Group comparisons across all study phases were significant for headache frequency and severity, ictal and interictal photophobia (p < 0.001, Quade analysis of covariance), with significantly higher values for the patients compared to the controls (Table 1). Interictal photophobia was significantly higher in the patients compared to the controls in all study phases regardless of the intervention. During the Flash intervention and in the Post Dark period headache frequency and severity as well as ictal photophobia converge between groups, meaning that no significant effect of group was found here. In all other phases (Baseline, Dark, Post Flash) these variables were significantly higher in patients. Due to the significant group differences, the effect of study phase was analyzed for both groups separately, using non-parametric Friedman tests. At baseline, mean interictal photophobia was 2.59 (SD = 1.82) in the patients, while the controls scored close to 0 on average (mean = 0.06, SD = 0.18; Table 1). Interictal photophobia remained low in all study phases in the controls (< 0.02 on average) mirrored by a non-significant effect of study phase (p = 0.406). In the patients, interictal photophobia scores dropped during and after the Dark intervention, while the score remained high during the Flash intervention but dropped afterwards, which is also obvious in most patients individually (Fig. 2a). Accordingly, the Friedman test revealed a significant effect for study phase in the patients (p = 0.026) with a significant post-hoc comparison between Flash and Post Dark (p = 0.050 FDR-corr.). Further, we observed trends for lower values for the Post Dark vs. Baseline (p = 0.048 uncorr., p = 0.120 FDR-corr.), for Dark vs. Flash (p = 0.034 uncorr., p = 0.120 FDR-corr.) and for Post Flash vs. Flash (p = 0.048 uncorr., p = 0.120 FDR-corr.).

Mean interictal photophobia (a) and mean headache frequency (b) at baseline, during the intervention (Flash or Dark) and post intervention, plotted for all patients (P01-P10) individually. Interictal photophobia scores dropped after the Flash and after the Dark intervention in most of the patients. Regarding headache frequency, individual responses to the interventions were heterogenous but returned to baseline levels after the interventions

The effect of study phase was not significant for headache frequency (patients: p = 0.261; controls: p = 0.722), headache severity (patients: p = 0.593; controls: p = 0.348) and for ictal photophobia (patients: p = 0.717; controls: p = 0.406). One reason for the insignificant results of these symptoms is probably the high variability in the individual responses to the interventions, as illustrated for headache frequency in Fig. 2b. Here, headache frequency increased in four patients, but dropped in three other patients during the Flash intervention compared to the baseline (three patients did not report headache at all). During the Dark intervention 7 patients reported increased headache compared to the baseline, while only two reported symptom decrease (one patient remained stable). After both interventions, headache frequency approximated baseline levels again.

Out of the 9 female patients, 7 noted a menses in the diary (one patient twice), on 23 days for all patients in total. As migraine tends to be more severe two days before and during menstruation [22], headache occurrence was calculated for this time (39 days in total) and the remaining period (276 days). As headache was present on approximately 18% of the days before or during the menses and on approximately 20% of the remaining days, no significant modulation of headache frequency by the menstrual cycle was observed here. In total, patients reported headache on 76 days and used standard abortive medication (e.g., acetaminophen, triptans) in approximately half of the cases (on 37 days). The occurrence of headache varied considerably between the patients, for example, P01 reported headache on 22 days (9 with medication), P06 on 14 days (12 with medication), whereas P08 and P09 noted headache on two days only.

At baseline, both groups display widespread activation in occipital areas encompassing posterior temporal regions and in the lateral geniculate nucleus of the thalamus bilaterally (Fig. 3). In the controls, but not in the patients, right-hemispheric dominant fronto-parietal activation was observed as well. A contrast between patients and controls revealed significantly higher activation in the cerebellum and in the calcerine cortex in the patients while the activation in the left angular gyrus was lower compared to the controls. For the Flash intervention, the only significant difference was found in the anterior insula, with significantly higher activation in the patients compared to the controls. No significant effect of group was found for the Dark intervention. In Table 2 the significant cluster location with cluster size, MNI coordinates and highest T values are listed for the sessions separately and for contrasts between them.

Functional activation induced by flickering light in patients and controls at baseline, and after Flash and Dark intervention, contrasted to the baseline (FWE 0.05 corr., cluster threshold = 10). At baseline, patients display higher activation in the primary visual cortex than controls. When contrasting Flash vs. baseline, patients show higher activation in the bilateral anterior insula compared to controls, while no difference was found for the contrast Dark vs. baseline

Compared to the baseline measurement, patients showed higher activation in the Flash condition in the left insula and in bilateral inferior parietal areas, while activation was lower in regions dedicated to visual perception (right superior occipital, right fusiform gyrus and in the right posterior middle temporal lobe (visual area MT, Table 3). In the Dark condition, patients displayed higher activation in the right middle occipital gyrus, in the left supramarginal gyrus and in the right superior frontal gyrus and showed lower activation in several occipital regions and in the cerebellum compared to the baseline. In the controls, no area was found to be significantly higher in the Flash condition compared to the baseline. Lower activation was observed in several occipital regions, in the right anterior insula and in the right inferior frontal gyrus. In the Dark condition, controls displayed lower activation in the right inferior frontal gyrus compared to the baseline while the right superior frontal gyrus was significantly higher activated compared to the baseline. A direct comparison between patients and controls revealed significantly higher activation in the bilateral anterior insula in the patients for the Flash condition, while no significant difference between the groups was found for the Dark condition (Table 3).

Over all participants and sessions, the habituation slope of the beta values within the V1 ROI, as an estimate of primary visual cortex activation, is characterized by the regression:

with the initial activation amplitude of a = 1.2 (= average group intercept) and the uncorrected regression slope b = -0.02. The regression analysis of all data with b as dependent variable and a as predictor resulted in the regression slope c = -0.044 that was used to calculate the corrected individual regression slopes b’. The average a and b’ for both groups are plotted in Fig. 4. Repeated measurements ANOVAs with group as between-subject factor and time as within-subject factor for the corrected habituation slopes (b’) and the initial amplitude estimate (a) as dependent variables revealed no significant effects for b’, but for the initial amplitude a significant effect of condition (p = 0.002) and a significant interaction between condition and group (p = 0.042) was found. For both groups together, post-hoc comparison showed a significantly higher initial amplitude for the baseline compared to the Flash condition (p = 0.018, FDR-corr.) and compared to Dark (p = 0.027, FDR-corr.). For both groups separately, the comparisons showed significantly higher initial amplitude at baseline compared to Flash (p = 0.045, FDR-corr.) and compared to Dark (p = 0.045, FDR-corr.) in the patients. In the controls, only the comparison between baseline and Flash showed a trend for a lower initial amplitude (p = 0.081, FDR-corr.).

Initial activation amplitude (a) and corrected habituation slope (b’) in the patients and controls for the baseline, Flash and Dark intervention. Patients exhibited a significantly higher initial activation amplitude at baseline compared to Flash and Dark

Over all sessions and groups (N = 60), correlation analyses between diary variables and habituation slopes in the primary visual cortex revealed a significant correlation between the initial amplitude estimate a and headache severity (rho = 0.308, p = 0.034, FDR-corr.) and ictal photophobia (rho = 0.348, p = 0.024, FDR-corr.). In addition, b’ was significantly correlated with headache frequency (rho = 0.378, p = 0.008, FDR-corr.) and severity (rho = 0.371, p = 0.008, FDR-corr.) as well as with ictal photophobia (rho = 0.289, p = 0.033, FDR-corr.). Analyzing patient data only (N = 30), resulted in even higher correlations (Fig. 5). The initial amplitude a significantly correlated with headache severity (rho = 0.469, p = 0.034, FDR-corr.) and ictal photophobia (rho = 0.432, p = 0.034, FDR-corr.). The corrected habituation slope b’ correlated significantly with headache frequency (rho = 0.651, p = 0.000, FDR-corr.), headache severity (rho = 0.538, p = 0.004, FDR-corr.) and with ictal photophobia (rho = 0.468, p = 0.012, FDR-corr.). For the controls (N = 30) no correlation turned out to be significant.

Correlations between primary visual cortex (V1) activation habituation estimates and diary variables in the patients. The initial amplitude was significantly correlated with headache severity and ictal photophobia, while the corrected habituation slope was significantly correlated with headache frequency and severity, and with ictal photophobia

We found high variability in the individual course of the corrected habituation slope b’, particularly in the patients (Fig. 6). In relation to the baseline, 6 patients showed an increased b’ (P01, P03, P05, P08, P09, P10) and 4 a declined b’ (P02, P04, P06, P07) in the Dark condition. When comparing Flash to the baseline, an increased habituation slope was found in 6 patients (P01, P02, P05, P06, P08, P10) and a decrease in 4 patients (P03, P04, P07, P09). Interestingly, all patients reporting headache more frequently during the Flash intervention compared to the baseline (P01, P02, P05, P06, compare Fig. 2b) also revealed an increased habituation slope in the Flash compared to the baseline. Two further patients (P08, P10) showed an increased habituation slope, but had no difference in the headache frequency (no headache reported in the baseline and in the Flash condition). In line with these observations, all patients with a decreased headache frequency in the Flash intervention (P03, P04, P07), revealed a decreased habituation slope in the Flash compared to the baseline (as well as P09 who did not report headache at baseline or during the Flash intervention).

Corrected habituation slope (b’) for the primary visual cortex (V1) in all migraine patients individually. In the Flash condition, an increased habituation slope was found in 6 patients (P01, P02, P05, P06, P08, P10) and a decrease in 4 patients (P03, P04, P07, P09) compared to the baseline. In the Dark condition, 6 patients showed an increased b’ (P01, P03, P05, P08, P09, P10) and 4 a declined b’ (P02, P04, P06, P07) relative to the baseline