In a prospective within-subject test-retest study, we investigated working memory in migraine using neuropsychological assessment and a cognitive task with fMRI, during and between spontaneous migraine attacks. We found that patients obtained comparatively lower cognitive scores during migraine attacks and also that performance in a working memory task during the attack increased the activation of areas related to inhibitory control and pain. To our knowledge, this is the first study documenting the neural correlates of a demanding executive task performed during migraine attacks, a finding that contributes to explain patients’ complaints.
Neuropsychological assessment
Cognitive scores of both sessions were within expected ranges based on age and education. However, an overall enhancement in neuropsychological performance was evident during the second, S-interictal assessment, compared to S-ictal, which was significant in the Stroop test and non-dominant finger tapping. A comparable trend of improvement has been noted in a test-retest study using the same battery in healthy controls and interictal migraine participants, showing a significant difference only in the Stroop test17. These observed improvements may be attributed to the learning effect on the second session. While the time interval between evaluations was longer in the earlier study (80 days), the absolute enhancement observed here surpassed what was observed in the previous data. This implies that the absence of pain could have influenced the magnitude of the expected learning effect, in our sample.
fMRI
The improvement in the N-back task performance and the reduced brain activity that were observed in the second session, although not correlated, could be explained either by a learning effect or a decline in mental effort, associated to the absence of pain. A study using a 2-back task in healthy individuals with a shorter test-retest interval (14.6 days), observed a decrease in functional activity in the parietal and dorsolateral prefrontal cortex between first and second session, which was attributed to the learning effect [
30]. The observed improvement in performance that increased from 93.2 to 95.5% of hit accuracy in that study was not statistically significant and less expressive than what was found in our data. The higher magnitude of difference observed in our participants suggests that there was an additional effect of the migraine attack affecting participants’ performance in the first session. This finding supports the hypothesis that pain impairs working memory function, as suggested by a report of a negative relationship between pain severity and N-back behavioural measures in subjects with diverse pain conditions [
31]. Altogether, this data indicates the cost of sharing limited attentional resources between cognitive functions and pain.
In fact, we identified a cluster on the left orbital prefrontal cortex that was significantly more activated in the ictal session during the N-back paradigm. This area is not generally considered to be part of the working memory network, regardless of the modality (auditory or visual) in which the cognitive effort is generated [
32]. For verbal working memory, namely with the letter N-back, areas activated include the middle frontal gyrus and parietal lobe regions [
33]. Conversely, the orbito-prefrontal cortex is relevant for inhibition, impulse control and decision-making [
34]. However, it is also observed under trigeminal pain stimulation and ictal phase, along with activation of somatosensory cortices and other brain areas indicating that it can be related to pain processing [
35]. Earlier research has also reported high functional connectivity between regions associated with pain and sensory processing, such as the right thalamus and contralateral pain processing regions, including the orbitofrontal cortex [
36]. In a similar study which also compared spontaneous attacks and interictal phase in migraine without aura but not during a cognitive task, significant functional connectivity between the insula and the frontal pole was noted during painful stimulation in patients with high frequency of attacks [
37].
Alternatively, the higher activation of the left orbital prefrontal cortex observed in the ictal phase could suggest that patients are employing greater inhibitory control to perform the task.
No significant correlation was identified between the prefrontal cortex activation during S-ictal and N-back performance, depression questionnaire, disease duration, attack frequency or the clinical features of the studied attack. However, individuals reporting higher disability (according to the HIT-6) displayed higher brain activity in that region during the attack, although statistical significance was not achieved. A preceding study investigating resting state in interictal migraine without aura established an inverse correlation between the HIT-6 score and impaired connectivity in the somatosensory cortex [
38].
Taken together, these findings reinforce the concept that activity in brain regions associated to pain processing might influence the neural correlates of cognitive performance during migraine attacks.
Limitations
In an ideal study design, the migraine phases would be balanced, and session repetition should be randomized to minimize the learning bias. However, for practical recruitment issues and to improve effective participation, the researchers prioritized ictal sessions. A similar design was used in a study which reported difficulties alternating the sessions while studying brain activity during spontaneous migraine attacks [
39]. Having a healthy control group would be necessary to compare the test-retest learning and familiarity effect of the task and the context of the fMRI exam. The sample size of eleven exclusively female participants limits the generalizability of the findings. Another relevant limitation is that the scans were made in different time frames within the attack, and there is evidence that brain activation may differ according to the timing from the onset [
40].