Burrowes et al.

The Journal of Headache and Pain (2021) 22:127

[Link]
The Journal of Headache
and Pain

RESEARCH ARTICLE Open Access

Differences in gray matter volume in

episodic migraine patients with and
without prior diagnosis or clinical care: a
cross-sectional study
Shana A.B. Burrowes1,2,3,4,7*, Olga Goloubeva5, Michael L Keaser2,3, Jennifer A. Haythornthwaite6 and
David A. Seminowicz2,3

Abstract
Background: Migraine sufferers face difficulties getting appropriate care and treatment. Migraine is associated with
reduced gray matter volume (GMV) in several brain regions, which could be related to various clinical characteristics
of the disorder.
Objectives: To examine differences in GMV in migraine patients with and without prior clinical care for migraine
and examine differences in migraine clinical variables, psychosocial symptoms and their relationship with GMV.
Methods: We utilized the baseline MRI scan and psychosocial symptom questionnaires from a longitudinal
randomized controlled trial. Prior care of migraine was determined by diagnosis by a medical practitioner or
prescription of migraine specific medication.
Results: 117 patients were included in the study. Patients without prior care (n=23) had reduced GMV in the right
dorsal medial prefrontal cortex (dMPFC) relative to patients who had prior care (p=0.034, FWE corrected). Both
patient groups had reduced GMV compared to healthy controls (n=36). Patient groups did not differ in headache
clinical variables. Regardless of care status, increasing scores on the stress (Perceived Stress Score) and depression
questionnaires (Patient Health Questionnaire) were associated with increased GMV in the dMPFC.
Conclusions: Clinical care may impact GMV in migraine patients. Patients may need different treatment options to
address this baseline deficit.
Trial registration: NCT02133209.
Keywords: Gray matter volume, Treatment naïve, Prior care, Depression, Stress, Migraine, Headache, MRI

* Correspondence: shanab@[Link]
1
Section of Infectious Diseases, Department of Medicine, Boston University
School of Medicine, 02218 Boston, MA, USA
2
Department of Neural and Pain Sciences, School of Dentistry, University of
Maryland Baltimore, 21201 Baltimore, MD, USA
Full list of author information is available at the end of the article

© The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if
changes were made. The images or other third party material in this article are included in the article's Creative Commons
licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons
licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain
permission directly from the copyright holder. To view a copy of this licence, visit [Link]
The Creative Commons Public Domain Dedication waiver ([Link] applies to the
data made available in this article, unless otherwise stated in a credit line to the data.
Burrowes et al. The Journal of Headache and Pain (2021) 22:127 Page 2 of 9

Introduction in the bilateral dorsolateral prefrontal cortex (DLPFC)

Migraine often goes undiagnosed and migraineurs face associated with increased attack frequency and longer
difficulties in ascertaining appropriate care and treat- disease duration [12], providing further evidence that
ment options. [1, 2] Nationwide studies in the US have structural changes to these areas are associated with se-
shown that among persons with episodic migraine (EM), verity of disease.
only 45.5 % had received a medical consultation in the However, there has not been any investigation into the
preceding year, and of those 86.7 % received a diagnosis pathophysiological association between ineffective or
of migraine. Furthermore, of those migraineurs who are lack of treatment and migraine and it is clear that this is
in need of clinical care only one quarter successfully a gap that needs to be addressed. In this study we exam-
achieved the minimum appropriate care. [3] These treat- ined GMV differences between two groups of EM pa-
ment patterns are not unique to the US. The Eurolight tients, those who have received care for their migraine
project comprising 10 European countries, reported that and those who have not. We hypothesized that at base-
in population-based samples the proportion of patients line GMV will differ in specific pain and affective pro-
who had seen a doctor ranged between 9.5 and 18 % and cessing areas such as the DLPFC, insula and ACC based
an even smaller proportion (3.1-15 %) were prescribed on prior diagnosis/care of migraine at baseline. Explora-
migraine specific abortive medication (Triptans). [4]. tory analysis examined the relationship between psycho-
In the American Migraine Prevalence and Prevention social symptoms and GMV.
Study (AMPP) it was found that there were three main
steps needed to attain minimum care: appropriate med- Methods
ical consultation, accurate diagnosis and effective treat- Study design and population
ment. [3] Though barriers exist at each level, the Participants included in this cross-sectional analysis of
greatest occur at the level of seeking care where predic- GMV were enrolled in the MRI Outcomes of Mindfulness
tors for consultation are, having access to health insur- Meditation for Migraine clinical trial (NCT02133209).
ance, and headache related variables such as headache The Johns Hopkins School of Medicine and University of
related disability and pain intensity. [3]. Maryland Baltimore Institutional Review Boards approved
The need to overcome the aforementioned barriers is the study. All participants provided written consent. Par-
extremely important for several reasons, the most im- ticipants were enrolled from June 2014 to February 2017.
portant being to reduce the risk of migraine patients Participants were aged 18-65 and were not eligible if they
transitioning to more chronic and severe forms of the had a history of mindfulness meditation practice. All mi-
disease. Ineffective treatment of episodic migraine is just graine patients were episodic migraineurs (EM) as defined
one of the risk factors for new onset chronic migraine by the International Classification of Headache Disorders
(CM). [5] The prevailing theory is that as a patient expe- Criteria-II for migraine. [13] Migraine patients were pro-
riences more frequent headache attacks there is pro- spectively recruited from patient registries at headache
longed activation of neuronal networks that are involved clinics at John’s Hopkins and University of Maryland
in pain processing during attacks and through neuro- Medical Center and via print, electronic and radio media.
plastic mechanisms have lower thresholds for subse- Migraine patients completed 28-day headache diaries at
quent attacks. [5–7] It is also hypothesized that enrollment and were eligible if they reported 4-14 head-
ineffective treatment leads to longer exposures to pain ache days at baseline. Screening and intervention visits for
which increase the risk of CM. [5]. migraine patients were conducted at Johns Hopkins Bay-
Gray matter volume (GMV) in motor/premotor, pre- view Medical Center. The study also enrolled 39 healthy
frontal, cingulate, posterior parietal, and orbitofrontal pain free controls, matched to the migraine patients on
cortices have been negatively correlated with headache age (±5 years), sex, body-mass index (BMI) (±5), educa-
severity, frequency, duration, and pain intensity. [8–10] tion (college and no college) and race. The main outcomes
The right anterior cingulate cortex (ACC) and bilateral of the clinical trial have been published elsewhere. [14].
insula have been associated with headache frequency in Neuroimaging visits for participants occurred at the
episodic migraine patients. [8, 11] Chen et al. also re- University of Maryland Baltimore. At each visit partici-
ported decreased GMV in the left superior frontal gyrus pants underwent an MRI, as well as completed several
(SFG) in migraine patients (both EM and CM) when psychosocial and quality of life questionnaires. Patients
compared to tension type headache patients, suggesting provided headache diary data and psychosocial and qual-
that differences in the SFG are specific to headache type. ity of life questionnaires at three time points, but this
[11] Additionally greater pain intensity and disease dur- study only utilizes baseline data. Headache diaries were
ation have been associated with reduced GMV in the bi- completed electronically (online via a link sent through
lateral posterior and the left anterior insula. [12] email) and collected detailed information on headaches
Hubbard et al. further found reduced cortical thickness over a 28 day period. Information on the quality of the
Burrowes et al. The Journal of Headache and Pain (2021) 22:127 Page 3 of 9

headache including type of pain, headache duration and of GMV in both patient groups. [16] Using the cross
associated symptoms were collected. These analyses uti- sectional segmentation pipeline in CAT12, the baseline
lized diagnosis information, MRI, headache diary and structural T1-weighted images were spatially normalized
questionnaire data at baseline. Participants who were en- to Montreal Neurological Institute (MNI) space
rolled but did not complete MRI visits, have unusable (resampled to a voxel size of 1.5mm x 1.5mm x 1.5mm),
MRI data or missing data on prior clinical care and segmented into gray matter (GM), white matter (WM)
treatment for migraine were not included in this study. and cerebrospinal fluid (CSF). Scans were preprocessed
with an absolute threshold mask of 0.1. This threshold
Assessment of prior clinical care excluded voxels with less than 10 % probability of being
Migraine diagnosis or clinical care prior to enrollment in gray matter. Finally, images were smoothed with an
the study was established by detailed interview and 8mm Gaussian Kernel prior to analysis.
health history form (HHF) during the baseline screening
at Johns Hopkins. The HHF form collected information Explicit mask analyses
on current and past medical diagnoses and problems, To assess the difference in GMV between EM patients
prescription and over the counter medications including with and without prior clinical care we conducted expli-
doses and life style factors such as smoking, drinking cit mask and whole brain analyses. The explicit mask
and exercise. Patients completed a separate screen about comprised the bilateral SFG, DLPFC, insula and ACC.
history of their migraine including but not limited to The inclusion of these regions in the explicit mask was
prior diagnoses and treatment information for migraine based on their relationship with migraine clinical charac-
headaches. Based on information from their HHF and teristics and were included in our preregistered analysis
migraine screening session, patients were categorized plan ([Link]
into two groups (prior care and no prior care for mi- All regions included in the analysis were made from the
graine) before study enrollment. Categorization was Atlas of Intrinsic Connectivity of Homotopic Areas. [17]
achieved using patient report of diagnosis from a med- Given no a-priori hypotheses about laterality all regions
ical practitioner (specialist or general practitioner) or were made for both the left and right sides of the brain.
prescription of migraine specific medication for which The regions were combined using the “-add” function in
the intended use was migraine prevention or treatment. FSLMATHS and displayed for inspection on a standard
Patients with either a diagnosis or prescribed treatment brain in MNI space (Additional File 1).
were categorized as having prior care. Patients who did
not have a history of migraine diagnosis or prescribed SPM12 group analysis and GMV extraction
migraine medication were classified as not having any In SPM12, two sample t-test adjusted for age and total
prior care for migraine. However, all patients were offi- intracranial volume (TIV) was used to assess differences
cially diagnosed with migraine at enrollment by a neur- between the two patients groups. A cluster forming
ologist at Johns Hopkins using ICHD-II criteria. threshold of p<0.001 was used and significant clusters
within the explicit mask as well significant clusters in
MRI procedures the whole brain analyses were extracted using Marsbar
All baseline MRI scans were performed from September ([Link] Secondary analysis ex-
2014 through March 2017 and completed on a Siemens tracted GMV for healthy controls from significant clus-
Tim-Trio 3T MRI scanner using a 32-channel head coil. ters (from the analysis between patients with and
This study utilized scan data from the T1 MPRAGE. Ac- without prior clinical care) to determine if patients dif-
quisition parameters for the T1 MPRAGE were as fol- fered from controls in these regions. Values were con-
lows: repetition time (TR) 2300ms, echo time (TE) verted to mm3 using the following equation: cluster size
2.98ms, slice thickness 1mm, field of view (FOV) x voxel size x beta value. The beta value extracted repre-
256mm, flip angle 9°, voxel sixe (1 × 1 × 1mm). sents the proportion of the voxel attributed to gray
matter.
MRI preprocessing and analysis methods
We used voxel based morphometry (VBM) to assess dif- Analysis of clinical variables
ferences in GMV in episodic migraine patients with and Linear regression models assessed the relationship be-
without prior clinical care/treatment for migraine at tween GMV differences in those with and without prior
baseline. [15] All images were realigned to the anterior- clinical care and clinical characteristics of migraine as
posterior commissure in Statistical Parametric Mapping independent predictors. Of interest were headache clin-
(SPM12) ([Link] prior to pre- ical variables such as severity (rated as mild, moderate or
processing. The computational anatomy (CAT12.1, severe), headache pain intensity (rated on a scale of 0-
r1250) toolbox within SPM12 was used to assess VBM 10) and headache frequency which were collected via
Burrowes et al. The Journal of Headache and Pain (2021) 22:127 Page 4 of 9

electronic headache diaries over a 28-day period. We with BMI in the normal range. Patient groups also
also collected information on psychosocial disorders showed no difference in headache frequency, intensity of
(anxiety, stress and depression) and examined the rela- headache pain or severity of headaches. In terms of
tionship between GMV and psychosocial factors as inde- medication use, 24 % of patients used preventatives, 55 %
pendent predictors. Anxiety was measured by the of patients with prior care used abortives and 67 % used
Generalized Anxiety Disorder-7 (GAD-7), depression- over the counter (OTC) pain treatments. Patients in the
Patient Healthy Questionnaire-9 (PHQ9) and stress- prior care group often used more than one kind of
Perceived Stress Scale (PSS). All scores were modeled as medication (see details in footnote of Table 1). 60 % of
both raw values and the following categories;GAD-7 was patients without prior care reported using OTC pain
categorized as 0-4 (no anxiety), 5-9 (mild), 10-14 (mod- treatments.
erate), 15≤ (severe). Similarly PHQ-9 was modeled cat- There were 39 healthy controls enrolled in the study
egorically in bins of 5 (0-4, 5-9, 10-14, 15-19 and ≥20) to but three did not have baseline MRI data, resulting in a
represent no depression, mild, moderate, moderately se- final sample of 36 controls. Similar to patients, controls
vere and severe. PSS was modeled where 0-13 is average, were primarily female, white and college educated. Con-
14-26 is moderate stress and high stress is 27-40. Other trols were of comparable age of patients with prior care
covariates assessed as potential confounders for these for migraine (38 years) but older than those without
models were duration of disease, sex, BMI, race, age, prior care (31 years) (Table 1).
education and employment status. Employment status
was modeled as it was reported by each individual and MRI- GMV differences between episodic migraine patients
as a dichotomous variable (full time vs. not-full time). with and without prior clinical care
Final models were adjusted for age as it was correlated In the explicit mask analyses no significant clusters dif-
with disease duration, was found to be a better predictor fered between the two groups. However whole brain
and improved the model fit. The Chi-Square, Fishers analyses showed that patients without prior clinical care
Exact, t and Wilcoxon tests were used to assess differ- had lower GMV in one cluster located in the right dorsal
ences in clinical and confounding variables. Secondary medial prefrontal cortex (R DMPFC) (p=0.034, FWE
analyses examined demographic and GMV differences cluster-level corrected) compared to those with prior
between patients and healthy controls utilizing Fishers clinical care. The average volume in patients with prior
Exact, t-tests, Wilcoxon and ANOVA tests. All statistical clinical care was 471.68mm3 compared to 435.92mm3 in
analyses were conducted using SAS (v.9.4, SAS Institute those without. Healthy controls had an average GMV of
Inc. Cary, NC). Testing was two-sided and done at the 599±32 mm3 in the R DMPFC cluster. This was signifi-
0.05 level of significance. cantly larger than both patient groups (<0.0001). The
cluster comprised 418 voxels and extends to include
Sample size and power Brodmann Area 10 (BA10), portions of the right middle
Using a two sample t-test with a two sided 0.05 level of frontal gyrus (MFG), medial superior frontal gyrus and
significance, sample size of 120 patients and a group ra- medial frontal gyrus (Fig. 1A and B). Table 2 details the
tio of 3:1 (patients with and without prior clinical care at size and location of the cluster as well as peak voxels
baseline), provided above 80 % power to detect a mean within the cluster.
difference ranging 8-15mm3 between migraine patients
with and without prior clinical care at baseline.
Difference in GMV in EM patients with and without prior
Results clinical care and the relationship with clinical
Description of demographic and clinical profiles characteristics
There were 120 migraine patients with MRI and diagno- We assessed disease duration, headache frequency, se-
sis data at baseline. Two patients were excluded due to verity (mild, moderate, severe), headache pain intensity,
abnormal brain morphology and one was excluded due and specific psychosocial disorders associated with mi-
to poor data quality. The final sample included 94 (80 %) graine (anxiety, stress and depression). There were no
migraine patients classified as having prior care for mi- significant associations with headache clinical character-
graine and 23 (20 %) who had not. Migraine patients istics. However, in the exploration of psychosocial mea-
with prior clinical care were significantly older than sures we observed a significant association between
those without (mean age 38 vs. 31) and reported longer GMV and depression and stress. Table 3 shows the re-
median duration of disease (18 vs. 10 years) (Table 1). sults from the linear regression models of GMV in the
However, across other clinical and sociodemographic right DMPFC with patient group (prior care vs. no prior
metrics both groups were comparable, being predomin- care) and depression/stress modeled as predictors. Both
antly female, white, college educated, employed full time, the depression and stress models show that those
Burrowes et al. The Journal of Headache and Pain (2021) 22:127 Page 5 of 9

Table 1 Sociodemographic and clinical factors in EM patients with and without prior clinical care for migraine and healthy controls
Variable Prior Medical Care /94 (80 %) No Prior Medical Care/23 (20 %) p value1,* Healthy Controls/36 p value2,*
Sex

Male 9 (10) 5 (22) 0.15 4 (11) 0.23

Female 85 (90) 18 (78) 32 (89)

Age/mean±SD 38±12 31± 8 0.001 38±13 0.02

Race/n (%)3

Black 17 (18) 6 (26) 0.14 9 (25) 0.35

White 67 (72) 12 (52) 25 (69)

Other 9 (10) 5 (22) 2 (6)

Education/ n (%)
High School/Technical School 6 (6) 1 (4) 0.86 2 (6) 0.97

College 56 (60) 15 (65) 21 (58)

Graduate 32 (34) 7 (30) 13 (36)

Employment/n (%)

Full-time 62 (66) 10 (43) 0.16 N/A

Part-time 13 (14) 5 (22)

Student 12 (13) 5 (22)

Homemaker 2 (2) 1 (4)

Retired 3 (3) 0
Unemployed 2 (2) 2 (9)

BMI/mean±SD 27±6 27±7 0.91 26±5 0.45

Headache Days
4
Headache frequency/28days median (range) 8 (3-17) 9 (3-16) 0.96 N/A
5
Raw Headache Days/ median (range) 8 (3-17) 8 (3-16) 0.92 N/A

Headache intensity/0-10 5±1.54 4±1.47 0.20 N/A

Headache severity

Mild 25 (27) 11 (48) 0.14 N/A

Moderate 64 (68) 11 (48)

Severe 5 (5) 1 (4)

Duration of disease/years Median (range) 18 (1.5-50) 10 (1-24) <0.001

6
Use of migraine/pain medications
Preventatives 7 23 (24) N/A

Abortives 52 (55) N/A

Over the Counter 63 (67) 14 (61) N/A

None 9 (10) 9 (39)
3 8
Use of non-migraine medication 10 (12) 3 (14)

PSS/ Median (range) 11 (0-28) 12 (2-32) 0.69

GAD-7/Median (range) 1 (0-16) 1 (0-16) 0.64

PHQ-9/Median (range) 2 (0-20) 3 (0-10) 0.42

Additional Chronic Pain 3 9

Yes 27 (30) 6 (30) 1.0

No 63 (70) 14 (70)
1
p value for the comparison of prior care and no prior care
2
p value for the comparison across three groups: prior care, no prior care, healthy controls
3
Variables with missing observations
4
Headache days per 28 days are adjusted headache days
5
Raw headache days do not account for incomplete diaries
6
Numbers for the prior care group add up to more than 94 because patients took more than one type of medication. Of the 52 patients prescribed abortives: 12 patients took abortives only,
23 took abortives and over the counter (OTC) pain killers, 8 took abortives and preventatives and 9 patients took abortives, preventatives and OTC. Of the remaining patients, 27 took OTC
only, 2 patients used preventatives only, 4 patients used preventatives and OTC and 9 patients stated they had been diagnosed with migraine but did not list any current medications
7
Patients reported using the following preventative medications: Amitriptyline, Atenolol, Candesartan, Inderal, Lamictal XR, Neurontin, Nortriptyline, Norvasc, Propranolol ER, Topamax,
Venlafaxine, Verapamil
8
There were five patients taking medication for anxiety and 2 patients taking medication for IBS, depression, and hypertension. The followings diseases all had only one patient taking
medication: ADD, bipolar, heart disease, Crohn’s disease, hypothyroid, ulcers and pituitary adenoma. Data was missing on 11 patients
9
The following pain chronic pain conditions were included: fibromyalgia, osteoarthritis, rheumatoid arthritis, chronic pelvic pain, low back pain, neck pain, temporomandibular joint disorder
*T test, Chi Square, Wilcoxon, Kruskal Wallis p value
EM Episodic Migraine, PSS Perceived Stress Scale, GAD Generalized Anxiety Disorder-7, PHQ-9 Patient Healthy Questionnaire-9
Burrowes et al. The Journal of Headache and Pain (2021) 22:127 Page 6 of 9

Fig. 1 Cluster results showing region of reduced GMV in the right dorsal medial prefrontal cortex. The results are from a t-test of EM patients
without prior clinical care compared to those with prior clinical care. A displays the cluster on the average brain created from all 117 patients.
Cluster forming threshold of p=0.001, k=418, pFWE<0.05 at cluster-level. B displays the cluster on the SPM12 cortical surface. L: left; R: right

patients without prior clinical care have less GMV (ap- patients from other headache types as well as EM from
proximately 53mm3) compared to those with clinical CM. [11].
care. Consistent in both models GMV increased with in- Conversely, a recent study in an Italian cohort found
creasing scores on the depression and stress question- that regions in the frontal and temporal lobes including
naire. With every point increase on the depression the R SFG, left middle temporal gyrus and right inferior
PHQ-9 questionnaire there was a 4.60mm3 increase in frontal gyrus of migraine patients had increased GMV
GMV. For the stress model there was a 2.15mm3 increase compared to controls and this was positively correlated
with every point increase on the stress PSS question- with headache symptoms. [19] However in those studies,
naire. There was no interaction between group and de- which reported reductions in GMV there were negative
pression or stress. (Additional File 2). correlations with headache severity, duration and head-
ache frequency. [8–10, 18] This negative association
Discussion with headache clinical variables was not present in this
Recently our knowledge of migraine pathophysiology has current study, as we found no significant relationship be-
improved tremendously, but gaps remain of how treat- tween GMV and headache symptomology. However,
ment differences are reflected in brain changes. We there was a positive association with stress and depres-
show that though migraine patients regardless of prior sion. Our results can be examined in the context of
clinical care, present with similar clinical and sociode- these psychosocial factors and clinical findings in other
mographic profiles there exists an underlying difference pain patients. It should be noted that this relationship is
in brain pathology. To our knowledge this is the first likely driven by small numbers of patients with high psy-
study to highlight GM differences between migraine pa- chosocial scores, thus the association we observe is hard
tients based on whether they have had prior care. Mi- to interpret. Plausible explanations are that patients with
graine patients who had not received prior care had less these scores may be more likely to seek clinical care, or
GMV in the right DMPFC compared to those with prior that the interaction of headaches and psychosocial co-
care. Both groups had less GMV compared to healthy morbidities affects the brain differently than each dis-
controls. The cluster included portions of the right order separately.
MFG, medial frontal gyrus, and medial SFG, all regions The medial prefrontal cortex (mPFC) has been impli-
which play critical roles in migraine pathophysiology cated in many pain disorders including IBS, fibromyal-
and clinical characteristics. [8, 10] The prefrontal cortex gia, back pain and migraine. [20–23] Chen et al.
has also been shown to differ between migraine patients reported smaller frontal regions in chronic compared to
and healthy controls in several studies. Specifically, the EM and a negative correlation between headache fre-
left MFG has shown reduced GMV volume in migraine quency and frontal pole volume. [24] In cluster headache
patients. [9, 10, 18] Further work to differentiate be- patients, GMV in the frontal pole was lower compared
tween subtypes of migraine patients, have used several to those with migraine. [25] Though these works indi-
approaches and localized the bilateral MFG and other cate that the frontopolar cortex (FPC) is important in
regions in the pre-frontal cortex to classify migraine migraine pathology, there have not been any studies in
Burrowes et al. The Journal of Headache and Pain (2021) 22:127 Page 7 of 9

Table 2 Decreased GMV from whole brain analyses comparing EM patients with and without prior clinical care
Brain Region Peak Voxel MNI Coordinates (x, y, z) Cluster Size p value
Right Dorsal Medial Prefrontal Cortex 18, 51, 20 418 0.03
30, 57, 30
16, 45, 27
*p value from SPM12 analyses. Analyses adjusted for age and total intracranial volume
MNI Montreal neurological institute, GMV Grey Matter Volume, EM Episodic Migraine

migraine patients examining how clinical care may im- significant main effects of migraine and depression in
pact brain structure. Research in other chronic pain con- the left mPFC where there was increased ALFF associ-
ditions has shown that with successful treatment there is ated with both disorders. It was concluded that these
recovery of abnormal brain structure in the prefrontal findings may indicate a common therapeutic target for
cortex. In chronic back pain patients, thinning in DLPF migraineurs with co-morbid depression. [27].
C in patients compared to controls was normalized after In light of previous studies our results add to a grow-
treatment and was correlated with reductions in pain ing body of work which highlight the mPFC as not only
and physical disability. [21] Though we found that an important role in chronic pain, but an important re-
healthy controls had significantly larger GMV compared gion in the treatment of pain patients. We show that this
to both patient groups, the fact that those with prior region may also adversely affected in patients who do
care had higher GMV than those without would suggest not receive care for their migraine headaches. However,
that treatment for migraine allowed them to recover we do have some limitations. In our sample almost 20 %
some GMV. One RCT in migraine found that baseline of patients had never received prior clinical care for their
mPFC volume could predict the hypoalgesia response migraine. Compared to population based studies where
with sham acupuncture after 8 weeks (mediated by de- less than half of migraine patients often report receiving
creased anxiety). [20] Fettes et al. showed abnormal medical consultation for their condition [3, 4] our sam-
functional connectivity in the medial and lateral FPC in ple of patients had a much higher proportion of diag-
treatment-nonresponsive major depressive disorder pa- nosed patients (80 %). This is likely due to the fact that a
tients which was correlated with symptom severity. [26]. major component of the recruitment strategy was tar-
This gives some insight into the role that the mPFC geted to headache clinics. This means that our sample
plays not only in migraine but in the way it interacts was not representative of the wider migraine community
with psychological disorders in migraine and subsequent where diagnosis of migraine is less common. It is pos-
migraine improvement. The interaction between mi- sible that this difference in sample size could bias our re-
graine and co-morbid psychological disorders is of grow- sults, either falsely finding a difference where there is
ing interest due to the prevalence of these disorders in none, or occluding additional differences which may be
the migraine population. A recent imaging study looked important to address in those patients who have not re-
at differences in intrinsic activity measured by amplitude ceived care. However we did conduct power analyses
of low-frequency fluctuation (ALFF) in migraineurs with which showed that even with a 3:1 ratio of patients be-
and without co-morbid depression. Researchers found tween groups, we were powered above 80 % to detect
differences in GMV between patients with and without
Table 3 Association between GMV in EM Patients and prior care. Our patients also had comparable sociodemo-
psychosocial factors graphic and clinical profiles, regardless of prior care sta-
Variable Beta Value (mm3) Standard Error p value tus which does address some concerns regarding
Depression differences in sample size. While most epidemiological
Prior Care (REF) 517 20 literature has found difference in socioeconomic and
No Prior Care -53 13 <0.001 headache severity status between those who seek care
Depression 4.60 1.72 0.009
and those who do not, this was not present in our study.
Given that we were unable to directly assess socioeco-
Stress
nomic status, and there was no association between
Prior Care (REF) 505.57 21.62 headache factors and GMV, there may be some unmeas-
No Prior Care -52.68 13.45 <0.001 ured confounding in the sample. The demographic simi-
Stress 2.15 0.78 0.007 larity between groups was likely because our patients
* Beta values represent the grey matter volume in the right dorsal medial were recruited for a clinical trial and will not be repre-
prefrontal cortex in mm3. All models are adjusted for age; it was correlated sentative of the general population of migraine patients.
with disease duration and was a better predictor in the model; REF reference
level for comparison, DMPFC dorsal medial prefrontal cortex, EM episodic
Patients were classified into groups based on patient re-
migraine, GMV Gray Matter Volume port of migraine diagnosis and migraine medication.
Burrowes et al. The Journal of Headache and Pain (2021) 22:127 Page 8 of 9

One drawback to this approach was the difficulty in col- Abbreviations

lecting information on the duration of therapy in pa- ACC: Anterior Cingulate Cortex; AMPP: American Migraine Prevalence and
Prevention Study; BMI: Body Mass Index; CAT12: Computational Anatomy
tients. Given that the data was based on patient recall Toolbox; CM: Chronic Migraine; CSF: Cerebrospinal Fluid; DLPFC: Dorsolateral
and not review of a medical record, patients provided in- prefrontal cortex; DMPFC: Dorsal Medial Prefrontal Cortex; EM: Episodic
formation on start dates in a manner that was not use- Migraine; FOV: Field of View; FPC: Frontopolar Cortex; FWE: Family wise error;
GM: Gray Matter; GMV: Gray Matter Volume; GAD-7: Generalized Anxiety
able for analysis or to estimate an average duration for Disorder-7; HHF: Health History Form; MFG: Middle Frontal Gyrus;
the sample. Since we found differences between patients MNI: Montreal Neurological Institute; MRI: magnetic resonance imaging;
based on prior care (including medication use) it is pos- OTC: Over The Counter; PHQ-9: Patient Healthy Questionnaire-9;
PSS: Perceived Stress Scale; SFG: Superior Frontal Gyrus; SPM12: Statistical
sible that duration of treatment is an important factor in Parametric Mapping; TE: Echo Time; TIV: Total Intracranial Volume;
our observed findings. Because we are unable to adjust TR: Repetition Time; VBM: Voxel Based Morphometry; WM: White Matter
for this in analysis we cannot elucidate differences be-
tween those who recently started therapy and those who Supplementary information
had been on treatment for a long time. Further, we can- The online version contains supplementary material available at [Link]
org/10.1186/s10194-021-01340-5.
not determine how they differ from those without prior
care. Finally, though preventative and abortive treat-
Additional file 1: Supplemental Fig. 1. Explicit mask used for region
ments work very differently and play different roles, of interest analyses overlaid on average brain of all participants. Brain
there was substantial overlap between medication use in image showing the regions of the brain included in explicit mask
analyses. Mask includes the bilateral SFG, DLPFC, insula, and cingulate
patients with prior care and thus differences between
cortex.
these two types of treatments could not be elucidated.
Additional file 2 Supplemental Fig. 2. Scatter plot of right dorsal
Additionally there could be some occlusion of our re- medial prefrontal gyrus GMV with stress (PSS) score (a) and depression
sults given patients in both groups used over the counter (PHQ-9) score (b) in EM patients with (blue) and without (red) prior
clinical care. Plot of the relationship between GMV in the right DMPFC
(OTC) pain medication for the treatment of acute mi-
and psychosocial factors to show how GMV varies with scores.
graine. However use of OTC between groups was similar
(67 % vs. 61 %) and thus the use of OTC in the no prior Acknowledgements
care group likely did not have an effect on the results. Not applicable.
We do however present a study with several strengths.
Though the proportion of undiagnosed patients was Authors’ contributions
SAB collected and analyzed all MRI and questionnaire data and drafted the
smaller than expected, we observed a much larger differ- manuscript. OG provided statistical guidance and reviewed and edited the
ence in GMV between patients than anticipated. Power manuscript. MLK collected MRI data provided MRI statistical guidance and
analyses were based on expected differences of 8- reviewed and edited the manuscript. JAH acquired funding, reviewed and
edited manuscript. DAS acquired funding, reviewed and edited manuscript.
15mm3 and we report a difference of almost 36mm3 be- All authors read and approved the final manuscript.
tween patient groups. This is also the first study which ex-
clusively compares GMV differences between migraine Funding
The authors disclosed receipt of the following financial support for the
patients, based on prior clinical care and though it has research, authorship, and/or publication of this article: This work was
some limitations, it provides a foundation upon which fur- supported by NCCIH/NIH R01 AT007176 to DAS.
ther research can be built. Taken together, research in sev-
Availability of data and materials
eral patient populations, points to the common theme that The datasets used and/or analysed during the current study are available
abnormal structure and function in several regions in the from the corresponding author on reasonable request.
prefrontal cortex (DLPFC, mPFC) may both predict treat-
ment success and be altered with successful treatment. Declarations
Ethics approval and consent to participate
Conclusions The Johns Hopkins School of Medicine and University of Maryland Baltimore
Institutional Review Boards approved the study. All participants provided
Our results provide further evidence that EM patients written consent.
who are not receiving care may have an added burden
associated with this lack of care even though clinically Consent for publication
they may present the same. Given these findings it is Not applicable.

clear that efforts to improve access to migraine treat- Competing interests

ment need to be expanded. Furthermore, these results The authors declare that they have no competing interests.
indicate that some patients may need different treatment
Author details
options to address this baseline deficit. Future studies 1
Section of Infectious Diseases, Department of Medicine, Boston University
should aim at following treatment naïve patients to de- School of Medicine, 02218 Boston, MA, USA. 2Department of Neural and Pain
termine if they take longer to regain GMV over time Sciences, School of Dentistry, University of Maryland Baltimore, 21201
Baltimore, MD, USA. 3Center to Advance Chronic Pain Research, University of
and if they require either different or more intense ther- Maryland Baltimore, 21201 Baltimore, MD, USA. 4Department of
apy to reach treatment success. Epidemiology and Public Health, School of Medicine, University of Maryland
Burrowes et al. The Journal of Headache and Pain (2021) 22:127 Page 9 of 9

Baltimore, 21201 Baltimore, MD, USA. 5Greenebaum Comprehensive Cancer 19. Messina R, Rocca MA, Colombo B et al (2018) Gray matter volume
Center, University of Maryland, University of Maryland Baltimore, 21201 modifications in migraine: A cross-sectional and longitudinal study.
Baltimore, MD, USA. 6Department of Psychiatry and Behavioral Sciences, Neurology. doi:[Link] [pii]
Johns Hopkins University School of Medicine, Baltimore, MD, USA. 7Boston 20. Liu J, Lan L, Li G et al (2013) Migraine-related gray matter and white matter
University School of Medicine, 801 Massachusetts Avenue Room 2004, MA changes at a 1-year follow-up evaluation. J Pain 14(12):1703–1708 doi: 10.
02118 Boston, USA. 1016/[Link].2013.08.013 [doi]
21. Seminowicz DA, Wideman TH, Naso L et al (2011) Effective treatment of
Received: 23 August 2021 Accepted: 6 October 2021 chronic low back pain in humans reverses abnormal brain anatomy and
function. J Neurosci 31(20):7540–7550. doi:[Link]
JNEUROSCI.5280-10.2011 [doi]
22. Apkarian AV, Hashmi JA, Baliki MN (2011) Pain and the brain: Specificity and
References plasticity of the brain in clinical chronic pain. Pain 152(3 Suppl):49 doi: 10.
1. Wu J, Hughes MD, Hudson MF, Wagner PJ (2012) Antimigraine medication 1016/[Link].2010.11.010 [doi]
use and associated health care costs in employed patients. J Headache Pain 23. Seminowicz DA, Labus JS, Bueller JA et al (2010) Regional gray matter
13(2):121–127. doi:[Link] [doi] density changes in brains of patients with irritable bowel syndrome.
2. Lipton RB, Scher AI, Kolodner K, Liberman J, Steiner TJ, Stewart WF (2002) Gastroenterology 139(1):48–57.e2. doi:[Link]
Migraine in the united states: Epidemiology and patterns of health care use. 03.049 [doi]
Neurology 58(6):885–894 24. Chen XY, Chen ZY, Dong Z, Liu MQ, Yu SY (2020) Regional volume changes
3. Lipton RB, Serrano D, Holland S, Fanning KM, Reed ML, Buse DC (2013) of the brain in migraine chronification. Neural Regen Res 15(9):1701–1708.
Barriers to the diagnosis and treatment of migraine: Effects of sex, income, doi:[Link] [doi]
and headache features. Headache 53(1):81–92. doi:[Link] 25. Giorgio A, Lupi C, Zhang J et al (2020) Changes in grey matter volume and
526-4610.2012.02265.x [doi] functional connectivity in cluster headache versus migraine. Brain Imaging
4. Katsarava Z, Mania M, Lampl C, Herberhold J, Steiner TJ (2018) Poor medical Behav 14(2):496–504. doi:[Link] [doi]
care for people with migraine in europe - evidence from the eurolight 26. Fettes PW, Moayedi M, Dunlop K et al (2018) Abnormal functional
study. J Headache Pain 19(1):10–11. doi:[Link] connectivity of frontopolar subregions in treatment-nonresponsive major
0839-1 [doi] depressive disorder. Biol Psychiatry Cogn Neurosci Neuroimaging 3(4):337–
5. Lipton RB, Fanning KM, Serrano D, Reed ML, Cady R, Buse DC (2015) 347. doi: S2451-9022(17)30233-1 [pii]
Ineffective acute treatment of episodic migraine is associated with new- 27. Ma M, Zhang J, Chen N, Guo J, Zhang Y, He L (2018) Exploration of intrinsic
onset chronic migraine. Neurology 84(7):688–695. doi:[Link] brain activity in migraine with and without comorbid depression. J Headache
0.1212/WNL.0000000000001256 [doi] Pain 19(1):48–49. doi:[Link] [doi]
6. Burstein R, Cutrer MF, Yarnitsky D (2000) The development of cutaneous
allodynia during a migraine attack clinical evidence for the sequential Publisher’s Note
recruitment of spinal and supraspinal nociceptive neurons in migraine. Brain Springer Nature remains neutral with regard to jurisdictional claims in
123:1703–1709 (Pt 8 (Pt 8) published maps and institutional affiliations.
7. Bernstein C, Burstein R (2012) Sensitization of the trigeminovascular
pathway: Perspective and implications to migraine pathophysiology. J Clin
Neurol 8(2):89–99. doi:[Link] [doi]
8. Jia Z, Yu S (2017) Grey matter alterations in migraine: A systematic review
and meta-analysis. Neuroimage Clin 14:130–140. doi:[Link]
[Link].2017.01.019 [doi]
9. Kim JH, Suh SI, Seol HY et al (2008) Regional grey matter changes in
patients with migraine: A voxel-based morphometry study. Cephalalgia
28(6):598–604. doi:[Link] [doi]
10. Hu W, Guo J, Chen N, Guo J, He L (2015) A meta-analysis of voxel-based
morphometric studies on migraine. Int J Clin Exp Med 8(3):4311–4319
11. Chen WT, Chou KH, Lee PL et al (2018) Comparison of gray matter volume
between migraine and “strict-criteria” tension-type headache. J Headache
Pain 19(1):4–6. doi:[Link] [doi]
12. Hubbard CS, Khan SA, Keaser ML, Mathur VA, Goyal M, Seminowicz DA.
Altered brain structure and function correlate with disease severity and pain
catastrophizing in migraine patients. Eneuro. 2014;1(1):e20.14. doi: https://
[Link]/10.1523/ENEURO.0006-14.2014 [doi]
13. Headache Classification Subcommittee of the International Headache
Society (2004) The international classification of headache disorders: 2nd
edition. Cephalalgia 24(Suppl 1):9–160
14. Seminowicz DA, Burrowes SAB, Kearson A et al (2020) Enhanced
mindfulness-based stress reduction in episodic migraine: A randomized
clinical trial with magnetic resonance imaging outcomes. Pain 161(8):1837–
1846. doi:[Link] [doi]
15. Ashburner J, Friston KJ (2000) Voxel-based morphometry--the methods.
Neuroimage 11(6 Pt 1):805–821. doi:[Link] [doi]
16. Gaser C, Kurth F (Updated 2018) Computational anatomy toolbox-12
manual. [Link] Accessed
04/07/, 2018
17. Joliot M, Jobard G, Naveau M et al (2015) AICHA: An atlas of intrinsic
connectivity of homotopic areas. J Neurosci Methods 254:46–59. doi:https://
[Link]/10.1016/[Link].2015.07.013 [doi]
18. Dai Z, Zhong J, Xiao P et al (2015) Gray matter correlates of migraine and
gender effect: A meta-analysis of voxel-based morphometry studies.
Neuroscience 299:88–96. doi:[Link]
066 [doi]