Brainstem Dysfunction in Chronic Migraine as Evidenced by Neurophysiological and Positron Emission Tomography Studies*


  • Sheena K. Aurora MD,

  • Patricia M. Barrodale RN,

  • Reda L. Tipton,

  • Ani Khodavirdi PhD

  • From the Swedish Headache Center, Seattle, WA, USA (Drs. Aurora, Barrodale, and Ms. Tipton); Neurosicence Division, Advanced Bionics, Valencia, CA, USA (Dr. Khodavirdi).

  • *

    This article is the winner of the 2007 Harold G. Wolff Award.

Address all correspondence to Dr. Sheena Aurora, Swedish Headache Center, 1101 Madison, 200, Seattle, WA 98104, USA.


Background.—The pathophysiology of chronic migraine (CM) is not fully understood. We aimed to examine transcranial magnetic stimulation (TMS) indices of cortical excitability in patients with CM and also performed PET studies to ascertain if there were any areas of activation and inhibition for possible correlation.

Methods.—Excitability of the cortex was assessed by a reliable parameter of magnetic suppression of perceptual accuracy (MSPA) profiles using transcranial magnetic stimulation in 25 patients with CM. Of these 10 patients were also studied with (18F-FDG PET) scans.

Results.—MSPA demonstrated decreased inhibition in CM compared to normal controls and episodic migraine. The percentage of letters reported correct at 100 ms was 84.37 for CM compared to 19.14 for normal controls and 57.41 for episodic migraine. The PET evaluation in 10 subjects demonstrated increased cerebral metabolism in areas of in the brainstem compared to the global flow. There were also decreased areas of cerebral metabolism in the medial frontal and parietal as well as the somatosensory cortex.

Conclusions.—Patients with CM appear to be characterized by reduced visual suppression correlating with high cortical excitability. In a cohort of these subjects there was brainstem activation and inhibition in certain areas of the cortex suggesting a potential dysfunction in the inhibitory pathways.

Headache on a daily or a near daily basis occurs in about 5% of the general population.1 Chronic daily headache (CDH) was2 a descriptive term used for a heterogeneous group of conditions, which includes headache on a daily or almost daily basis as a common feature. The newly accepted International Headache Society classification of daily headache initially proposed by Silberstein and colleagues3 has the advantage of differentiating the working criteria for clinically distinct phenomena: chronic tension type headache (TTH); new daily persistent headache (NDPH); and hemicrania continua and chronic migraine (CM).4 Within this classification scheme problems with definition and classification of CM with relevance to clinical practice have been identified and recent studies have proposed a new definition for the more prevalent form of CM.5

In most patients CM evolves from episodic migraine (EM). The underlying mechanisms of CM are not well known, but medication overuse is implicit in the popular term “transformed migraine.”6 Routine clinical imaging for typical CM is usually normal, with no evidence of structural abnormality in the presence of a normal neurological examination.7,8 There is, however, an increasing body of research evidence on the involvement of nociceptive pathways in CDH and migraine. The first reports came from Raskin and colleagues,9 who observed a migraine-like headache developed in patients with electrode implantation in the periaqueductal gray (PAG). Recently the rostral brainstem has been identified as being specifically involved in migraine.10–13 The influence of the PAG on the trigeminovascular nociception has also been studied using neurophysiological studies. Using a cat model, Knight and Goadsby14 stimulated the venterolateral part of the PAG and demonstrated a reduction in evoked trigeminal neuronal activity inferring that the PAG was involved in inhibition of facial pain. Consistent with this animal work, we have previously demonstrated changes in the same anatomical regions of humans with CM. In patients with CDH, we noted a positive correlation between the duration of illness and the increase in MRI index of tissue iron levels in PAG.15 With this study we aimed to study the metabolism of the brain with particular attention to the brainstem in CM using positron emission testing (PET). Neuroimaging with positron emission tomography has shed light on the genesis of 2 of the most important headache syndromes, documenting activation in the midbrain and pons in migraine16 and in the hypothalamic gray in cluster headache (CH).17,18

We also aimed to study the neurophysiological effects of the brainstem activation on the cortical pathways. Although progress is clearly being made in understanding the long-term effects of CM on the brainstem, relatively little attention has been paid to the question of how pathophysiological mechanisms in the cortex—which are well established for EM-–operate in patients with CM. Three a priori possibilities are distinguishable: first, attacks of CM arise through additional pathophysiological mechanisms brought about by, ie, exposure to sustained large doses of analgesics; second, CM shares pathophysiological mechanisms with standard migraine; third, some combination of additional and shared mechanisms is responsible for CM.

Recent advances in noninvasive techniques for studying components of migraine pathophysiology permit straightforward examination of the second possibility outlined above. In standard migraine, we19 and others20,21 have used a variety of transcranial magnetic stimulation protocols to adduce converging evidence for functional cortical hyperexcitability in migraine. One technique, termed magnetic suppression of perceptual accuracy (MSPA) is an objective and reliable way of demonstrating excitability differences between migraine patients and controls.22 In the MSPA protocol, participants see a series of 3-letter trigrams flashed briefly on a computer screen. Each trigram is followed by a short interval of 40-190 ms, then a single magnetic pulse is delivered by a stimulation coil held against the occipital skull. In normal participants, plots of MSPA data (henceforth MSPA profiles) exhibit a characteristic U-shaped function; accuracy of reporting the letters is good at short (40 ms) and long (190 ms) intervals, but no better than chance for medium (100 ms) intervals. The cellular basis for suppression of perception at the 100 ms interval is likely to be the preferential activation of inhibitory neurons by a high-intensity TMS stimulus.23 Using MSPA we have previously demonstrated preliminary evidence of a spectrum of illness where CM patients had reduced visual suppression.24 In this paper we further confirm decreased inhibition and therefore hyperexcitability of the occipital cortex in migraine and also an activation of the brainstem as measured by blood flow in a subset of these patients.


Participants.— Twenty-five patients recruited who had CM with or without medication overuse using the revised proposal of the new International Headache Society classification (IHS 2004). The diagnosis was confirmed by a prospective electronic baseline diary. Patients who had a history of seizures, or any implanted medical device (eg, a pacemaker) were excluded. All participants had a normal neurological examination at the time of study. None of the participants was taking preventive drugs as well as drugs known25 to alter central nervous system excitability (sedatives, hypnotics, anticonvulsants or β-blockers) for at least 4 weeks prior to the study. Analgesic and abortive medications were permitted during this 4-week period. None, however, were taken in the 48-hour period prior to the study except for analgesics in the CM group. Studies in EM and CM patients were performed in the interictal period with at least 48 hours between the previous and next migraine attack. All studies were approved by Western Institution Review Board.

Apparatus.— All eligible participants underwent occipital cortex stimulation using the Magstim 200 (The MagStim Company Ltd, Whitland, Wales, UK). A 90-mm circular coil was used, which has 14 turns giving a peak magnetic field strength of 2 Tesla and 530 V/m of peak electric field strength.

Procedure: MSPA.—

Stimuli.— Visual target stimuli consisted of low contrast letter trigrams presented centrally within a frame. The letters were presented in upper case Arial 48-point font. The presence of a frame helped equalize any crowding effect on the letters and, therefore, also improved their legibility.26 The letters used were chosen from a subset of letters of approximately equal legibility.27 The letter trigrams, within the frame, subtended 1.21°× 0.55° of visual angle when viewed from a distance of 175 cm. There was no color contrast between the trigrams and the background: they differed only on a gray scale. Visual stimuli were presented for short durations (30-250 ms) on a Gateway 2000 monitor and PC, running SuperLab (Cedrus Corp., Phoenix, AZ, USA) software. For each trigram presented, participants were asked to verbally report the letters in the correct order, which was then recorded by an experimenter. The participants were asked to say “blank” or “don't know” if they were unaware of a letter at a specific position.

Phase 1—Practice.— Before magnetic stimulation began, participants completed a series of practice trials. The first 3 trigrams were presented for 250 ms to familiarize participants with the stimuli. Participants then completed 10 practice trials in which the trigrams were presented for 30 ms. If participants were unable to report the letters accurately by the end of this session, the practice trials were rerun to ensure that participants were familiar with the procedures and were able to accurately perceive the stimuli.

Phase 2—Time Course of Suppression.— In Phase 2, participants were presented with 54 trials in which the letter trigrams were followed at a variable interval (the stimulus onset asynchrony, SOA) by the magnetic pulse. Six SOAs were tested: 40, 70, 100, 130, 160, and 190 ms, with participants completing 9 trials at each SOA. The order of presentation of trials was randomized for each participant. The intensity of the TMS pulse was set at 75% of maximum stimulator output for this phase of testing. Throughout, an interval of at least 5 seconds between successive magnetic pulses was observed. Participants responded verbally by trying to name the letters in the order that they were presented.

PET Scans.— 18F fluoro-deoxyglucose positron emission test (18F-FDG PET) was performed for 10 of the 25 CM patients with an interval of at least 30 days. All scans were performed in the interictal period with the last reported migraine being at least 24 hours prior. One subject did report a migraine in the 6 hours following the PET scan. All patients received approximately 370-MBq 18F-FDG and were allowed to rest in a dimly lit room for approximately 45 minutes. Next, PET scanning of the brain was performed under standard resting conditions in a quiet, dimly lit room. FDG PET scans were performed with an Allegro PET scanner (Philips, Amsterdam, Netherlands) in 3-dimensional mode with RAMLA reconstruction and standard attenuation correction. Attenuation correction was performed with a transmission scan using a Cesium 137 source. Images were reconstructed by 3D RAMLA iterative technique through a 128 × 128 pixel image matrix (pixel size 2.0 × 2.0 mm2).

Images were reviewed by 1 nuclear medicine physician with additional training in neurologic PET imaging. The nuclear medicine physician was blinded to clinical information. A Sun workstation (Sun Microsystems Inc., Santa Clara, CA, USA) was used for visual image analysis. Automated analysis was used with the 3-dimensional stereotactic surface projections (3D-SSP) program (Neurostat, MI, USA). This program has been described previously in detail and has been evaluated for clinical and scientific use for analysis of PET and SPECT brains.28–31

Rotational alignment and centering of each brain scan were performed in 3 dimensions followed by spatial standardization along the anterior commissure-posterior commissure line. Linear scaling and nonlinear warping of each dataset were performed to adjust each individual brain to a proportional grid system, proposed by Talairich and Tournoux, resulting in a standardized image set with a uniform voxel size of 2.25 mm.32

Approximately 16,000 surface pixels along the surface of the brain and including along the mid-sagittal plane, reflected the metabolic activity with an average depth of 13.5 mm. All pixels were normalized to the average global pixel value. The resulting projections for each individual and for the summed group were compared to a global cerebral metabolism.


MSPA Profiles.— An average of MSPA profiles for CM group are shown in Figure 1 and are compared to MSPA profiles for normal controls and EM. For each participant, the percentage of correctly reported letters at each SOA was calculated. These percentages were then submitted to standard analysis of variance, with 3 levels of diagnosis (C, EM, CM) as a between-subjects factor, and 6 levels of SOA (40, 70, 100, 130, 160, 190 ms) as a within-subjects factor. There was a significant interaction between diagnosis and SOA, reflecting the difference in MSPA profiles evident in Figure 2 (F[2,34]= 71.67; P < .001). In order to examine the degree of visual suppression further across the 3 groups, a suppression index score was calculated for each participant as follows (maximum accuracy – minimum accuracy)/maximum accuracy. This index may be regarded as an estimate of the deepness/shallowness of the typically U-shaped MSPA profile for each patient. Index scores were submitted to standard analysis of variance with 3 levels of diagnosis (C, EM, CM) as the single between-subjects factor. There was a significant effect of diagnosis: the suppression index was largest in the C group, and smallest in the CM group, with the EM group falling in between (F[2,34]= 85.395; P < .001). All pairwise differences between diagnostic groups were also significant by post-hoc Tukey's HSD tests.

Figure 1.—.

Magnetic suppression of perceptual accuracy profiles by diagnostic group, separated by plot symbol. SOA is stimulus onset asynchrony, the time in milliseconds between the appearance of the letter trigram and the delivery of the TMS pulse. C = controls; EM = episodic migraine; CM = chronic migraine. Bars show standard errors.

Figure 2.—.

Suppression index for individual patients. (Note: no standard error bars for controls and episodic migraine because of smaller n's and more homogenous groups).

PET Results.— The mean scores for cerebral metabolism are shown in detail in the Table. There was increase in the cerebral metabolism in the pons (0.793) compared to global cerebral metabolism (0.037) (P < .01). There were also areas of increased metabolism in the right temporal cortex (0.589) compared to global metabolism (0.037) (P < .01) (Fig. 3). There were areas of decreased metabolism in several areas of bilateral medial frontal (and parietal as well as the somatosensory cortex [details in the Table]) (P < .1) (Fig. 3). Reduction in cerebral metabolism was also seen in bilateral caudate nuclei compared to global metabolism (details in the Table) (P < .1). There was no statistical significance for the increased metabolism in frontal and visual cortices. Decreased metabolism was seen in the cerebellum, anterior, and posterior cingulate cortices but this was also not statistically significant (details in the Table and Figs. 4 and 5).

Table Table.—.  Regional Average of Cerebral Metabolism
Brain RegionMeanSEMP valueBrain Region
PRT-R0.18540.2106 Parietal
PRT-L0.16890.2232 CTX
TMP-R−0.51160.1903P < .05Temporal
TMP-L−0.15890.2113 CTX
FRT-R−0.18790.161 Frontal
FRT-L−0.04240.1825 CTX
OCT-R−0.56380.2928 Occipital
OCT-L−0.53590.3404 CTX
PCING-R0.07260.1874 Posterior cingulate
PCING-L0.01970.1615 CTX
ACING-R−0.00710.2033 Anterior cingulate
ACING-L−0.09790.2407 CTX
MFRT-R0.43880.1482P < .01Medial frontal
MFRT-L0.26490.159P= .06CTX
MPRT-R0.80650.2022P < .01Medial parietal
MPRT-L0.67840.2183P < .01CTX
SMC-R0.7790.1558P << .01Somatosensory
SMC-L0.7340.1795P << .01CTX
VC-R−0.23520.27 Visual
VC-L−0.18450.4355 CTX
CAD-R0.79830.1601P << .01Caudate
CAD-L1.13680.3217P << .01 
CBL-R0.24650.514 Cerebellum
VER-R−0.34040.2943 Vermis
HEMI-R−0.23920.1281 Cerebral
HEMI-L−0.04990.1047 Hemispheres
PNS−0.92780.2962P < .01Pons
Figure 3.—.

Regional metabolism as compared to global meta- bolism.

Figure 4.—.

Positron emission testing scan decrease in regional metabolism.

Figure 5.—.

PET scan increase in regional metabolism.


In previous work we have demonstrated a significant difference in MSPA profile between EM patients (with aura) and matched controls22 as well as preliminary evidence in CM.24 The results we report in this paper both confirm and extend this previous finding. We continue to demonstrate that CM patients have MSPA profiles that are shallower again than EM patients-–Figure 1 demonstrates clearly that virtually no suppression of perception takes place in these patients. Statistical analysis of a suppression index reveals that this difference is robustly significant. These results of lack of visual suppression and, therefore, reduced inhibition in CM are confirmed by imaging studies of decreased metabolism by PET studies. The reduction in cerebral metabolism in medial frontal and parietal as well as somatosensory cortices may infer that normal inhibitory capacity of the cortex is reduced.

Previous PET techniques have used O15 to demonstrate increase in blood flow in the pons in spontaneous16 and glyceral trinatrate induced migraine.33 The 18F-FDG technique is a more robust way of assessing activation since it has the advantage of directly studying metabolism instead of blood flow. In this study we confirm the activation demonstrated in the pons in EM. Contrary to previous PET studies in migraine we did not see activation in the cingulate cortex. The reason for this is that none of the PET studies were performed during active migraine but were performed in the interictal period. The finding of activation in the pons despite being in the interictal period may be because of chronicity of the disorder in our patient population.

A compelling explanation for these converging results is that cortical excitability is raised still further in CM patients as compared to EM patients. Animal models suggest strongly that increases in cortical excitability, brought about by manipulation of intracortical inhibitory network activity, predispose cortex to pathophysiological activity.34 It is plausible, therefore, to argue that CM patients may be exceptionally susceptible to external or intrinsic triggers of migraine, precisely because of their high cortical excitability. The frequency of attacks is thus high.

We recognize limitations in this study in that it is a cross-sectional study of cerebral metabolism. We are in the process of doing follow-up PET scans on 6 of the 10 individuals following sham and active occipital nerve stimulation. We will study regions of activation and correlate the findings with the headache diaries. We hope that some patients as a result of the treatment35 may revert to EM and we will therefore then be able to demonstrate differences in EM and CM. The second limitation is a lack of controls in our study with the 18F-FDG PET. However, this technique is well established for studying several disorders28,30 and similar findings of pontine activation have been noted in EM.

Nonetheless, the robust statistical significance and the consistency of our neurophysiological findings with previous work permit the formulation of some intriguing research questions for the future. First, what can account for the very high excitability of CM cortex observed here? One interesting hypothesis is that of a vicious circle between brainstem and cortex. Exposure to repeated attacks and sustained doses of analgesics appears to result in brainstem changes as patients progress from EM to CM.15 If these brainstem changes also upregulate cortical activation, they will tend to increase attack frequency and hence analgesic consumption. Second, it will be of interest to develop longitudinal studies of the mechanisms by which EM may transition into CM and vice versa after treatment. Some preliminary work along these lines has already been reported: brain stem changes in CM appear to resolve after effective treatment.36 A complete picture of brainstem–cortex relationships in CM, over the evolution of the disorder, should be the goal of investigators in this area: we feel that the goal is attainable with the appropriate techniques of both neuroimaging and TMS-based functional studies as reported here.

Conflict of Interest::  Dr. Khodavirdi is an employee of Advanced Bionics.