This meta-analysis protocol complies with the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) 2015 statement,[35] and will have been registered on PROSPERO (https://www.crd.york.ac.uk/PROSPERO/) at the beginning of the study (registration number: CRD42021257300).
Selection criteria
Inclusion criteria
The published studies will satisfy the following inclusion criteria: (1) peer reviewed, (2) patients diagnosed with migraine according to the International Classification of Headache Disorders (ICHD), (3) original rs-fMRI studies with ReHo and FC analysis, (4) compare migraine patients with HC by whole-brain seed-based FC, and (5) written in English. If the peak coordinates are not provided in the full-text or appendix material, the corresponding authors will be contacted for this information.
Exclusion criteria
The exclusion criteria are as follows: (1) no seed-to-whole-brain FC studies, (2) no comparative HC group, (3) comorbidity or secondary with other diseases, (4) other types of migraine (e.g., vestibular migraine), (5) the results of significant effect cluster coordinates missing stringent correction information (e.g., p < 0.05), (6) duplicate studies, (7) systematic reviews, and (8) case reports. The detailed screening process is exhibited as a flow diagram in Figure 1.
Search method
The following terms will be searched in the title, abstract, and full-text: “migraine” OR “migraine with aura” OR “migraine without aura” OR “chronic migraine”, AND “magnetic resonance imaging” OR “functional magnetic resonance imaging” OR “fMRI”. In addition, keywords such as “functional connectivity” OR “regional homogeneity” will be used to screen the articles in the online databases (i.e., PubMed, Web of Science, Cochrane library, and Medline) published from inception to June 1, 2021. The reference list of retrieved studies will also be screened to include more feasible studies (Table 1).
Quality evaluation
At present, there is no consensus on the quality evaluation of neuroimaging research. In this study, the quality of included studies will be assessed with the Newcastle-Ottawa Quality Assessment Scale (http://www.ohri.ca/programs/clinical_epidemiology/oxford.htm) as published in previous meta-analyses.[36, 37] (Table 2)
The quality score of each study will be assigned as high (10-12), medium (5-9), or low (0-4). Two investigators will evaluate the quality of all included studies independently based on this scale. Any disagreement will be resolved by consensus with a third investigator (CZH).
Data collection
Two investigators (CYL and XQ) will extract data from all included studies. Any uncertainty or disagreement will be settled through consensus with a third investigator (CZH). The extracted data will comprise all authors’ names, publication year, sample size, demographic and clinical characteristics of subjects, as well as the main findings. The parameters of the MRI scan, seed point, statistical correction methods, and the main findings will also be included.
ALE meta-analysis
The Brainmap Ginger ALE 3.0.2 software (http://www.brainmap.org/) will be used to perform the coordinate-based meta-analysis. The algorithm uses the kernel method to evaluate the spatial uncertainty and convergence of coordinates, treated as a tridimensional Gaussian probability distribution, of the included individual articles. Then, a random-effects model of ALE will be used to assess the consistency of reported coordinates across studies in order to correct for the within-experiment and within-group effects.
The modeled activation (MA) maps will be generated using reported coordinates in the Montreal Neurological Institute (MNI) space. When computing the MA maps, Gaussian kernel smoothing will be applied to coordinates based on the sample size.
The 3D probability distribution centered on the coordinate MA score will be the activation foci. Then, the union probability of the union ALE map will be calculated using individual MA maps. The actual convergence of the activation foci will be determined with random spatial clustering.[38, 39]
A cluster-level family-wise error (FWE) with a corrected threshold of p < 0.05[38] and an uncorrected threshold of p < 0.001 for the initial cluster will be used for multiple comparison correction, and permutation tests will be repeated 5000 times. The balance between the sensitivity and specificity will be provided with the cluster-level FWE thresholding. Talairach coordinates will be translated to the MNI coordinates using the Ginger ALE software.
Subgroup or meta-regression analysis
When enough studies (>10) are included,[34] the duration of the disease, the frequency of the attacks, the types of migraine with or without aura or chronic migraine, the age or gender of patients, and the methods of the trials will be used to divide patients into different subgroups. The potential heterogeneity of the meta-analysis will also be analyzed.[40]
Sensitivity analysis
To test the replicability of the results, we will perform a sensitivity analysis using the leave-one-out method. An identical analysis will be applied to all remaining studies, with only one study being removed. Then, the process will be repeated with another study being removed. The advantage of this method is that it can evaluate the stability of the meta-analysis outcomes and eradicate the influence from specific outlier studies. Overall, the sensitivity analysis improves the credibility and repeatability of the results.