Enhanced learning in migraine: False positive finding or discovery of a new migraine trait?

In the publication Enhanced audiovisual associative pair learning in migraine without aura in adult patients: an unexpected finding, Tót et al. report better associative learning in migraine patients than controls (1). In this interesting and novel study, the authors applied a well-developed and standardized paradigm to investigate cognitive performance in migraine. The findings indicate enhanced learning in migraine, and the authors speculate if the sensory hypersensitivity characteristic of migraine could explain this finding. However, the study also has some limitations, that are well-known in the field of migraine physiology research.

When comparing subjects with and without migraine, it is important that these group do not differ considerably in other aspects other than the diagnosis. Tót et al. did not include controls specifically for this study but retrieved control measurements from an existing database (1). This approach, in which the same controls are used in multiple studies, has several advantages. The cost and effort required for each study are reduced (2), and, as demonstrated by Tót et al., controls matched for important factors such as age, sex and level of education, can be retrieved specifically for each study. Furthermore, this approach can allow bootstrapping-techniques to better utilize a larger database with measurements from controls (3).

There are also challenges with using pooled control data. First, randomized selection of measurements from the database may not be practically possible. Second, a blinded study design cannot be applied. Blinded procedures are rarely applied in physiological migraine research (4–6), but evidence from several other scientific fields show that lack of blinding may result in overestimation of differences between groups (7,8). Third, studies must apply the same methods as used in development of the control data. This may, at least to some extent, prevent implementation of methodological improvements and other novel, and perhaps more relevant, methods. Lastly, with changes in personnel and equipment, the use of an aging control dataset will become problematic. For instance, some of the control measurements in the study by Tót et al. are more than five years old. Arguably, measurements in controls and patients should ideally be done within the same time period and conducted by the same investigators. For instance, quantitative sensory testing, which is commonly used in migraine studies (9), is known to be influenced by training and gender of the investigator (10).

The migraine patients in the study by Tót et al. showed signs of superior learning compared to controls, in that they needed less trials to learn associations, had fewer errors and faster reaction times (1). However, these types of measurements may depend on motivation and voluntary effort. Such factors may differ in controls recruited among university students up to several years earlier, and patients recruited specifically to investigate a disease that is of great importance to them. In addition, in a study of cognitive performance, there is risk of self-selection bias, i.e., including a subset of patients that perform better on the task than migraine patients in general. These limitations may have a larger impact when comparing outpatients to control measurements in a database, than when recruiting persons with and without migraine from the same population.

Migraine is understood as brain disorder with interictal, preictal, ictal and postictal phases (11). It is well documented that these phases differ clinically (12,13), and in neurophysiological and functional imaging studies (12–19). Tót et al. aimed to investigate cognitive performance in migraine patients during the interictal phase and ensured that five days had passed since the last migraine attack (1). For the postictal phase, this is likely too restrictive. The definition of this phase varies between studies, but it has been suggested that the postictal phase may last up to 24 h (11). More importantly, Tót et al. did not exclude preictal measurements. Since 2003, our research group have conducted cross-sectional and longitudinal studies on migraine physiology, in total including over 500 laboratory visit for patients with episodic migraine (5,17–21). The patients in these studies completed headache diaries during the study period, including at least one week before and after each laboratory visit. In total, 61% of the measurements were classified as interictal, while 15%, 15% and 9% were classified as preictal, ictal and postictal respectively. A good and less resource demanding alternative to headache diaries is to call the participant a few days after the laboratory visit (16).

The migraine patients in the study by Tót et al. had from 0.2 to 12 migraine attacks/month, and it is therefore reasonable to assume that their study may include a significant number of preictal measurements (1). Symptoms in the preictal phase encompass several cognitive symptoms, including difficulties with concentration (13). For that reason, it may seem unlikely that including preictal measurements has influenced the conclusion of the study. On the other hand, alterations in sensory processing and cortical responsivity, which according to the authors may explain their findings, may be more pronounced in the preictal than the interictal phase (22,23). This example illustrates that making the effort to detect preictal measurements is important when studying the interictal migraine phase.

In summary, the study of Tót et al. provides evidence for enhanced learning in migraine, and their findings indicate a positive effect of the well-known altered sensory processing in migraine. However, as also stated in the publication, the findings are surprising and in need of replication.