We used phosphorus magnetic resonance spectroscopy (31P MRS) to investigate in vivo the brain and skeletal muscle energy metabolism of 15 children with migraine with aura in interictal periods. Brain 31P MRS disclosed low phosphocreatine and high inorganic phosphate contents, and high intracellular pH in all patients. Calculated [ADP] and the relative rate of mitochondrial oxidation were higher in the brain of patients than in control subjects, whereas the phosphorylation potential was lower. Brain intracellular free Mg2+ concentration was reduced by 25% in patients. Abnormal skeletal muscle mitochondrial respiration was also disclosed in 7 of 15 patients as shown by the slow rate of phosphocreatine postexercise recovery. The multisystem bioenergetic failure found in patients with juvenile migraine is comparable to that found in adults with different types of migraine.
Migraine onset occurs in about 40% of cases in childhood and adolescence(1). Affecting 4-10% of schoolchildren(2, 3), it is a frequent cause of headache in the pediatric age(4). Clinical differences occur between pediatric and adult migraine in terms of duration, location, intensity of attacks, and associated symptoms(5). Moreover, in the pediatric age, migraine can be associated with so-called “periodic syndromes” such as alternating hemiplegia(6) and paroxysmal vertigo(7).
An altered brain mitochondrial functionality is a common feature of different types of adult migraine as shown in vivo by 31P MRS during migraine attacks(8) and interictal periods(9–13). A similar bioenergetic impairment has also been found in other tissues such as skeletal muscle(9–14) and platelets(15) of adult migraine patients. Although the pathogenesis of bioenergetic failure is still hypothetical, the evidence of a low ictal free [Mg2+] in the brain(16) and of a low interictal [Mg2+] in serum and erythrocytes of migraine patients(17–19) suggested a role of hypomagnesemia in the pathogenesis of mitochondrial deficit and of the migraine attacks(20).
To test whether a mitochondrial dysfunction is an interictal feature also of juvenile migraine we used 31P MRS to study in vivo the brain and skeletal muscle energy metabolism in 15 children affected by different types of migraine with aura. Brain free intracellular [Mg2+] was also assessed by means of 31P MRS.
Patients. The patient group (Table 1) consisted of 15 patients (age 13.0 ± 2.05 y, mean ± SD; 9 boys and 6 girls) affected by migraine with aura, all with a positive family history. The diagnosis was made according to the definition of the Headache Classification Committee of the International Headache Society(21). In all patients, laboratory investigations excluded diabetes mellitus, connective tissue disorders, and coagulopathies. ECG, EEG, and brain magnetic resonance imaging were normal. To avoid metabolic changes related to therapies and headache attacks, all patients were free from any medication and were in an attack-free period lasting at least 7 d when they underwent 31P MRS. Informed consent was obtained from a parent of each patient, and studies were carried out with the approval of the hospital ethical committee.
31P MRS. 31P MRS was performed using a General Electric (GE) 1.5 Tesla Signa system and a spectroscopy accessory with a surface coil provided by GE and according to the quantification and quality assessment protocols defined by the EEC Concerted Research Project on“Tissue Characterization by MRS and MRI,” COMAC-BEM II.1.3(22).
Brain 31P MRS was performed on occipital lobes by positioning a surface coil on the occipital region after imaging the brain(23). The depth-resolved surface-coil spectroscopy localization technique(24) was used to avoid contribution to the signal by neck muscles, skin, and other interposed tissues. The repetition time was 5 s. The flip angle in the selected volume was approximately 30°, and it was not necessary to introduce any correction for saturation effects due to repetition time. Four hundred FIDs were accumulated to have a signal-to-noise ratio of 9-12 for β-ATP. A computerized curve-fitting program supplied by the data station (RUN GENCAP) was used to quantify the individual peaks of the spectrum(9). In particular, the critical analysis of the Pi peak was done using two Lorentzian curves to fit separately extracellular and intracellular Pi(25). By assuming a cytosolic [ATP] of 3 mM(26), we calculated the concentration of Pi and PCr. We do not have absolute data on ATP content in the brain of our patients. Nevertheless, if [ATP] were lower in these patients, [PCr] would be even lower. Mitochondrial functionality was assessed by calculating the ADP concentration(27), a major regulator of mitochondrial respiration(28), the phosphorylation potential = [ATP]/[ADP] × [Pi], a measure of the readily available free energy in the cell(29), and the percentage of the maximal rate of ATP biosynthesis (V/Vmax)(30).
Brain cytosolic free [Mg2+] was assessed by a semiempirical equation that correlates the chemical shift of the β-ATP signal from PCr to the free [Mg2+] and taking into account most cytosolic phosphate compounds binding Mg2+ More precisely, the in vitro calibration was obtained using solutions mimicking the in vivo brain cytosolic concentration of PCr. creatine, Pi, ATP, the cytosolic pH, the intracellular ionic strength, and temperature(31). To handle the relevant number of different equilibria and variables, an algorithm and a software previously developed were used(32).
Muscle 31P MRS was performed on the right calf muscle(33) by the pulse-and-acquire technique (repetition time of 5 s) at rest, during in-magnet aerobic isokinetic exercise(34), and during recovery from exercise. Sixty FIDs were accumulated during rest (5 min); then exercise was begun and data were collected for 1 min (12 FIDs) for each level of work. As soon as the last minute of work was completed and the corresponding 12 FIDs recorded, one 2-FIDs data block (10 s) was also recorded to be considered zero time, and the exercise stopped immediately afterward. During postexercise recovery 10-s data blocks (2 FIDs) were recorded during the first 60 s, and 30-s data blocks thereafter for another 4 min. The limits of all the peaks were marked manually on each spectrum after phasing, and areas were calculated between the limits(34).
The efficiency of mitochondrial ATP production was assessed by measuring the rate of PCr resynthesis during recovery that is entirely oxidative(35). The rate of PCr resynthesis was calculated from the monoexponential equation best fitting the experimental points of recovery, and reported as TC. In consideration of the linear relationship linking the PCr recovery to the minimum value of intracellular pH reached during recovery(33), the single patients' TC values were plotted as a function of the minimum pH.
Intracellular pH was calculated from the chemical shift of Pi relative to PCr(36). The chemical shift was carefully determined from the centroid of PCr peak to the centroid of the phosphate peak. All MRS data were analyzed blinded to the status (patient/control) of the subjects.
Control data. Control subjects were healthy children aged 9-16 y (mean ± SD); 12 (age 12.9 ± 1.9 y; 5 girls), for brain studies, and 15 (age 13.1 ± 2.08 y; 6 girls), for muscle studies. Informed consent was obtained from a parent of each subject. Data are given as mean ± SD. Statistically significant results, determined by an unpairedt test, were taken as p < 0.05.
The upper section of Figure 1 shows typical 31P MR spectra of occipital lobes from a patient with migraine with aura (spectrum A, case no. 8) and a sex- and age-matched control subject (spectrum B). Brain PCr concentration was remarkably reduced in this patient as well as in the other migrainous patients, the mean value of the patients' group being below 2 SD of the control group mean (Table 2). In the migrainous group we also found a parallel increase of Pi and a small, although significant, increase of cytosolic pH (Table 2). Calculated [ADP] and the relative rate of mitochondrial ATP biosynthesis(V/Vmax%) were increased in the patients, whereas the phosphorylation potential was decreased (Table 2), all changes having high statistical significance. The young migraine patients also showed a low cytosolic concentration of free Mg2+ in their brains(Table 2).
31P MR spectroscopy of resting calf muscles did not show any difference between patients and matched healthy volunteers (data not shown). All patients and control subjects exercised inside the magnet, reaching the PCr depletion required to assess the rate of PCr resynthesis during recovery-a sensitive measure of the efficiency of mitochondrial ATP production. The middle section of Figure 1 shows the calf muscle spectra from a patient (spectrum C, case no. 7) and a matched control subject(spectrum D) at the end of work, both having depleted PCr to the same extent. The bottom section of Figure 1 reports the spectra collected from 30 to 40 s of recovery from the same individuals (spectrum E, case no. 7; spectrum F, matched control subject). PCr in the patient was lower than in the control subject at this time of recovery, indicating delayed PCr resynthesis in this patient. The rate of PCr postexercise recovery was quantified by a monoexponential equation best fitting the experimental points and reported as TC, as shown in the top section of Figure 2.
In view of the different degree of cytosolic acidification during recovery, and the effect of cytosolic pH on the rate of PCr postexercise resynthesis(33), TC values of PCr recovery from single patients were reported as a function of the lowest value of cytosolic pH reached during recovery (Fig. 2, bottom section). Seven migrainous patients had TC values outside the reference range, as defined by the 95% confidence band around the regression line linking the rate of PCr recovery and the lowest value of cytosolic pH reached during recovery (termed minimum pH). No correlation was found between type of aura, age of onset, and number of attacks experienced at the time of MRS, and extent of bioenergetic brain and/or muscle deficit and brain [Mg2+] reduction (data not shown).
The main finding of our study was an altered brain energy metabolism in children with migraine with aura and a reduced brain cytosolic free Mg2+ concentration. It is worth underlining that l) all patients experienced a limited number of attacks (two to eight per year) at the time of MRS examination, the migraine onset being from 1 to 5 y; 2) all patients were studied during an attack-free period; and 3) no patient was taking any medication.
The impairment of brain oxidative metabolism was shown in young migraine patients by reduced PCr and increased Pi contents. Cytosolic pH was also altered showing a small although significant increase. As a consequence, the calculated ADP concentration was remarkably greater in patients' than in control subjects' brain. ADP is a major regulator of mitochondrial respiration(28). Therefore, the high ADP concentration shows that brain cells are operating nearer to the asymptote of the hyperbola describing the ADP control of respiration, and indicates that they are less able to handle any further energy demand. Quasistable brain bioenergetics were also shown by increased relative rate of energy metabolism(V/Vmax%, Table 2). A high relative rate of metabolism may be due to reduced Vmax in malfunctioning mitochondria or to increased rate of respiration because of increased neuronal energy demand(20), of both. Unstable metabolic state was further shown in our patients' brain by a remarkable reduction(Table 2) of the phosphorylation potential, a measure of cell's free energy availability(29).
Brain cytosolic free [Mg2+] in our healthy control subjects is lower than the concentration reported by others using 31P MRS(37, 38). This is not surprising because the model underlying our calibration curve was based on a larger number of species binding Mg2+. We developed a new calibration curve for the in vivo assessment of free Mg2+ as the existing ones were not specifically developed for brain conditions. Our calibration curve relies onin vitro high resolution NMR measurements performed on solutions mimicking in vivo conditions, and on a chemical model of a multiequilibrium system of 25 species in the physiologic region of pH, taking into consideration both ionic strength and temperature(31).
Migraine patients showed in their occipital lobes an intracellular free[Mg2+], measured interictally, significantly reduced. Ramadan et al.(16) showed a reduced brain [Mg2+] in a group of migraine patients with and without aura studied during the migraine attack, but in the same study they failed to demonstrate a significant interictal reduction of brain [Mg2+] in nine migraine patients with and without aura(16). The absence of an interictal significant reduction in brain [Mg2+] in Ramadan's study can be due to different factors, such as the smaller number of cases studied and the heterogeneity of the migraine patient group and of the brain area studied(i.e. frontal, frontotemporal, parieto-occipital, and occipital cortex). A preliminary report from the same group shows that in patients with hemiplegic migraine the interictal [Mg2+] is reduced in occipital lobes, whereas being normal in frontal lobes(39). The 25% reduction in occipital lobes free [Mg2+] found in vivo in our patients can be due, at least in part, to low total [Mg2+] as found by others in serum and erythrocytes of migrainous patients(17–19).
[Mg2+] is known to influence the equilibrium constant of several cytosolic reactions including creatine kinase(40). In our study [ADP] was calculated from the creatine kinase equilibrium using the same equilibrium constant for patients and control subjects. Changes of[Mg2+] in our patients were relatively small, thus being responsible for a very little change, if any, of the equilibrium constant of creatine kinase. Nevertheless, if this were the case, reduced [Mg2+] would reduce the value of the equilibrium constant(40) and lead to even higher calculated [ADP].
Alkalotic pH, found in our young patients, can be a consequence of ionic abnormalities present in the brain of migraine patients, which in turn may affect proton pumps. Also in familial hemiplegic migraine, due to a P/Q-type Ca2+ channel defect(41), increased pH in occipital lobes has been reported in association with low Mg2+(39). The suggestion that some cases of migraine with aura too may be linked to abnormalities of the P/Q-type Ca2+ channel(42) leads to the intriguing speculation that the low content of Mg2+, which interferes with Ca2+ channels(43), may be related to calcium channel dysfunction. Low[Mg2+] may contribute to brain bioenergetic impairment, as Mg2+ is essential for mitochondrial membrane stability and coupling oxidative phosphorylation(20). Moreover, the alteration of plasma membrane permeability due to low Mg2+ and, possibly, P/Q-type Ca2+ channel dysfunction, may increase the ATP consumption by ATP-dependent ion pumps.
Mitochondrial function was also impaired in the skeletal muscle of several of our young patients as previously found in adult migraine patients(9, 10, 12, 14). Their bioenergetic deficit was not large enough to be seen in the resting muscle. However, when a metabolic stress was applied to their muscles by exercising, 7 of 15 patients showed a failure of energy metabolism, as shown by a slow rate of PCr resynthesis after exercise (bottom section of Fig. 2). It is worth underlining that postexercise PCr biosynthesis, depending entirely on mitochondrial respiration, is a very sensitive index of mitochondrial functionality(44, 45). It seems interesting to point out that the percentage of young migraine patients with abnormal muscle mitochondrial respiration is somewhat similar to that found in adult patients with migraine with aura(46).
We do not know why in our young migraine patients a similar bioenergetic and Mg2+ deficit resulted in different forms of aura. We conceive that other environmental or genetic influences may modify the phenotypic expression of a similar biochemical deficit. Along these lines is worth recalling that in mitochondrial encephalomyopathies a similar bioenergetic abnormality(23) and even the same mitochondrial DNA mutation may result in different clinical expressions(47).
Our present findings, by showing a multisystem bioenergetic defect in young migraine patients. further strengthen our hypothesis(9, 10, 12) that a bioenergetic defect, unrelated to the age of migraine onset, migraine subtype, number of attacks experienced, medications used, and aging, is: 1) an intrinsic feature of migraine, and 2) a pathogenic background able to enhance the susceptibility to develop headache when brain energy demand is increased and/or the supply of oxidizable substrates and oxygen is decreased.
magnetic resonance spectroscopy
relative rate of mitochondrial ATP production
free induction decay
Raskin NH 1988; Migraine: clinical aspects. In: Headache, 2nd Ed. Churchill-Livingstone, New York, 41–51.
Billi B 1962; Migraine in school children. Acta Paediatr 51: suppl 136 1–151.
Abu-Arafeh I, Russel G 1995; Prevalence and clinical features of abdominal migraine compared with those of migraine headache. Arch Dis Child 72: 413–417.
Hockaday J 1988; Definition, clinical features, and diagnosis of childhood migraine. In: Hockaday J (ed) Migraine in Childhood. Butterworth, London, 5–24.
Winner P, Martinez W, Mate L, Bello L 1995; Classification of pediatric migraine: proposed revision to the IHS criteria. Headache 35: 407–410.
Silver K, Anderman F 1993; Alterning Hemiplegia of childhood: a study of 10 patients and results of flunarizine treatment. Neurology 43: 36–41.
Abu-Arafeh I, Russel G 1995; Paroxysmal vertigo as a migraine equivalent in children: a population-based study. Cephalalgia 15: 411–414.
Welch KMA, Levine SR, D'Andrea G, Schultz LR, Helpern JA 1989; Preliminary observations on brain energy metabolism in migraine studied by in vivo phosphorus 31 NMR spectroscopy. Neurology 39: 538–541.
Barbiroli B, Montagna P, Cortelli P, Martinelli T, Sacquegna P, Zaniol P, Lugaresi E 1990; Complicated migraine studied by phosphorus magnetic resonance spectroscopy. Cephalalgia 10: 263–272.
Barbiroli B, Montagna P, Funicello R, Iotti S, Monari L, Pierangeli G, Zaniol P, Lugaresi E 1992; Abnormal brain and muscle energy metabolism shown by 31P magnetic resonance spectroscopy in patients affected by migraine with aura. Neurology 42: 1209–1214.
Sacquegna T, Lodi R, DeCarolis P, Tinuper P, Cortelli P, Zaniol P, Funicello R, Montagna P, Barbiroli B 1992; Brain energy metabolism studied by 31P-MR spectroscopy in a case of migraine with prolonged aura. Acta Neurol Scand 86: 376–380.
Montagna P, Cortelli P, Monari L, Pierangeli G, Parchi P, Lodi R, Iotti S, Frassineti C, Zaniol P, Barbiroli B 199431; 31 P-magnetic resonance spectroscopy in migraine without aura. Neurology 44: 666–669.
Uncini A, Lodi R, DiMuzio A, Silvestri G, Servidei S, Lugaresi A, Iotti S, Zaniol P, Barbiroli B 1995; Abnormal brain and muscle energy metabolism shown by 31P-MRS in familial hemiplegic migraine. J Neurol Sci 129: 214–222.
Lodi R, Kemp G, Montagna P, Pierangeli G, Cortelli P, Iotti S, Radda G, Barbiroli B 1997; Quantitative analysis of skeletal muscle bioenergetics and proton efflux in migraine and cluster headache. J Neurol Sci 146: 73–80.
Sangiorgi S, Mochi M, Riva R, Cortelli P, Monari L, Pierangeli G, Montagna P 1994; Abnormal platelet mitochondrial function in patients affected by migraine with and without aura. Cephalalgia 14: 21–23.
Ramadan NM, Halvorson H, Vande-Linde A, Levine H, Low brain magnesium in migraine. Headache 29: 590–593.
Sarchielli P, Costa G, Firerize C, Morucci P, Abritti G, Galla V 1992; Serum and salivary magnesium levels in migraine and tension-type headaches. Results in a group of adult patients. Cephalalgia 12: 21–27.
Thomas, Tomb E, Thomas E, Faure G 1994; Migraine treatment by oral magnesium intake and correction of the irritation of buccofacial and cervical muscles as a side effect of mandibular imbalance. Magn Res 7: 123–127.
Soriani S, Arnaldi C, DeCarlo L, Arcudi D. Mazzotta D, Battistella PA, Sartori S, Abbasciano V 1995; Serum and red blood cell magnesium levels in juvenile migraine patients. Headache 35: 14–16.
Welch KMA, Ramadan NM 1995; Mitochondria, magnesium and migraine. J Neurol Sci 134: 9–14.
Headache Classification Committee of the International Headache Society 1988; Classification and diagnostic criteria for headache disorders, cranial neuralgias and facial pain. Cephalalgia 8: 1–96.
EEC Concerted Research Project 1995; Quality Assessment in in vivo NMR spectroscopy. Results of a concerted Research Project of the European Economic Community (6 papers). Magn Reson Imaging 13: 115–176.
Barbiroli B, Montagna P, Martinelli P, Lodi R, Iotti S, Cortelli P, Funicello R, Zaniol P 1993; Defective brain energy metabolism shown by in vivo31P MR spectroscopy in 28 patients with mitochondrial cytopathies. J Cereb Blood Flow Metab 13: 469–474.
Bottomly PA, Foster TH, Darrow RD 1984; Depth-resolved surface-coil spectroscopy (DRESS) for in vivo1H, 31P, and 13C NMR. J Magn Reson 59: 338–342.
Barbiroli B. Lodi R, Iotti S 1996; Phosphorus MR spectroscopy of human brain. An overview after several years' experience studying some 1,200 patients. In: Podo F. Bovée WMMJ, de Certaines J, Henriksen O, Leach M. Leibfritz D (eds) Eurospin Annual. Istituto Superiore di Sanitá, Rome, 385–396.
Bottomly PA, Hardy CJ 1989; Rapid, reliable in vivo assay of human phosphate metabolites by nuclear magnetic resonance. Clin Chem 59: 392–395.
Degani H, Alger J, Shulman R, Petroff O, Prichard J 1987; 31P magnetization transfer studies of creatine kinase kinetics in living rabbit brain. Magn Reson Med 5: 1–12.
Chance B Willams GR 1955; Respiratory enzymes in oxidative phosphorylation. III. The steady state. J Biol Chem 217: 409–427.
Veech RL, Lawson JWR, Cornell NW, Krebs HA 1979; Cytosolic phosphorylation potential. J Biol Chem 254: 6538–6547.
Nioka S, Chance B, Hilberman M, Subramanian H, Leigh J, Veech R. Forster R 1987; Relationship between intracellular pH and energy metabolism in dog brain as measured by 31P-NMR. J Appl Physiol 62: 2094–2102.
Iotti S, Frassineti C, Alderighi L, Sabatini A, Vacca A, Barbiroli B 1996; In vivo assessment of free magnesium concentration in human brain by 31P-MRS. A new calibration curve based on a mathematical algorithm. NMR Biomed 9: 24–32.
Sabatini A, Vacca A, Gans P 1992; Mathematical algorithms and computer programs for the determination of equilibrium constants from potentiometric and spectrophotometric measurements. Coord Chem Rev 120: 389–405.
Iotti S, Lodi R, Frassineti C, Zaniol P, Barbiroli B 1993; In vivo assessment of mitochondrial functionality in human gastrocnemius muscle by 31P MRS. The role of pH in the evaluation of phosphocreatine and inorganic phosphate recoveries from exercise. NMR Biomed 6: 248–253.
Zaniol P, Serafini M, Ferraresi M, Golinelli R, Bassoli P, Canossi I, Aprilesi GC Barbiroli B 1992; Muscle 31P-MR spectroscopy performed routinely in a clinical environment by a whole-body imager/spectrometer. Phys Med 8: 87–91.
Kemp GJ, Taylor DJ, Radda GK 1993; Control of phosphocreatine resynthesis during recovery from exercise in human skeletal muscle. NMR Biomed 6: 66–72.
Petroff OAC, Prichard JW, Behar KL, Alger JR, Shulman T 1985; Cerebral metabolism in hyper- and hypocarbia: 31P and 1H nuclear magnetic resonance studies. Neurology 35: 1681–1688.
Taylor JS, Vigneron DB, Murphy-Boesch J, Nelson SJ, Kessler HB, Coia L, Curran W, Brown TR 1991; Free magnesium levels in normal human brain and brain tumours: 31P chemical-shift imaging measurements at 1.5 T. Proc Natl Acad Sci USA 88: 6810–6814.
Halvorson HR, Vande-Linde AMQ, Helpern JA, Welch KMA 1992; Assessment of magnesium concentration by 31P NMR in vivo. NMR Biomed 5: 53–58.
Barker PB, Boska MD, Butterworth EJ, Hearshen D, Aurora TK, Norris L, Duyn JH, Mooen CTW, Welch KMA 1995; Phosphorus and proton MR spectroscopic imaging in migraine. 3rd Annual Meeting, Society for Magnetic Resonance, Nice, France, 139
Lawson JWR, Veech RL 1979; Effects of pH and free Mg2+ on the Keq of the creatine kinase reaction and other phosphate hydrolyses and phosphate transfer reactions. J Biol Chem 254: 6528–6537.
Ophoff R, Terwindt G, Vergouwe M, vanEijk R, Oefner P, Hoffman S, Lamerdin J, Mohrenweiser H, Bulman D, Ferrari M, Haan J, Eindhout D, vanOmmen G, Hofker M, Ferrari M. Frants R 1996; Familial hemiplegic migraine and episodic ataxia type-2 are caused by mutations in the Ca2+ channel gene CACNLIA4. Cell 87: 543–552
May A, Ophoff R, Terwindt G, Urban C, vanEijk R, Haan J, Diener H, Lindhout D, Frants R, Sandkuijl L, Ferrari M 1995; Familial hemiplegic migraine locus on 19p13 is involved in the common forms of migraine with and without aura. Hum Genet 96: 604–608.
Altura B 1985; Calcium antagonist properties of magnesium: implications for antimigraine actions. Magnesium 4: 169–175.
Arnold DL, Matthews PM, Radda GK 1984; Metabolic recovery after exercise and the assessment of mitochondrial function in vivo in human skeletal muscle by means of P-31 NMR. Mag Reson Med 1: 307–315.
Taylor DJ, Bore PJ, Gadian DG, Radda GK 1983; Bioenergetics of intact human muscle. A 31P nuclear magnet resonance study. Mol Biol Med 1: 77–94.
Montagna P, Cortelli P, Barbiroli B 1994; Magnetic resonance spectroscopy studies in migraine. Cephalalgia 14: 184–193.
DiMauro S, Moraes CT 1993; Mitochondrial Encephalomyopathies. Arch Neurol 50: 1197–1208.
Supported by Telethon-Italy (R.L.).
About this article
Cite this article
Lodi, R., Montagna, P., Soriani, S. et al. Deficit of Brain and Skeletal Muscle Bioenergetics and Low Brain Magnesium in Juvenile Migraine: An in Vivo 31P Magnetic Resonance Spectroscopy Interictal Study. Pediatr Res 42, 866–871 (1997). https://doi.org/10.1203/00006450-199712000-00024
Nature Reviews Neurology (2019)
Canadian Journal of Neurological Sciences / Journal Canadien des Sciences Neurologiques (2019)
Evidence of an increased neuronal activation-to-resting glucose uptake ratio in the visual cortex of migraine patients: a study comparing 18FDG-PET and visual evoked potentials
The Journal of Headache and Pain (2018)
Current Pain and Headache Reports (2018)