Habituation of laser-evoked potentials by migraine phase: a blinded longitudinal study

Seventy-four migraineurs and 40 controls responded to an advertisement in the local newspaper, on the local hospital’s web page [46] and the Intranet within our university [47]. We screened both groups over telephone and migraineurs were evaluated by neurologists per the ICHD-II criteria for migraine with or without aura [48]. Controls could not have a headache more than once a month. If they occasionally had a headache, we asked if they had consulted a physician regarding the headache, if they experienced the headache as painful and if they used abortive medication for their headache. We excluded controls if they confirmed more than one of these three questions. Included migraineurs had an attack frequency between two and six per month and had no more than ten days with migraine attacks per month. They could use symptomatic, but not prophylactic migraine treatment. Exclusion criteria were: coexisting tension-type headache seven days or more per week in migraineurs, neurological or psychiatric diseases, sleep disorders, active infectious diseases, connective tissue diseases, metabolic, endocrine or neuromuscular diseases, other clinically relevant painful conditions including recent injuries, malignancy, previous craniotomy or cervical spine surgery, heart disease, cardiopulmonary or cerebrovascular diseases, pregnancy, medication for acute or chronic pain, antipsychotics, antidepressants, anticonvulsants or other drugs that may influence neuronal, vascular or muscular function, substance abuse, ferromagnetic implants and prophylactic allergy treatment.

Fifty migraineurs and 31 controls participated in the study. One migraineur withdrew consent after the first examination and was not included in the analysis. Three migraineurs attended only once, twice and three out of four times respectively. We excluded one control because we were unable to obtain reliable LEPs as most trials were rejected. Thus, 49 migraineurs completed a total of 190 examinations, and 30 controls completed one examination each. Table 1 shows demographic and clinical data. We report details of exclusions and dropouts in Additional file 1: Table S1.

The Regional Committees for Medical and Health Research Ethics approved the protocol, and all subjects gave their written informed consent. Migraineurs and controls received an equivalent of $ 125 and $ 30 respectively to cover expenses.

Painful heat stimuli were generated by a pulsed solid-state (Nd:YAP) laser (STIMUL 1340, DEKA M.E.L.A. SRL, Calenzano (FI), Italy) with a wavelength of 1340 nm. The laser stimulator settings were the same as in a previous study at our lab [49]: The pulse duration was 6 ms, a relatively short stimulus duration to maximize the N1-amplitude [50]. We set the laser beam diameter to 8 mm (area ≈ 50 mm²) with an energy ranging from 2 to 6.5 J (4.0–12.9 J/cm²). The diameter and durations are comparable to other researchers using the same type of laser [51, 52]. A diode laser aiming beam visualized the stimulation site. We recorded LEPs with a Viking Select system (Nicolet Biomedical Inc., Madison, WI, USA). The recording silver disc electrodes were placed at the Fz, Cz, Pz, T3, T4, A1 and A2 sites of the 10–20 system. The impedance was kept below 5 kΩ. The two most important analysis channels, Cz referred to the nose, and T3 referred to Fz, were preselected as recommended by the international IFCN-guidelines [53]. We used the other channels as back up to account for interindividual variation in field topography and to improve detectability of waves. For control of artifacts, we monitored the electrooculogram from a left infraorbital electrode referred to T4. The onset of stimuli triggered the recording system. The sampling rate was 1000 Hz, the sweep time was 750 ms and the filter setting was 0.2–100 Hz. Rejection level was set to ±225 μV, and total rejection rate after exclusions was 3 %. We applied online averaging [49, 54] since rejection effectively canceled artifacts and eye movements also were included in a separate channel.

Subjects lay comfortably on an examination table with laser safety glasses and acoustic earmuffs to avoid any acoustical interference at the time of stimulation [53, 55]. We delivered laser stimuli to the dorsum of the right hand between the carpal bones, metacarpophalangeal joints and second and fourth metacarpal bone. The laser beam was moved randomly within this area to avoid skin lesions and nociceptor fatigue or sensitization [56]. We measured skin temperature before the test. Because we previously observed that the recommend fixed intensity (equal to twice the mean pin-prick threshold [12]) did not always elicit pain and LEP in every healthy subject [49], we used stimulus intensities based on intraindividual thresholds [51, 57, 58]. First, the individual thresholds for pinprick pain were identified, starting at 2 J and increasing with 0.5 J steps [49]. The subject had to differentiate between burning pain and pinprick pain. Subjects scored pinprick pain on a verbal, numerical rating scale (NRS) with range 0 = “no pain” to 10 = “unbearable pain.” We measured the threshold twice, and we defined the pinprick threshold as the lowest intensity inducing pain in at least one of the two trials at that intensity. A tolerable intensity was 4.5 J (9.0 J/cm²), in most subjects, corresponding to about two times the pinprick threshold. However, 4.5 J generated too much pain in 22 subjects, and too little pain in five subjects and the energy had to be adjusted up or down (range: 3–5.5 J ≈ 6.0–10.9 J/cm²). The chosen intensity did generally elicit reliable N2P2 potentials [59]. We recorded two blocks of 21 stimulations with six to ten seconds between each stimulation since six seconds was recommended as the minimum interval to avoid peripheral nociceptor habituation [53]. The between-block interval was also short, between 6 and 10 s, to prevent recovery of central habituation. Subjects kept their eyes open and rated perceived pain verbally (NRS 0–10) after each stimulation to prevent LEP-amplitude decrease by distraction and drowsiness [12, 60, 61]. We stored pain scores for analysis. We used identical energy levels for all sessions within subjects. However, we did not tell participants that the energy was constant.

Examinations were classified by the headache diary as interictal (more than one day before attack onset or one day after the attack ended), preictal (less than one day before attack onset), ictal (a migraine headache during the examination) and postictal (less than one day after the attack ended). We applied this definition in previous studies of pain physiology related to migraine phase [6, 62]. Eleven of the 190 examinations were unclassifiable and excluded from data analysis, mainly because they had migraines both the day before and the day after examination.

We analyzed in LabChart® (Version 7 pro, ADInstruments, Dunedin, New Zealand). A random number identified each LEP session and we randomized the order of the two blocks within each session. Thus, the investigator who analyzed the LEPs was blinded to diagnosis, migraine phase and order of the two blocks. The N1 and the N2P2 components were assessed, N1 at the contralateral temporal electrode (T3) against Fz (best bipolar derivation to show N1 [63]) and N2P2 at Cz against nose [53]. We measured the N1-amplitude from baseline (start of N1) to the N1 peak and the N2P2-amplitude from the most negative to the most positive peak. We had to discard some LEPs due to unrecognizable responses, too much noise/artifacts or latencies far from normal values [53]. The N1-amplitude may have a low signal-to-noise ratio, and it was not detectable in 15% of LEPs in migraineurs and 22% of LEPs in controls. These responses were included in the analysis as interval censored responses [64, 65] by setting the lower bound to zero and the upper bound to the maximal negative noise peak within the N1-time window. The exact N1-amplitude was then unknown, but we presumed that it was between zero and the largest noise peak, and we included the amplitude as an interval, a rough estimate, instead of a point estimate. We discarded recordings with technical errors, 17 of 358 in total in migraineurs. We present the grand average of all recordings by phase (Fig. 1) and by group (Fig. 2).

We analyzed data with STATA version 13.1 (StataCorp LP). We applied separate multilevel models [66] for the response variables N1, N2P2 and pain scores. The first set compared the within-subject change in migraineurs and the second set compared interictal recordings from the first exam and controls. The interaction effects were the main outcomes. For the first set, interactions-of-interest reflect differences between phases. For the second set, interactions-of-interest reflect differences between migraine and healthy subjects. We have included full model specifications in the Additional file 1.

We added explorative post hoc contrasts that tested if the slopes (i.e., habituation) at each phase, and in all phases combined, were different from zero. Also, based on the main results, we performed data-driven explorative post hoc contrasts comparing the first-block amplitudes and habituation slope of N1 between the preictal and ictal phases. We used diagnosis as a fixed factor to compare migraineurs in the interictal phase and controls. Only interictal recordings from the first exam were compared to controls because the investigator was blinded to diagnosis only on the first session, and to avoid possible long-term habituation/sensitization effects in subsequent exams. N1 and N2P2 in both sets were square rooted to improve normality.

We analyzed phase and group differences in pain scores with pain scores from both blocks combined consecutively into one continuous time variable which was interacted with phase or group, respectively. The time variable was centered at its mean and divided by 10. Thus, the regression constant shows the average pain score and the habituation coefficient the change in pain score per ten stimulations. We tested group differences in pain thresholds and laser intensity with independent samples Student’s t-test and present results as 95% confidence intervals (CI). Thus we consider intervals not containing zero to be significant at a level of p < 0.05. We have back transformed the data and tabulate result-parameters in the original scale. However, results are presented as transformed units in the Additional file 1.

We extended the original models to test the effect of additional variables. We specified four separate models that estimated the phase-differences in habituation for 1) migraineurs with and without aura, 2) sessions differentiated by headache laterality, 3) by years lived with migraine, and 4) by pain scores. Headache laterality was classified by the related attack if the phase was preictal, ictal or postictal. Interictal recordings were classified by the side the subject most commonly experienced headache, either left, right or bilateral. Sixteen interictal recordings had an equal amount of left and right-sided unilateral migraine and were not included in the laterality-analysis. We included age, migraine intensity, and migraine frequency as control variables (not included in the interactions) in the extended model 3 that estimated the effect of years lived with migraine.

We conducted three additional analyses to explore the relationship between habituation and number of days to next attack. We conducted these analyses in two steps, first with interictal phase only and then with both interictal and preictal phases included. We interpret the interaction effects in the latter analyses as the interictal-preictal day-to-day change in habituation towards the next migraine attack. Also, we performed a secondary set of analyses with a three-day limit to test if postictal phase-related LEP-changes last longer than 24 h after the attack.

With 30 controls and 50 migraineurs, the statistical power to detect a low medium-sized effect equal to 0.65 SD [67] based on a two-sample Student’s t-test is 80%. As we estimated to have approximately 20 pairs for intraindividual phase-related comparisons, power (based on paired Student’s t-tests) to detect a similar medium-sized effect (0.65 SD) was calculated to 83%.

Table 2 shows means and standard deviations of N1 and N2P2-amplitudes, and pain scores by phase.

N1-habituation was significant in the interictal phase as shown by the negative coefficient of block (Fig. 3 and Table 3). The degree of habituation was not different between the interictal phase and the preictal, ictal and postictal phases respectively (interaction effects in Table 3). However, post hoc contrasts showed significant habituation in the preictal phase (95% CI [−2.71, −0.33] μV/block), but not in the ictal and postictal phases (95% CI [−1.29, 0.57] and [−3.44, 0.00] μV/block, respectively). The habituation in all phases combined was significant (95% CI [−1.40, −0.38] μV/block). The contrast of the difference in habituation between the preictal and ictal phase was not significant (95% CI [−0.54, 2.86] μV/block). Neither the first-block amplitudes nor the combined first and second-block amplitudes differed between phases, but the post hoc contrast that compared first-block amplitudes between the preictal and ictal phases showed a tendency towards lower first-block amplitudes in the ictal phase (95% CI [−3.20, 0.04] μV).

The N2P2-amplitude change from the first to the second block was significant in the interictal phase, and none of the interactions were significant (Fig. 3 and Table 3), interpreted as interictal habituation with no differences between phases. Post-hoc contrasts showed significant habituation in both the preictal and ictal phases (95% CI [−6.66, −1.58] and [−7.38, −0.77] μV/block, respectively), but not in the postictal phase (95% CI [−7.01, 1.22] μV/block). The habituation in all phases combined was significant (95% CI [−4.90, −2.55] μV/block). N2P2-amplitude sizes did not differ between phases.

Pain scores increased linearly in the interictal phase (95% CI [0.11, 0.33] NRS-change/10 stimuli, Table 3). The linear increase, i.e., sensitization of pain scores, was not different between phases. Mean pain scores did not differ between phases.

We present complete results from the secondary analyses in the Additional file 1. N1 and N2P2 first-block amplitudes and habituation did not differ between migraineurs with and without aura as none of the three-way or two-way interactions were significant. Amplitudes and habituation did not differ between a left and right-sided migraine. Subjects with a bilateral migraine had reduced N2P2-habituation (more positive slope) in the postictal phase compared to the interictal phase (95% CI [0.07, 25.3] μV/block), and the same tendency was present in the ictal phase (95% CI [−0.01, 19.8] μV/block). The more years lived with migraine; the less was the N2P2-habituation in the preictal phase compared to the interictal phase (95% CI [0.11, 0.82] μV/block/year adjusted for age). No other interaction effects were significant.

Both N1 and N2P2 interictal first-block amplitudes correlated with pain scores (95% CI [0.08, 0.93] and [1.74, 3.65] μV/unit pain score for N1 and N2P2, respectively). The interactions between phase and pain score were not significant, that is, the correlations were not different between phases. Habituation of N1 and N2P2-amplitudes did not correlate with a change in pain scores from the first to the second block.

The analyses that explored the relationship between habituation and number of days to next attack showed no significant interactions. Thus, there was no interictal day-to-day linear change in habituation towards the next migraine attack. Changing the definition of the postictal phase from a one-day limit to a three-day limit did not alter the interpretation of LEP-habituation. Habituation of pain scores did not change by changing the definitions of the phases, although the mean pain score was significantly increased in the postictal compared to the interictal phase (95% CI [0.11, 1.28] unit pain score).

Table 4 shows means and standard deviations of N1 and N2P2-amplitudes, and pain scores by group.

Controls did not show habituation of the N1 amplitude, and habituation did not differ between migraineurs and controls (Fig. 4 and Table 5). Post-hoc contrasts showed no habituation in the groups combined (95% CI [−1.32, 0.30] μV) and no habituation in the migraine group (95% CI [−1.77, 0.49] μV). Neither the first-block amplitudes nor the combined first and second-block amplitudes differed between migraineurs and controls.

The N2P2-amplitude decreased from the first to the second block in controls, and the decrease was not different between controls and migraineurs (Fig. 4 and Table 5). Post-hoc contrasts confirmed a significant habituation in migraineurs (95% CI [−8.07, −2.48] μV). Overall amplitudes were not different between groups (95% CI [−9.87, 5.11] μV).

The linear change in pain scores was not significantly different from zero in controls (95% CI [−0.06, 0.27] NRS-change/10 stimuli) and did not differ between migraineurs and controls (95% CI [−0.09, 0.35] NRS-change/10 stimuli, Table 5). Pain thresholds and stimulation intensities were not different between groups (Student’s t-test 95% CI [−0.37, 0.34] and [−0.04, 0.44], respectively).

The major strengths of this study are its relatively large size with a longitudinal design and rigorous blinding both during data collection and analysis. The level of arousal, attention, and distraction may affect LEPs [60, 61]. Therefore, within-study consistency of the laser stimulation procedure is important. Especially when comparing groups, blinding of the investigators performing the stimulations becomes a necessity. Unfortunately, none of the previous migraine LEP-studies reported blinding of the investigators during stimulation, although the majority analyzed the LEPs blind to diagnosis (Table 6).

The solid-state laser used in this study differs from the CO₂-lasers employed in previous studies in that it produces a laser beam with shorter wavelength with deeper skin penetration that activates nociceptors more directly. This increases the amplitude of N1 and N2 and shortens the latency of N2P2, but the distribution of brain generators remains equal [80]. We do not believe that these differences, or the subtle differences in target-intensity, can explain the discrepancy between previous and present results.

The longitudinal design ensured a substantial number of interictal and preictal measurements and an acceptable number of ictal measurements. The postictal estimates are the least reliable due to the lowest number of measurements in that phase [77], although the number is comparable to previous migraine LEP-studies [21, 23, 24]. Only first-session responses were included in the migraineur versus control analyses because the investigator could not be blinded to diagnosis for the subsequent sessions, and to avoid possible long-term habituation/sensitization effects in later exams. Nevertheless, the number of interictal responses in this study was equal to [17, 22] or considerably larger than in the previous studies whose findings we attempted to reproduce [18‐21].

We always stimulated the right hand regardless of the side the migraineur predominantly experienced headache. We found no habituation differences of LEPs obtained ipsilateral and contralateral to a migraine headache in accordance with previous findings [17]. Hence, it is seemingly not necessary to adjust the stimulated side according to headache laterality. We did not collect information on clinical allodynia, which could be of importance as an explanatory variable. We recruited both migraineurs and controls from the general population, and this design may enhance the generalizability of our results to the standard migraine population [81]. Having a first-degree relative that suffers from migraine may influence the habituation in controls [82]. However, we found habituation in both groups, not lack of habituation, which would be the expected finding if migraine-related genes biased our control group. Also, only four of the controls in our study had a positive family history of migraine, and excluding them from the analyses did not change the conclusions (results not reported).

The use of symptomatic treatment may have influenced the results as both triptans and non-steroidal anti-inflammatory drugs may reduce the amplitudes during the ictal phase [83]. It is unlikely that the medication influenced other phases than the ictal phase due to short half-life. Lack of facial stimulation can also be considered a limitation. However, hand and face LEP habituation seems to agree quite well in previous studies (Table 6), and this is to be expected as LEP reflects activation of a large part of the bilateral cortical pain network, and our aim was to study the generalized effects. Also, other modalities like pain thresholds have shown abnormalities in hands (and face) [84‐86], suggesting an eventual thalamocortical dysfunction in migraine, in line with the development of cutaneous allodynia demonstrated by Burstein et al. [87]. Nevertheless, the present results are valid only for the more global pain function in migraine. It is necessary to do a similar study with face stimulation to conclude about lateralized second order trigeminal medullar afferent sensitization can be detected by LEP-abnormalities.

Previous studies have calculated habituation differently (Table 6). The method we chose included all available data and estimated the amplitude-change without prior calculation/manipulation of the dependent variable. Also, the approach did not use listwise deletion of cases with missing values, as would be the case with ANOVA. We were thus able to compare all four phases in one model. We also included the N1-responses where the signal to noise ratio was too low, as interval censored variables instead of discarding them, to avoid exclusion bias [65].