BIOMIGA: Results From Magnetic Resonance Imaging

The Hamburg data was processed as stated below and will be used as independent data set for hypotheses generation. The data of the two other sites (Italy and Spain) will be processed identically to the Hamburg data set:

rs-FMRI

Preprocessing of the resting-state functional MR data followed the SPM12 pipeline (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/) using slice time correction, realignment, coregistration to the structural image, normalization into MNI space, and smoothing with an 8mm^3 Gausian kernel. Data was further analyzed using the CONN-toolbox (https://web.conn-toolbox.org/) where a 0.003-0.08 Hz temporal filter and denoising with linear regression of confounding effects (white matter, CSF, movement) and linear detrending. We choose:

• Local Homogeneity (LCOR),
• Global Correlations (GCOR),
• Seed-Based Correlations (SBC): Seed to voxel and ROI to ROI with an additional mask of the hypothalamus
• Independent Component Analyses (group-ICA),
• Amplitude of Low-Frequency Fluctuations (ALFF)

Statistical comparisons calculated are defined as primary and secondary outcomes below and include (i) prediction of treatment outcome (reduction in headache days) from T0 (further PR), (ii) alterations between T0 and T1, (iii) differences between healthy controls and migraine patients at T0 (further HvsPAT), and (iv) differences between responders and non-responder at T0 and T1.

In the hypothesis generating Hamburg data significant results at a threshold of cluster-wise FDR-corrected p<0.05 with an entry threshold of voxel-wise uncorrected p<0.001, when corrected for age and sex in were found:

• LCOR: PR: Two clusters in left lateral occipital cortex and right temporal fusiform cortex; HvsPAT: bilateral thalamus left frontal operculum.
• SBC Seed to Voxel: PR: Hypothalamus 2 clusters, Sensori Motor Lateral R 1 cluster, Visula Lateral L 1 cluster, Visual Lateral R 3 cluster; HvsPAT: Sensori Motor Lateral L 5 cluster, Sensori Motor Lateral R 1 cluster, Sensori Motor Superior 3 cluster, Visual Medial 3 cluster, Visual Occipital 1 cluster, Visual Lateral L 4 cluster, Visual Lateral R 1 cluster,
• ICA: ICA 5 2 cluster, ICA 16 2 cluster, ICA 28 1 cluster, ICA 34 2 cluster; HvsPAT: ICA 4 3cluster, ICA 6 4 cluster, ICA 14 4 cluster, ICA 18 2 cluster, ICA 27 3 cluster, ICA 29 14 cluster, ICA 33 3 cluster, ICA 36 7 cluster, ICA 38 6 cluster
• There were no significant results for SBC ROI to ROI, GCOR and ALFF
• There were no significant results for the chosen threshold for the comparisons between responders and non-responder at T0 and T1 Preliminary results for local and global connectivity were recently presented at a scientific conference.

To reproduce the significant results from the Hamburg data there is choosen a small volume-corrected threshold of p<0.05. For non-significant result, the statistical threshold for the data of the two other sites will be set to voxel-wise FWE-corrected p<0.05.

MPRAGE Raw T1-images (aka MPRAGE) were processed using the CAT12-toolbox (https://neuro-jena.github.io/cat12-help/) which extents SPM12 (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/). For controls and migraine patients with data from only the first day, images segmented into compartments of gray and white matter and normalized. For patients with data from 2 days, data was longitudinally segmented. Images of gray and white matter were smoothed with an isotropic Gaussian kernel of 6 mm3. T-test and F-tests implemented in the SPM12 toolbox were used to estimate significant differences at a threshold of cluster-wise-correct p<0.05. The individual total intracranial volume (TIV), which is also estimated by the CAT12-toolbox, was used as covariate to control for different brain sizes. Age and gender were used as further covariates when indicated. All results were masked by gray or white matter masks stemming from a segmentation of the average normalized T1s of the participants. The statistical threshold for the hypotheses generating data set from Hamburg was set to cluster-wise FDR-corrected p<0.05 with an entry threshold of voxel-wise uncorrected p<0.001. As no significant results were achieved, the statistical threshold for the data of the two other sites will be set to voxel-wise FWE-corrected p<0.05.

ASL ASL was not yet analyzed for any of the sites. For the statistics we will use Cerebral Perfusion Images stemming from the toolbox ASLtbx (https://www.cfn.upenn.edu/zewang/ASLtbx_manual.pdf). Nevertheless, as different imaging protocols were available in Spain and Italy, a comparison of the equality has to be proofen first. Therefore, we acquired both protocols in the Hamburg data. As we do not expect any differences here, the statistical threshold will be set to voxel-wise FWE-corrected p<0.05.

Power calculation based on Hamburg MR data Power analysis of the primary outcome "prediction of headache reduction with resting-state functional connectivity measures from fMRI of day 1" for which rs-fMRI data of 54 migraine patients from the hypothesis generating site Hamburg were analyzed using the CONN-toolbox revealed that 37 migraine patients are necessary to reproduce the result of significant comodulation in local correlation when corrected for age and gender (80% Power, FWE-corrected p<0.05 calculated with PowerMap). Initial quality checks of the data from the other two sites reveal enough available data for reproduction.