Participants
Forty patients with MwoA were enrolled from outpatient in department of neurology or acupuncture at the Shuguang Hospital affiliated to Shanghai University of Traditional Chinese Medicine. Thirty-six aged-matched, education level-matched with patients and right-handed healthy controls (HCs) were recruited. All participants signed written informed consent. This trial was approved by the Ethics Committee of the Shuguang Hospital affiliated to Shanghai University of Traditional Chinese Medicine and is registered on www.chictr.org.cn (ChiCTR1900023105). The diagnosis of MWoA was established according to the International Classification of Headache Disorders, 3rd Edition ICHD-III criteria [1]. Inclusion criteria required that all patients: 1) were 18-65 years old and right-handed; 2) had two to eight times of migraine attacks during the past month; 3) had at least six months of migraine history; and 4) had no prophylactic headache medications during the past month, had no psychoactive or vasoactive agents during the last three months. The exclusion criteria included the following: 1) suffered from other type of primary or secondary headache; 2) had a history of head trauma or brain tumor; 3) had any other neurological or psychiatric disorder; 4) were pregnancy or breast-feeding; 5) had MRI or acupuncture contraindications.
Study design
The total observation period for Patients with MwoA of this study was ten weeks. Weeks 1 to 4 served as a baseline phase,and all patients had recorded headache diaries at baseline. Weeks 5–10 served as an intervention phase. During this period, patients with MwoA standard acupuncture treatment. All the patients maintained a headache diary during the study period. FMRI scans were administered before and immediately after the first and 12th acupuncture sessions for patients with MwoA (all fMRI were scanned within 1 h before and after acupuncture). All patients with MwoA had been migraine-free for at least 72h at the time of the fMRI scans. HCs group only received the baseline MRI scan (Fig. 1).
Acupuncture treatment
In our study, the patients with MwoA were performed 12 sessions of acupuncture (twice a week, finished in 6 weeks), and every session lasted for 20 minutes. Acupoints were selected according to the standardized acupuncture protocol: Baihui (DU20), Taiyang (EX-HN5), bilateral Fengchi (GB20), Shuaigu (GB8), Xuanlu (GB5), Toulinqi (GB15), Hegu(LI4), and Taichong (LR3) [4, 20]. Two licensed acupuncturists (Wang B and Liu S) were responsible for all the acupuncture treatments. Sterile disposable acupuncture needles of 25-40 mm in length and 0.25 mm in diameter were inserted to achieve the sensation of deqi. Electrical stimulation was applied bilaterally at GB20 and GB8 at a frequency of 2Hz and intensity ranging from 0.1 to 1.0mA until the patient felt bearable. All participants agreed not to take any conventional medication for migraine during the study period. In cases of severe pain, ibuprofen (as 300 mg extended-release capsules) was allowed as a rescue medication.
Clinical assessments
During the four weeks before the first fMRI scans and after all the acupuncture sessions, the frequency of migraine attack (days/month), VAS (0–10 scale, 10 being the most intense imaginable pain), Self-Rating Anxiety Scale (SAS), Self-Rating Depression Scale (SDS) and MSQ were assessed. Adverse events associated with acupuncture, including bleeding, subcutaneous bleeding, severe pain, fainting and local infection, were recorded at each treatment.
Data acquisition
MR scans were acquired on a 3.0-T MRI scanner (uMR780 Platform, United Imaging Medical Systems, Shanghai, China) with an 12-channel flexible head coil at the Shuguang Hospital MRI Center. The rest fMRI images were obtained axially by a multislice gradient-echo echo-planar imaging (EPI) sequence and the parameters were as following: repetition time (TR)=2000 ms, echo time (TE)=30 ms, flip angle=90°, field of view=240×240mm2, matrix=64×64, 33 contiguous slices with 3.5 mm slice thickness, 240 time points. Structural images were acquired by a three-dimensional turbo fast echo (3D-TFE) sequence with voxel size of 1mm3, and the parameters were as following: TR=7.2ms, TE=3.1ms, slice thickness, 1.0mm, flip angle=10°, field of view=256×256mm2, matrix=256×256, 176 slices without interslice gap. A cushion was placed into the coil to fix the head and reduce motion.The participants were instructed to keep still with eyes close, to relax but not to fall asleep, and to try not to think about anything.
Data preprocessing
FMRI data preprocessing was performed by DPABI software (http://www.rfmri.org/) in MATLAB. The preprocessing course consisted of the following steps: 1) the first 10 images were discarded and the remained 230 images were used for data analysis; 2) slice timing correction; 3) head motion correction (the translation or rotation motion in any given data did not exceed 2.0mm or 2.0°); 4) The co-registered functional images were spatially normalized to the Montreal Neurological Institute (MNI) space and resampled to 3-mm cubic voxels; 5) linear trend removal was performed to reduce the effect of low-frequency drifts; 6) nuisance covariates regression (the white matter signal, the cerebrospinal fluid signal and 24 head motion parameters; 7) lintemporal band-pass filtering at a frequency band of 0.01–0.08 Hz. After these head motion controls, eleven of the subjects (three HCs and eight patients with MwoA) were excluded.
Dynamic ALFF analysis
The dynamic ALFF (dALFF) for each participant was performed by DynamicBC (v2.2 www.restfmri.net/forum/DynamicBC) toolbox. Specifically, a temporal rectangular window was firstly chosen. Then, the ALFF values in each window were calculated. Window length was an important parameter in resting-state dynamics computation. The ‘rule of thumb’ of sliding-window length is that the minimum window length should be no less than 1/fmin, fmin=0.01 Hz. Here, a window length of 50 TR was considered as the optimal parameter to maintain the balance between capturing a rapidly shifting dynamic relationship and obtaining reliable estimates of the correlations between regions [21, 22], The sliding window was systematically shifted with a step size of 5 five TR (10 s) to calculate the dALFF of each participant. The preprocessed data of each individual were segmented into 37 windows, and the ALFF map was obtained for each sliding window. Subsequently, we measured the variance of these maps using standard deviation (SD) to evaluate the temporal variability of dALFF across 37 windows. The dALFF variability of each voxel was further transformed into a z-score by subtracting the mean and dividing by the SD of global values. Finally, the mean normalized dALFF maps were spatially smoothed using an isotropic Gaussian kernel of 8 mm full-width at half-maximum.
Dynamic effective connectivity analysis
In this study, we performed seed-based dynamic Granger causality analysis (GCA) by the DynamicBC toolbox to detect the dynamic effective connectivity (DEC). The time series of the each ROI based on dALFF results was defined as the seed time series X, and the time course of voxels within the whole brain was defined as Y. A bivariate coefficient GCA to investigate the Granger causal influence between the per ROI and each voxel of the whole brain. A positive coefficient indicates that activity in region X exerts a positive influence on activity in region Y, whereas a negative coefficient indicated that the activity of region X exerted a negative influence on the activity of region Y. The dynamic GCA was estimated using the sliding window approach above, the time series of each participant was also divided into 37 windows. Thus, For each subject, the averaged time course of GCA coefficient of each ROI was extracted across 37 windows and concatenated to form a 2 × W × N matrix (where W denotes the number of windows and N denotes the number of ROIs). The DEC variability for each ROI was assessed with the SD of the averaged time course of GCA coefficient across 37 windows. Finally, the dynamic GCA coefficient maps for all subjects were then converted to z-scores by Fisher z-transformation.
Statistical analyses
Demographic characteristics were evaluated between MwoA and HCs. Differences of two groups in age and education level were analyzed with two-sample t test; χ2 test was used for analyzing the difference of gender in two groups. Two-sample t -tests or Mann-Whitney U test was used to compare differences in the clinical variables between the two time points. P < 0.05 existed statistical difference.
Two-sample t-tests were performed to compare dALFF variability maps between MwoA at baseline and HCs within a gray matter mask with age, gender, education and head motion as covariates. The resultant T-maps were corrected for multiple comparisons using the Gaussian random field (GRF) theory (voxel p<0.001, cluster p<0.05, two tailed).
To find the differences effect of acupuncture during the different periods of treatment, we first performed repeated-measures one-way ANOVA to investigate the dALFF variability among the different periods. The SD value of each brain region with significant difference between groups was extracted for statistical analysis in SPSS version 25.0 (SPSS, Inc., Chicago, IL, United States), and post hoc t-tests were performed to detect differences of dALFF variability among two periods (false discovery rate corrected, P < 0.05).
For group level analyses on DEC of the ROIs, the SD values of Zx→y and Zy→x dynamic GCA coefficient maps were calculated for each group. These maps were entered into repeated-measures one-way ANOVA to determine the difference among the different periods with age, sex, and education level included as covariates. Multiple comparison correction was performed based on Gaussian random field theory (GRF, voxelwise p < 0.001, cluster-wise p < 0.05, two-tailed). Post hoc t-tests were performed to detect the differences in DEC variability between two periods (false discovery rate corrected, P < 0.05).
Finally, the SD value of the dALFF variability and DEC variability in regions with significant differences in each MwoA individual were extracted, Based on these regions, Pearson/Spearman correlation was analyzed to probe the relation of alterations in dALFF variability/DEC variability to the clinical data of MwoA. The significance was set at a threshold of p < 0.05 using Bonferroni correction.
Validation analyses
To validate the main findings of dALFF variability and DEC variability obtained from sliding-window length of 50 TR, we carried out auxiliary analyses with different sliding window lengths(30 and 80 TR).