P3 latency as a biomarker for the complexity of migraine with aura: Event-related potential study

Abstract

Abstract

Background

This study aimed to compare the P3 component between patients who have migraines with aura and healthy subjects, and to compare different subtypes of migraine with aura relative to the complexity of migraine aura.

Methods

Average Migraine Aura Complexity Score was calculated for each MwA patient. Visual oddball paradigm was used to elicit the P3 component. P3 amplitudes and latencies elicited from frequent and rare stimuli, as well as from difference wave, were compared with healthy subjects. Subsequently, subtypes of migraine with aura were compared and Average Migraine Aura Complexity Score was used to explore the connection between features of the P3 and complexity of migraine with aura.

Results

37 patients who have migraine with aura (16 with simple aura and 21 with complex aura) patients and 28 healthy subjects were studied. Patients who have migraine with aura had significantly prolonged latencies compared to healthy subjects (411 ± 39 ms vs 372 ± 34 ms, p < 0.001) relative to a rare condition. Patients who have complex aura significantly differs from patients who have simple aura (427 ± 34 ms vs 389 ± 35 ms, p = 0.004) and healthy subjects (372 ± 34 ms, p < 0.001) relative to P3 latency in a rare condition and the patients who have complex aura significantly differs from healthy subjects (442 ± 37 ms vs 394 ± 33 ms, p < 0.001) relative to P3 latency in difference wave. P3 latency from rare condition positively correlated with the Average Migraine Aura Complexity Score (p < 0.001).

Conclusions

Visual oddball paradigm, particularly rare stimuli, could serve as a potential new tool for deep profiling of different clinical complexities among patients who have migraine with aura. Also, the present pattern of P3 components provided new evidence for the cognitive dysfunctions in patients who have migraine with aura.

Introduction

Event-related potential (ERP) is a reliable and non-invasive neurophysiological examination that can reflect underlying brain activities during cognitive processing and has been increasingly employed as a cognitive biomarker in various neurological diseases (1,2). P3, the most investigated ERP component, is considered an effective index of cerebral information processing during a cognitive task (3). In the literature, there are no consistent differences of P3 characteristics in ERP evoked by an oddball paradigm between patients who have migraine, other headache types, and healthy subjects (HSs) (4–6). Several ERP studies have assessed the cognitive function in migraine patients and showed reduced P3 amplitudes (6,7), while there was also evidence that the P3 component was enlarged and delayed in migraine patients (8,9). Moreover, P3 amplitude was significantly reduced during mind wandering relative to on-task periods in migraine patients, which is in contrast to what was observed in HSs (10). All these opposite findings could be because researchers in previous studies recruited patients with different stages of the migraine cycle or different migraine phenotypes (migraine without aura (MwoA) and migraine with aura (MwA) or even different MwA subtypes) in different proportions in a single group (11). Also, an important finding in P3 studies is the lack of habituation during an interictal phase in patients who have a migraine, where results showed an acceleration of the P3 latency during the second trial (5,9).

It is well known that some MwA patients report abundant symptomatology of higher cortical disturbances during MwA attacks (12,13). Moreover, it is suggested that MwA patients could suffer from subtle cognitive changes during the interictal period (14,15). The most frequently reported cognitive changes were impaired visual and verbal memory, reduced information processing speed, executive dysfunction and attention deficit (11,16). The presence of cognitive impairment in MwA patients reinforces the complexity of this disease, which is not exclusively associated with pain symptoms (11). In fact, recent studies also suggest different layers of MwA pathophysiology among MwA subtypes (17–20). Moreover, a migraine aura complexity score (MACS) has been proposed in a recent study which could serve for quantification of the complexity of MwA attack (21). Knowing that MwA attacks could differ from one attack to another in the same person, it is suggested to use MACSs to calculate average MACS for the evaluation of mean MwA complexity in one patient and consequently help with the stratification of MwA patients into subgroups relative to theirs MwA complexity (22).

P3 studies conducted on a homogenous group of patients who have typical migraine aura are scarce and there is no comparison between MwA subgroups in terms of migraine aura complexity by using an oddball paradigm and recording ERP. Considering the aforementioned, the present study aimed to compare the P3 component between MwA patients and HSs, as well as to compare the P3 component between MwA patients who have only visual symptoms and those who have visual and somatosensory or dysphasic aura. Also, we aimed to correlate features of the P3 component with the estimated complexity of migraine aura.

Methods

This is an observational, descriptive, cross-sectional and quantitative study, which was approved by the Scientific Ethics Committee of Clinical Center of Serbia and Neurology Clinic (reference number: 23-690). All standards and guidelines of the World Medical Association Declaration of Helsinki were respected. The participants signed a written informed consent form before participation.

Participants

Patients who had episodic typical MwA (23) were recruited between 2019 and 2020 from the migraine population referring to the Center for headaches, Neurology Clinic, Clinical Center of Serbia. Patients with neurological (other than MwA), psychiatric, cardiovascular and metabolic disorders were excluded because they did not fit the profile chosen for the study. The minimum age for participating in this study was 21 years. Also, all MwA patients did not take any migraine preventive medications at the time of the enrolment in the study. Additionally, 28 HSs, balanced with MwA by age, sex and education, and with no family history of migraine were recruited. HSs were voluntarily recruited from clinical staff or their relatives and friends, who upon acceptance underwent physical and neurological examinations. Also, all participants underwent an MRI examination to exclude intracranial lesions.

Clinical data, such as MwA attack frequency, the average disease duration and migraine aura symptoms, were obtained via patient's diary and structured electronic questionnaire which patients regularly filled after every MwA attack. From the electronic records, the last six consecutive MwA attacks before the ERP recording were used to calculate the average MACS for each patient. The range of MACS is 0–9, with higher values indicating a more complex aura (21). Average MACS was used in the study to correlate MwA complexity with the P3 component derived from an oddball paradigm. Because the MACS are still undergoing validation process, we decided to stratify MwA patients into a subgroup of patients who have simple MwA (only visual symptoms during the aura – MwA-s) and those who have complex MwA (visual symptoms plus somatosensory and dysphasic symptoms during the aura – MwA-c), which is more recognizable from the point of the International Headache Society terminology (23).

ERP study design and processing of signals

All MwA patients were both migraine-free and not taking any medications at least three days before or after ERP recordings. Every MwA patient and HS received instructions before electrophysiological recordings started. Then, they were seated in a chair with a monitor placed 60 cm in front of them. To ensure the instructions were understood and followed correctly, short subject training was performed.

Each trial started with a fixation cross in the center of the screen with a jittered time range between 300 and 700 milliseconds (ms) that varied from trial to trial. Next, the letter O (frequent stimulus; 0.8 probability of occurrence) or X (rare stimulus; 0.2 probability of occurrence) appeared, which remained on the screen for three seconds or until response. Participants were instructed to press the left mouse button for the frequent stimulus and the right mouse button for the rare stimulus. Letters appeared in black Mono 24 px font against a light gray background. There were 300 trials in total including 240 trials for frequent condition and 60 trials for rare condition. For stimuli presentation, we used OpenSesame 3.3.9 (24).

Electroencephalography (EEG) signals were recorded continuously from the scalp in monopolar setup from 35 electrode sites positioned according to the international 10/20 standard: Fp1, Fp2, F7, F8, FT9, FT10, T7, T8, F3, Fz, F4, FC5, FC6, FC1, FC2, FCz, C3, Cz, C4, CP5, CP6, CP1, CP2, P3, Pz, P4, TP9, TP10, P7, P8, PO9, PO10, O1, Oz, and O2. All electrodes were referenced to the left earlobe, and the ground electrode was positioned at the AFz location. Skin-electrode contact impedance levels were maintained below 5 kO. EEG was recorded with a sampling rate of 1000 Hz.

Offline signal processing was conducted using custom MATLAB code (Version 2015a, The Mathworks, Natick, MA, USA). EEG channels were band-pass filtered using a zero-phase 4th order Butterworth filter in a range of 0.1–25 Hz. Individual 1000-ms EEG epochs (from –100 to 900 ms), with 0 marking the stimulus, including 100 ms pre-stimulus baseline and 900 ms post-stimulus data, were extracted from the continuous filtered EEG. All EEG channels were baseline corrected by subtracting the mean amplitude of the baseline from each epoch. The trials were inspected for artifacts (eye movement, blinks, high amplitude drifts). Only the noise-free trials associated with correct subjects’ responses were included in further analyses. Data from 1 MwA patient was rejected due to the presence of noise which resulted in a high number of rejected epochs per experimental condition (>30). For each participant and each condition at each electrode site, individual ERPs were calculated by averaging all remaining trials. Additionally, the difference ERP waveforms were calculated for each subject by subtracting the average ERPs (of each channel) of the frequent condition from the averaged ERPs of the rare condition. The positive peak between 300–500 ms was used to define the P3 component. Amplitudes and latencies for the P3 component were extracted from waves elicited by frequent stimuli (P3-f) and rare stimuli (P3-r), as well as from the difference wave (P3-d).

Statistical analyses

For the analyses of demographic and clinical variables among groups, we used descriptive statistics (mean ± standard deviation and percentage), parametric test (the Independent Student T-test for age and education) and nonparametric test (the Chi-square test for sex). Also, parametric and nonparametric tests were used for the analysis of behavioral data. P < 0.05 was considered statistically significant.

The effect of the oddball paradigm was assessed with an analysis of variance with grouping factors of participant status (MwA vs HSs), experimental factors (frequent vs rare stimuli) and recording site. The recording site included two dimensions: anterior-posterior distribution and laterality. The anterior-posterior dimension grouped frontal (F3, Fz, F4), central (C3, Cz, C4), and parietal (P3, Pz, P4) electrodes. The laterality dimension grouped left (F3, C3, T3), middle (Fz, Cz, Pz), and right (F4, C4, P4) electrodes. The amplitudes and latencies of P3-f and P3-r were used for repeated analysis of variance (ANOVA) In the case of significant interactions, they were broken down following subsequent analysis in an attempt to understand the locus of the interaction. A Greenhouse-Geisser correction was used in cases of sphericity violation. Significant main effects were further explored by follow-up t-tests. For the analysis of amplitudes between groups and subgroups, we used the Mann-Whitney U test, while the Independent Student T-test was used for the analysis of latency differences. Pearson and Spearman correlation coefficients were used to examine connections between clinical, behavioral and ERP data.

Results

Clinical and demographic data

A total of 37 MwA (16 MwA-s and 21 MwA-c) patients and 28 HSs were studied. They were balanced in age (37 ± 9 years vs 36 ± 9 years, p > 0.9), sex (70% females vs 71% females, p = 1.0) and education (15 ± 2 years vs 15 ± 3 years, p > 0.9). MwA attack frequency per year was 7 ± 8 attacks. The average disease duration was 19 ± 10 years. The mean of average MACS of all MwA patients included in this study was 3 ± 2 points.

MwA-s did not differ significantly from MwA-c in age (37 ± 9 years vs 36 ± 9 years, p = 0.8), sex (62% females vs 77% females, p = 0.5), education (15 ± 2 years vs 15 ± 2 years, p = 0.7), attack frequency (4 ± 2 attacks vs 9 ± 10 attacks, p = 0.2) and average disease duration (19 ± 9 years vs 18 ± 12 years, p = 0.9). The mean of average MACS significantly differs between subgroups as expected (0.9 ± 0.7 points vs 4.6 ± 1.8 points, p < 0.001).

Behavioral data

There was a significant difference between MwA and HSs in the reaction time for frequent (330 ± 97 ms vs 265 ± 62 ms, p = 0.002) and rare (415 ± 79 ms vs 360 ± 63 ms, p = 0.003) stimuli. There was no significant difference between MwA and HSs in the number of errors for frequent (0.2 ± 0.6 errors vs 0.2 ± 0.4 errors, p = 0.9) and rare (3.0 ± 2.8 errors vs 3.2 ± 2.6 errors, p = 0.6) stimuli during the task.

Subsequently, MwA-s and MwA-c subgroups were compared to the HSs group. MwA-c significantly differs from the HSs group relative to the reaction time for frequent (p = 0.009) and rare (p = 0.029) stimuli, while MwA-s did not differ from the HSs group (p = 0.1 and p = 0.051, respectively). There was no significant MwA-subgroup difference in the reaction time for frequent (318 ± 97 ms vs 339 ± 99 ms, p = 0.5) and rare (414 ± 80 ms vs 415 ± 80 ms, p > 0.9) stimuli. Also, the number of errors during the task did not differ between MwA-s and MwA-c subgroups for frequent (0.1 ± 0.2 errors vs 0.3 ± 0.7 errors, p = 0.4) and rare (2.4 ± 2.0 errors vs 3.4 ± 3.2 errors, p = 0.5) stimuli.

ERP data

The grand averaged ERP curves of both experimental conditions, including difference wave, for all participants at the frontal (F3, Fz, F4), central (C3, Cz, C4) and posterior (P3, Pz, P4) regions are shown in Figure 1. Regarding the analysis of P3 amplitudes, repeated measures ANOVA showed the main effect of condition (p < 0.001), with higher amplitudes in a rare condition. There was no significant interaction between effect and groups (p = 0.3).

Figure 1.

The grand averaged ERP curves of frequent (black line) and rare condition (red line) with confidence intervals (dotted lines) were presented at the frontal, central and posterior channels. The blue line represents the difference wave which reveals the P3 effect.

Regarding the analysis of P3 latencies, repeated measures ANOVA showed the main effect of condition (p < 0.001), with longer latencies in rare condition. An effect × region × lateralization interaction was also detected (p < 0.001), as well as effect × group (p < 0.001), effect × region (p < 0.001) and effect ×lateralization (p < 0.001) interactions. Subsequent analysis at the Pz site revealed that the MwA group had significantly longer latencies compared to HSs (411 ± 39 ms vs 372 ± 34 ms, p < 0.001) relative to a rare condition, while P3 latencies in frequent condition did not differ between groups (344 ± 31 ms vs 345 ± 48 ms, p > 0.9). Further analysis of subgroups showed that the MwA-c subgroup significantly differs from MwA-s and HSs at the Pz site with regards to P3 latency in a rare condition and the MwA-c subgroup significantly differs from HSs with regards to P3 latency in difference wave (Table 1, Figure 2 and Figure 3).

Table 1.

Amplitudes and latencies derived from the Pz channel for the frequent and rare conditions and difference wave in the MwA-c, MwA-s and HSs.

	MwA-c (n = 21)	MwA-s (n = 16)	HSs (n = 28)	Statistics*
P3-f amplitude (µV)	6 ± 3	7 ± 3	8 ± 3	p = 0.2
P3-f latency (ms)	345 ± 36	343 ± 26	345 ± 48	p > 0.9
P3-r amplitude (µV)	15 ± 5	14 ± 5	16 ± 4	p = 0.4
P3-r latency (ms)	427 ± 34	389 ± 35	372 ± 34	p < 0.001 (MwA-c vs MwA-s, p = 0.004; MwA-c vs HSs, p < 0.001)
P3-d amplitude (µV)	12 ± 5	11 ± 4	12 ± 4	p = 0.6
P3-d latency (ms)	442 ± 37	418 ± 38	394 ± 33	p < 0.001 (MwA-c vs HSs, p < 0.001)

HSs: healthy subjects; MwA-c: patients who have migraine with complex aura; MwA-s: patients who have migraine with simple aura; ms: milliseconds; µV: microvolts; P3-d: P3 component derived from difference wave; P3-f: P3 component derived from frequent stimuli; P3-r: P3 component derived from rare stimuli. *ANOVA was used for all 3 groups analysis and T-test with Bonferroni correction for post hoc analysis if needed.

Figure 2.

EEG topography derived from the grand averaged ERPs elicited by rare task stimuli in HSs (a), MwA-s (b) and MwA-c (c). Topographic maps show mean ERP amplitude in five 20-ms windows in a range of 350–450 ms. The gradual delay of the P3 peak is observed in MwA-s compared to HS and MwA-c compared to MwA-s and HS. Topographic maps show also a trend of wider distribution and higher P3 peak in HSs compared to MwA-s and MwA-c subgroups.

Figure 3.

The grand averaged ERP curves of HSs (black line), MwA-s (blue line) and MwA-C (green line) groups derived from the rare stimuli at the Pz site.

Further, clinical and behavioral data of MwA patients were correlated with P3 latency derived from frequent and rare conditions from the Pz site. There was a positive correlation between time response for frequent stimuli and P3-f latency (Pearson correlation coefficient = 0.6, p < 0.001), as well as between time response for rare stimuli and P3-r latency (Pearson correlation coefficient = 0.4, p = 0.023). Also, there was a positive correlation between MACS and P3-r latency (Spearman correlation coefficient = 0.6, p < 0.001). There was no significant correlation between the MACS and P3-f latency (Spearman correlation coefficient = 0.05, p = 0.8). MwA attack frequency and the average disease duration did not correlate significantly with P3-f and P3-r latencies.

Discussion

Despite mixed results, most studies found that MwA patients have lower cognitive performance than HSs (11). Our results resonate with this observation. Moreover, our findings support the theory of different levels of complexity of disease in MwA patients (22,25). This is, in fact, the first study of visually evoked P3 component in the largest number of carefully clinically studied patients who have typical MwA allowing us to investigate the influence of aura complexity on interictal cognitive processing in MwA patients. In the present study, the latency of the P3-r was prolonged in MwA compared to HSs group and MwA-c subgroup had prolonged P3-r and P3-d relative to HSs and P3-r compared to MwA-s subgroup.

Neurophysiological tools, such as the P3 component, have shown promising usefulness as an indicator of certain impairments of cognitive processes (11). The P3 component in many ERP studies is obtained using the so-called oddball paradigm, in which a sequence of standard (non-target) stimuli is randomly interrupted by infrequent target stimuli (26). This kind of paradigm elicits the P3 component, located at the centro-parietal region, which seems to correlate with attention, information processing and executive functions (27). Also, there is a hypothesis that the P3 may result from the operation of neural inhibition caused by brain mechanisms intending to inhibit extraneous brain activation when cognitive processes are engaged by stimulus and task demands (2,28). Furthermore, since rare stimuli can be biologically important, it is adaptive to inhibit unrelated activity to promote processing efficiency thereby yielding large P3 amplitudes (2). Regardless of these numerous functions of the P3 component, it is widely accepted that P3 latency reflects the length of stimulus evaluation processes when a two-choice reaction time is required (29) and its amplitude is largely determined by the amount of attention allocated to the stimulus (30).

Despite many electrophysiological studies of MwA (25), there is no sufficient evidence to characterize the P3 component in MwA patients. This is because most previous studies did not analyze MwA patients separately from MwoA patients. Furthermore, most of the published studies investigated visual evoked potentials focusing on early sensory and perceptual components to search for brain signatures associated with MwA (31,32). Since those visual evoked potentials provide a method of detecting only sensory but not cognitive processing, they cannot be compared to our findings. Evers et al. (9), investigating cognitive processing in primary headache, showed that MwA patients had increased P3 amplitude and longer latency compared to healthy controls. Moreover, they found an acceleration of the P3 latency during the second trial, which indicated a loss of cognitive habituation in MwA patients (9). We did not demonstrate significant changes in the P3 amplitude, although MwA patients had a trend towards decreased P3-f and P3-r amplitudes, as well for P3-d with exception of MwA-c patients, who had trends toward increased P3-d amplitude compared to MwA-s patients and HSs. However, in the previously mentioned study (9), investigators instructed participants to press a button whenever the red light occurred on the screen (15%) and to ignore the white light (85%), which is different from our paradigm and can influence comparison to our results. On the other hand, our results showed significantly increased P3-r latency in MwA patients, which is in line with the previous results and can suggest the prolongation of the cognitive processing time in MwA patients when facing rare stimuli (1,33). This is further confirmed with prolonged reaction time for both (frequent and rare) stimuli in MwA patients, which is in line with previous findings (9).

There were significant correlations between reaction time for frequent stimuli and P3-f latency and reaction time for rare stimuli and P3-r latency, which is expected because reaction time is influenced by cognitive processes and evaluation of the stimuli, as well as by selection processes and activation or execution of the motor response (34). Our behavioral data analysis revealed significant differences between MwA-c and HS groups for reaction times to both frequent and rare stimuli. However, P3 latencies differed significantly between the MwA-c and HS only for rare but not for frequent stimuli. This reveals group differences in cognitive processing of stimulus type (frequency), reflected in P3-r latency, but not reaction time measures. Thus, our behavioral findings could suggest impaired activation and execution of the motor response in MwA patients in general, but prolonged P3-r and P3-d latencies point out cognitive dysfunction during stimulus evaluation. Moreover, reaction time did not differ between MwA subgroups, while P3-r latency was significantly prolonged in the MwA-c relative to the MwA-s subgroup, suggesting more impaired cognitive dysfunction during stimulus evaluation in the MwA-c subgroup. This reveals that P3-r and P3-d latencies induced within the visual oddball task are sensitive measures for characterizing migraine complexity unlike behavioral outcome measures in our study which further confirms its biomarker potential.

To our knowledge, an ERP study focusing on the P3 component, that investigates subgroups of MwA patients has never been done. Thus, we are first to report that P3-r and P3-d latencies are prolonged in MwA-c patients compared to MwA-s patients and HSs. Moreover, the MACS positively correlated with the P3-r latency, suggesting P3-r latency as a promising electrophysiological biomarker for evaluation of the MwA complexity and reliable tool for further investigation of underpinnings of multilayered MwA pathophysiology. It is also worth noting that MwA-c and MwA-s groups did not differ in age, since ERPs are age-dependent (35). Altogether, these results suggest subtle abnormalities in attentional processing in MwA patients, particularly linked to the complexity of MwA attacks, and indicate that further studies should focus on understanding the impact of MwA on everyday life in patients during the interictal period (36). Also, knowing that MwA patients frequently experience interictal cognitive difficulties such as heightened sensitivity to extraneous sensory inputs (37,38), supported by the results which point to an increase in grand-average neural response to any kind of sensory stimuli due to deficient short-term and long-term adaptive processes to external stimuli (10), prolonged latency of P3 component, in the light of inhibitory theory, may suggest abnormalities in the compensatory strategy for reducing stimulus overload in the cortex (10). If this hypothesis is valid, this explanation can elucidate the positive correlation between P3 latency and MwA complexity. This can be further strengthened by the result that stratification of MwA patients according to the distinctive manifestations in typical aura pointed to the same finding, which challenges the point of view that patients who have only visual symptoms and someone who has visual and somatosensory or dysphasic aura should be equally weighted and placed in the same group, as already discussed in the previous neuroimaging study (22). Furthermore, the frequency of MwA attacks and disease duration did not seem to influence cognitive performance and ERP parameters, which is also found in other studies of migraine with aura (9,39).

In closing, we would like to restate the procedural decisions that can somewhat constrain the interpretation of the present findings. First, this study is limited by the lack of comparison with MwoA patients, thus our results could not be interpreted as specific for MwA patients. Second, the inclusion of fMRI findings would significantly improve the interpretation of our results, which we hope will be a future step for confirming our ERP study. Third, we did not investigate the habituation of latencies in MwA patients due to the lower number of rare stimuli used to evoke the P3 component. Our study aimed to explore the potential of the P3 component as a clinically applicable biomarker of the MwA and its complexity. We, therefore, aimed to apply a simple and fast protocol of visual oddball paradigm with an optimized number of stimuli per experimental condition. The strength of the study is that MwA patients were carefully divided into homogenous groups according to their clinical phenotypes and they did not have any comorbidity, nor did they take prophylactic treatment, such as Topiramate, which could influence the study results (40). Finally, the results of this study should be confirmed using a new and independent cohort of subjects.

Conclusion

P3 component of ERP as a response to rare stimuli within the visual oddball paradigm is a promising tool for investigation of cognitive processes in patients who are affected by numerous and varied clinical MwA features and can help in more deep profiling of different clinical complexities among MwA patients. Overall, the present pattern of P3 components provided new evidence for the dysfunction of cognitive function in MwA patients. Finally, we strongly believe that a better characterization of clinical and electrophysiological phenotypes of MwA will lead to novel, individually oriented and tailored therapeutic interventions.

Clinical implications

The P3 component could serve as an auxiliary biomarker in distinguishing patients who have complex aura symptoms from those who have simple visual auras.

The visual oddball paradigm, particularly rare stimuli, could be a promising tool for further investigation of cognitive processes in brains who are affected by numerous and varied clinical features of migraine with aura.

Combined characterization of clinical and electrophysiological phenotypes of MwA patients could point to novel and individually oriented therapeutic treatment.