In a large cohort of users of a German headache app, we observed slightly reduced self-reported stress levels during the first month of the COVID-19-related lockdown. Headache frequency and intensity remained unchanged with a small reduction of acute headache medication days in the first lockdown month. An extension analysis up to the third lockdown month revealed no significant changes in headache characteristics or lifestyle factors.
This is the first study to assess changes in headache frequency longitudinally before and after lockdown implementation using daily data from a commercial headache app. Parodi et al. evaluated through personal interviews changes in migraine severity, migraine intensity, and number of triptans per week in 49 subjects in the 2 months before and during quarantine in Italy [12]. All outcomes showed a significant improvement during lockdown. A larger Italian telephone survey also revealed a mild improvement of headache frequency and intensity as well as days with acute medication intake in 433 patients with migraine during the first lockdown month [13]. Similarly, a cohort study from the Netherlands, using a time-locked e-diary, showed a significant decrease in monthly migraine days and acute medication days during the first month of lockdown [14]. While our analysis did not detect any differences in pain frequency or intensity, the reduction of acute medication use in the first month is in line with findings from these previous studies. One possible explanation relates to the acute effects of decreased stress levels during lockdown. Changes in work and social routines might have led to a more relaxed way of living. In home office, people are usually more flexible in their time management and do not need to take acute medication immediately to “function” again. The public discussion of a potential link between NSAIDs and a negative course of COVID-19 infections could also have contributed to the slight reduction of acute medication days [15]. Of note, there was a numeric increase in the subsequent months, which might reflect habituation to the new living conditions or might also be explained by return to office / workplace with higher pressure to “function”.
All observed changes in lifestyle factors during the first lockdown month were modest and their clinical significance remains to be determined. In the extension analysis, we could not detect any significant changes anymore. This could be due to softening of the lockdown measures but could also be a result of related to the smaller sample size and thus reduced power. Importantly, our aggregated results cannot provide information on individual fates, but only an overall view of average values. While some individuals may have benefited from the lockdown measures in terms of more relax and self-care, some others might have become isolated or lost their job. A more direct influence of the disease itself through symptomatic infections is also possible, although certainly rare, as fewer than 200,000 COVID-19 cases were reported in Germany during the observation period (≈ 0.25% of the German population).
Previous research on sleep patterns during the COVID-19 pandemic is scarce: In an Indian cohort, 325 students reported an increased sleep duration during lockdown, while 203 office workers indicated reduced sleep hours [16]. Our findings are similar to the data of the Indian students. Although the differences in sleep time are small and no longer significant in the extension analysis, they are in line with reduced stress levels.
A major concern for patients with primary headache disorders in the COVID-19 pandemic may be the worsening of their condition due to reduced access to medical care [8]. However, our data does not support such hypothesis. Accordingly, a study in patients with multiple sclerosis reported improved health-related quality of life during lockdown [17]. A similar pattern may apply to headache patients, as reflected by the slightly higher levels of mood in our cohort. As the lockdown started in early spring, seasonal influences might also have contributed to mood improvement in the first lockdown month.
The reduction of physical activity due to home confinement in the pandemic has already been reported in large survey-based cohort studies [18, 19]. We detected similar reductions only in the first month of lockdown, while activity levels increased again in the subsequent months. This may be explained by the selective softening of lockdown measures (e.g. opening of gyms, increase in mobility) but also by the increased popularity of home-based exercise [20].
This study is an example on how headache data collected by an app can be used for research purposes in specific situations like the COVID-19 pandemic. Digital data collection offers the possibility of a large sample size. Nevertheless, several methodological limitations should be considered when interpreting our results. The M-sense algorithm classifies the single headache attacks, but cannot provide a final headache diagnosis. Due to the analysis of aggregated anonymous data patients could not be assessed for individually diagnosis. Previous assessments have shown that more than 70% of users have headaches compatible with the diagnosis of migraine [21]. Therefore, it is likely that our cohort consists mostly of patients with migraine but other headache disorders cannot be excluded. In addition, the app did not provide sufficient information about changes in preventive treatment or medical consultations during the observation period. Lifestyle factors were assessed using numeric rating scales, which is a simple and intuitive method, but not validated for this purpose. About 85% of M-sense users are estimated to live in Germany. The lockdown started in Austria and Switzerland (15% of users) a few days earlier than in Germany, which may have had a small impact on the data of the last baseline week. The subsequent lockdown measures were similar in all three countries. Place of residence is not collected by default in the app and subgroup analyses depending on the living place were not possible. Moreover, regional differences in continuation of lockdown measures after the first month could not be controlled for. However, considering that the majority of restrictions remained in place during the entire observation period, this is unlikely to have had relevant effects on our results. Due to the particular analysis procedure with only aggregated, anonymized data for predefined endpoints, subgroup analyses or timeline extensions were not retroactively possible. Further studies should assess which patients are more affected by lockdown measures and on which factors this depends. A longer follow-up period with the inclusion of the subsequent lockdown phases should also be considered.