of headache frequency rose gradually from group 2 until reaching its peak at group 7 (Figure 2). However, the difference of the right amygdala volume between group 2 and 7 was not significant (p = 0.07). The bilateral amygdala volumes were comparable between each group of patients and the HC.
Association between limbic structural volumes and migraine outcome
Analysis of the available headache outcome data for 96 (78.6%) patients showed good headache outcomes in 51 patients and poor headache outcomes in 45 patients. The adjusted right hippocampus volume was positively associated with good headache outcome after controlling for age, sex, HADS score, and headache frequency (adjusted OR 4.72, p = 0.024). However, headache outcome was not associated with the left hippocampus volume or bilateral amygdala volumes.
Discussion
In the present study, hippocampus and amygdala volumes changed with headache frequency, and these volumes were correlated with frequency in specific ranges. In addition, the right hippocampus volume was positively associated with a good 2-year outcome of migraine.
A few studies have investigated the variation of hippocampus volume with migraine frequency. One longitudinal study demonstrated reduced hippocampus volume in patients with newly diagnosed migraine at 1-year follow-up, with a concomitant increase in headache frequency from 4.4 to 9.5 days/month (22), but whether the volume reduction was due to increased headache frequency or increased disease duration could not be determined. Another cross-sectional study by Maleki et al. showed that the bilateral hippocampi were larger in migraine patients with 1–2 headache days than those with 8–14 headache days/month, and the hippocampus volumes were negatively correlated to the estimate of total number of attacks (9). The present study demonstrated the volumetric variation of bilateral hippocampi in a full spectrum of headache frequencies, with the volumes peaked at 5–7 days/month on both sides and correlated with headache frequency in different directions within specific ranges. Of note, the disease durations did not differ across patient groups, and anxiety and depression were adjusted in data analysis. Thus, our study, for the first time, detailed the association of migraine frequency with hippocampus volume, without unwanted confounding from disease duration or psychiatric comorbidities. We did not reveal differences in hippocampus volumes between headache frequencies 1–2 and 8–14 days/month (p = 0.57 on the left side and p = 0.35 on the right side), as shown in Maleki’s study (9). The discrepancy may be attributed to adjustment of hippocampus volumes with psychiatric symptoms in our study, use of different imaging processing tools to normalize hippocampus volumes to TIV, and difference in the ratio of male vs. female patients (1:2.3 in Maleki’s and around 1:1.65 in ours).
We proposed the hippocampal volume changes with migraine frequency in this study are attributed to neuroplasticity induced by stress and pain. Adaptive increase and maladaptive decrease of hippocampus volume in response to stress have been demonstrated in both human and animal models (23,24); in addition, persistent pain can produce stress-like remodeling of hippocampus (25). Our findings imply that migraine attacks per se act as stressors that induce subsequent hippocampal volume changes. The increase of hippocampus volume in patients with 3–7 headache days/month suggests hippocampus adaptation to the increased physical and emotional demands in this frequency range, and can be viewed as a learning process of adaptation and adjustment to “migraine experiences” (24). This process may involve enhanced neurogenesis in the dentate gyrus and increased neural connections between hippocampus and the other neural structures (23,24). On the other hand, hippocampus volume decreased with increasing headache frequency in the range of 5–30 days/month, suggesting that frequent attacks of migraine (beyond a threshold) exceed the coping ability of the neural system and lead to a maladaptive volume reduction (26). Suppression of neurogenesis, retraction of dendrites, and loss of synapses in hippocampus have been demonstrated in animal models of chronic pain and chronic restraint stress (25,27), which may explain the reduced volumes in our patients with frequent migraines.
Another explanation of our findings is that changes in hippocampus volume reflect the overall perceived stress from any source. As stress is a well-known trigger of migraine headaches (28), it is reasonable to suggest the relationship of stress-related hippocampal volume change to headache frequency. The two hypotheses mentioned above are not contradictory. The hippocampus serves inhibitory feedback to the hypothalamic–pituitary–adrenal axis (HPA axis) that controls reactions to stress (3), and hippocampal neurogenesis has been demonstrated to buffer stress responses (29). A smaller hippocampus may reflect reduced neurogenesis in the dentate gyrus with insufficient feedback on the HPA axis, which leads to a delayed shutdown and an over-activation of stress responses that trigger further headaches. In fact, a smaller hippocampus has been shown to enhance vulnerability to stress, and is associated with increased occurrence of stress-related disorders and the persistent state of pain (8,30). In agreement, our study demonstrated a larger right hippocampus was associated with headache reductions in migraine. Taken together, escalating headache frequency and migraine chronification may result from a feedforward cascade of hippocampus volume reduction and stress maladaptation (5), implying a bidirectional relationship between limbic structural change and headache frequency [i.e., headache frequency affects the limbic structural volumes, and the volume per se also determines future clinical outcome (headache frequency) of migraine]. It is notable that interacting effects between hippocampus and other neural structures, such as the amygdala or prefrontal cortex (7), may also play roles in structural plasticity of hippocampus. Individual genetic heterogeneity is also associated with hippocampus volume but is beyond the scope of this study (31).
The functional dissociations between the bilateral amygdala may explain their different patterns in volume changes across headache frequency. The right amygdala has been reported to be predominately involved in negative emotions, fear conditioning, processing of prolonged nociceptive inputs, and development of sensitization (32,34), whereas the left amygdala is more involved in positive emotions (33). In addition, the amygdala plays a feedforward role in the regulation of the HPA axis (3), and an animal study reveals enhanced dendritic arborization and increased spine density of the amygdala in response to chronic immobilization stress (35), which may explain a trend of volume increase with increasing headache frequency of the right amygdala in our study. Valfre et al. once reported a smaller amygdala volume in chronic migraine than in episodic migraine (10); however, the amygdala volume did not differ between these two groups in our study (p = 0.42 on the left side and 0.21 on the right side, data not shown). It is notable that the mean age between the patients with episodic and chronic migraine was different in Valfre’s study but not in ours, and the impact of psychiatric comorbidities was additionally adjusted in our study. Taken together, the above inconsistent findings in amygdala volumes between sides and chronicity states of pain may reflect a complex role of amygdala in the affective components of pain, and the potential influence of comorbid psychiatric symptoms on the volumetric change of amygdala.
Some caveats should be considered before interpreting our findings. First, our study relied on cross-sectional data (amygdala and hippocampal volumes) rather than longitudinal data; therefore, the exact causal relationship between limbic volumes and headache frequency could not be determined. Second, although the analyses of amygdala and hippocampal volume changes to headache frequency were adjusted for the effects of age, sex, and psychiatric comorbidities, it did not adjust for other possible influencing factors, including stressors other than migraine attack, menstrual cycles in female subjects, medication overuse, and individual genetic heterogeneity. Third, we used headache days but not migraine days to categorize our migraine patients, therefore, non-migraine headache days may also be included. Fourth, usage of acute or preventive medications was not controlled at follow-ups in this study. Fifth, the results of the volumetric differences between specific headache frequency groups should be considered exploratory in the figures. Thus, the p-values were adjusted only by the number of tests we have done. Finally, whether sub-regions of the hippocampus and amygdala responded differently to stress and pain stimuli could not be revealed in this study due to methodological constraints.
To conclude, the hippocampus and amygdala change structurally with headache frequency in patients with migraine, and the right hippocampus volume is positively associated with a favorable prognosis of migraine. Future longitudinal studies are required to determine whether plastic change in these two structures may serve as a surrogate imaging measure or brain signature for migraine phenotypes, evolution, and prognosis.
Article highlights
Hippocampus and amygdala volumes change with headache frequency. The right hippocampus volume is associated with migraine outcome.
Footnotes
Declaration of conflicting interests
The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: H-Y Liu, K-H Chou, P-L Lee, D-M Niddam, K-L Lai, F-J Hsiao, C-P Lin, Y-Y Lin and W-T Chen report no disclosures. S-J Wang has served on the advisory boards of Allergan, and Eli Lilly Taiwan. He has received speaking honoraria from local companies (Taiwan branches) of Pfizer, Elli Lilly and GSK. He has received research grants from the Taiwan National Science Council, Taipei-Veterans General Hospital, and Taiwan Headache Society. J-L Fuh is a member of a scientific advisory board of Novartis, and has received research support from the Taiwan National Science Council and Taipei-Veterans General Hospital.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: grants from Ministry of Science and Technology (MOST 103-2628-B-075-001-MY3 to WT Chen, 101-2314-B-010-068-MY3 to YY Lin, MOST 104-2745-B-010-003, 103-2321-B-010-017, 102-2321-B-010-030, 100-2314-B-010-018-MY3, and 99-2314-B-075-036-MY3 to SJ Wang), Taipei-Veterans General Hospital (V104C-115, V105C-092, V105E9-005-MY2-1 and VGHUST105-G7-1-2 to WT Chen, VGHUST105-G7-1-1, V105C-127, V105E9-001-MY2-1, and VTA105-V1-1-1 to SJ Wang), NSC support for Center for Dynamical Biomarkers and Translational Medicine, National Central University, Taiwan (NSC 101-2911-I-008-001, 102-2911-I-008-001 to SJ Wang), Brain Research Center, National Yang-Ming University, a grant from Ministry of Education, Aim for the Top University Plan, and grants from the Ministry of Health and Welfare (MOHW104-TDU-B-211-113-003).
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
