Abstract
Introduction
Medication-overuse headache (MOH) is a debilitating consequence of analgesic use, yet its pathophysiology remains incompletely characterised. Impaired glymphatic clearance may contribute to headache chronification. We investigated whether glymphatic–structural alterations progress across the migraine–medication exposure continuum and their biomarker potential.
Methods
This cross-sectional study enrolled 149 right-handed participants: 29 MOH patients, 34 patients with migraine and moderate medication use (MM), 46 medication-naive migraine patients (MN), and 40 healthy controls (HCs). The diffusion tensor imaging along the perivascular space (DTI-ALPS), perivascular space volume fraction (PVSVF), free water, and regional brain volumes were quantified. Group differences, imaging–clinical correlations, and diagnostic utility were assessed.
Results
Right-hemispheric centrum semiovale (CSO) PVSVF increased progressively across the continuum (HC < MN < MM < MOH; linear trend p = 0.003), with elevation in MOH versus controls (Cohen's d = 0.59, p = 0.014). Hippocampal volume was reduced in all migraine groups, most in MOH (d = − 0.61, p = 0.001). DTI-ALPS and free water showed no group differences. Lower DTI-ALPS was associated with higher hippocampal free water in MOH (r = − 0.539, pFDR = 0.043). Hippocampal decline accelerated beyond approximately 11 years disease duration (exploratory). A combined right PVSVF-CSO and hippocampal volume model distinguished MOH from controls, with an area under the curve (AUC) of 0.75.
Conclusion
A graded, cross-sectional increase in perivascular space burden and reduced hippocampal volume, with right-hemispheric predominance, were observed across the migraine-medication exposure continuum. These markers may serve as adjunctive tools for MOH evaluation, meriting independent cohorts.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1007/s40122-026-00883-2.
Keywords: DTI-ALPS, Glymphatic system, Hippocampus, Medication-overuse headache, Migraine, Perivascular space
Plain Language Summary
Medication overuse headache is a condition in which headaches become more frequent because a person takes acute headache medication too often. It is common in people with migraine, and how migraine shifts into this more constant form of head ache is still unclear. The brain has a system that clears waste, partly through small fluid-filled spaces that surround its blood vessels. Researchers have suggested that poor waste clearance may make headaches more frequent, and we used MRI to look for such changes. We studied 149 adults in four groups: patients with medication-overuse headache, patients with migraine who used a moderate amount of medication, patients with migraine who used none, and people without migraine. In each person we examined several brain regions, measuring the fluid filled spaces around blood vessels, the sizes of brain structures, and two further fluid measures. Most of these measures were similar across the four groups, but two changes stood out. The fluid-filled spaces in the centrum semiovale became enlarged on the right side of the brain; more so as medication use increased. The hippocampus was smaller in all migraine groups and smallest in medication-overuse headache. The two fluid measures showed no difference between the groups. Using these two measurements could help tell patients with medication-overuse headache apart from people without migraine. They may serve as supporting markers in future, but larger studies that follow people over time are needed before they could be used in practice.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1007/s40122-026-00883-2.
Key Summary Points
| Why carry out this study? |
| Medication-overuse headache is a disabling condition that develops when people with frequent headache-overuse acute pain medication, yet the brain changes that accompany it remain poorly understood. |
| Because impaired waste clearance by the brain's glymphatic system has been proposed as a contributor to headache chronification, this study used multimodal MRI to assess glymphatic function, perivascular space burden, free water, and brain volumes across several brain regions, spanning the spectrum from migraine to medication-overuse headache |
| What was learnt from the study? |
| Among the brain regions examined, perivascular space burden was increased specifically in the centrum semiovale, and this right-sided elevation rose progressively from healthy controls through medication naive migraine and moderate medication use to medication-overuse headache. |
| Hippocampal volume was reduced across all migraine groups and was lowest in medication overuse headache, whereas the DTI ALPS index and free water content showed no significant differences between groups. |
| A model combining right centrum semiovale perivascular space burden and right hippocampal volume identified medication-overuse headache with moderate accuracy (AUC 0.75), and these structural measures may serve as adjunctive markers once confirmed in larger, longer-term studies. |
Introduction
Medication-overuse headache (MOH) is a debilitating secondary disorder characterised by a paradoxical increase in headache frequency resulting from the regular overuse of acute symptomatic medication in patients with a pre-existing primary headache, predominantly migraine [1]. MOH affects approximately 1–2% of the general population. It accounts for a substantial portion of the socioeconomic burden associated with headache disorders, driven by high disability rates and resistance to preventative therapies [2, 3]. While central sensitisation and impaired descending pain modulation are established mechanisms [1, 4], the pathophysiology underlying the transition from migraine to medication overuse remains incompletely understood. The transformation of episodic migraine into MOH is increasingly understood through the lens of nociplastic pain, in which central sensitisation, impaired descending modulation, and dysfunction of reward and cognitive-control circuits act alongside pharmacological and behavioural factors [5–7].
The glymphatic system, a glial-dependent perivascular network, is the brain’s primary waste-clearance pathway, facilitating the removal of interstitial metabolic waste, neurotoxic proteins and inflammatory mediators via perivascular spaces (PVS) [8–10]. Impairment of this system has been implicated in migraine pathophysiology, potentially contributing to the retention of nociceptive signalling molecules and migraine-associated biomarkers [11]. Notably, MOH presents a distinct physiological context compared to episodic migraine, where the brain is subjected to the continuous metabolic demand of processing acute analgesics [such as non-steroidal anti-inflammatory drugs (NSAIDs) or caffeine-containing compounds] alongside the intrinsic burden of the primary headache [12].
We hypothesise that this sustained metabolic demand may exceed the capacity of the glymphatic clearance system. Such "clearance failure" could manifest not merely as functional impairment in diffusivity but as structural maladaptation, including the enlargement of PVS due to fluid stagnation, and subsequent parenchymal damage [10]. While previous studies have examined glymphatic function in migraine using diffusion tensor imaging analysis along the perivascular space (DTI-ALPS) [13–15], findings have been inconsistent. Crucially, it remains unknown whether glymphatic alterations progress with medication exposure and whether they manifest as measurable structural changes in MOH.
To address this gap, we conducted a multi-modal neuroimaging study to characterise glymphatic-related alterations across the migraine-medication exposure continuum. We focused on brain regions implicated in pain processing, reward, and medication response, including limbic structures and subcortical nuclei. We hypothesised that greater medication exposure might be associated with graded, cross-sectional increase in perivascular space burden and downstream hippocampal volume loss, and that these structural changes would demonstrate right-hemispheric predominance consistent with known autonomic asymmetries. Furthermore, we evaluated whether combined glymphatic–volumetric markers could contribute to objective MOH identification.
Methods
Study Design and Participants
This cross-sectional study was conducted at the Headache Center of Ningbo No. 2 Hospital between April 2024 and November 2025. This study was approved by the Ethics Committee of Ningbo No. 2 Hospital (approval numbers YJ-NBEY-KY-2023-199-01 and SL-NBEY-KY-2026-239-01) and was performed in accordance with the Helsinki Declaration of 1964 and its later amendments. All participants provided written informed consent prior to enrolment.
Patients were enrolled based on the following inclusion criteria: (1) diagnosis according to the International Classification of Headache Disorders, 3rd edition (ICHD-3) [16], with confirmation by two independent neurologists with over 15 years of experience in headache disorders; (2) age 18–65 years; (3) right-handed; (4) normal neurological examination; (5) MRI was performed during the interictal period. For episodic and medication-naive patients this was defined as no migraine symptoms for at least 72 h before and 24 h after acquisition; for MOH patients, in whom prolonged headache-free periods are seldom attainable, scanning was performed during the longest available symptom-light interval, with possible residual subclinical activity considered in the limitations.
Patients with migraine were stratified into three groups based on their history of acute medication use (restricted to non-steroidal anti-inflammatory drugs, caffeine, or compound analgesics containing these agents): (1) MOH defined according to ICHD-3 criteria (A8.2), requiring headache ≥ 15 days/month with acute medication use on ≥ 15 days/month for simple analgesics (NSAIDs) or ≥ 10 days/month for combination analgesics, for > 3 months; (2) Patients with migraine and moderate medication use (MM) defined as acute medication use 1–14 days/month for ≥ 3 months without meeting MOH criteria; (3) medication-naive migraine patients (MN), defined as no acute medication use.
Inclusion criteria for healthy controls (HCs) were: (1) age 18–65 years; (2) right-handed; (3) no personal or family history of migraine or other primary headaches; (4) normal neurological examination.
Exclusion criteria for all participants were: (1) other neurological diseases (e.g., Parkinson's disease, Alzheimer's disease, stroke, multiple sclerosis); (2) systemic diseases (e.g., uncontrolled hypertension, cardiac disease, diabetes mellitus, thyroid disorders); (3) history of alcohol or substance abuse; (4) contraindications to MRI; (5) primary sleep disorders.
Clinical Assessments
All participants completed the following assessments: Montreal Cognitive Assessment (MoCA) [17], Hamilton Anxiety Rating Scale (HAMA), 17-item-Hamilton Depression Rating Scale (HAMD) [18, 19], Patient Health Questionnaire-9 (PHQ-9) [20], Pittsburgh Sleep Quality Index (PSQI) [21] and Fatigue Severity Scale (FSS) [22].
MOH and migraine patients underwent additional disease-specific assessments: (1) disease characteristics: duration, attack frequency and medication history; (2) symptom severity: visual analogue scale (VAS) for headache; (3) disease impact: Migraine Disability Assessment (MIDAS) [23], Migraine-Specific Quality of Life Questionnaire (MSQ) (version 2.1) [24], and Headache Impact Test-6 (HIT-6) [25].
MRI Acquisition and Image Processing
All examinations were performed on a 3.0-Tesla scanner (MAGNETOM Prisma; Siemens Healthineers, Erlangen, Germany) with a 64-channel head-neck coil.
The imaging protocol comprised a high-resolution 3D T1-weighted magnetization-prepared rapid gradient-echo (MPRAGE) sequence and a diffusion tensor imaging (DTI) sequence.
Structural T1 images: repetition time (TR) = 2000 ms; echo time (TE) = 2.31 ms; inversion time (TI) = 900 ms; flip angle = 8°; field of view (FOV) = 256 × 256 mm2; matrix = 256 × 256; and slice thickness = 1.0 mm.
DTI images: Multi-band EPI sequence; TR = 2800 ms; TE = 67 ms; FOV = 240 × 240 mm2; matrix = 192 × 192; and slice thickness = 4.0 mm. Diffusion gradients were applied in 20 non-collinear directions (b = 1000 s/mm2, 3 repetitions) with 10 b0 images, utilising a multi-band acceleration factor of 2.
Image Preprocessing
Data preprocessing employed a combined pipeline utilising dcm2niix (v1.0.20241211), MRtrix3(www.mrtrix.org), FSL(www.fmrib.ox.ac.uk/fsl), and FreeSurfer (v7.4.1). Diffusion data underwent thermal noise attenuation (MP-PCA), Gibbs ringing removal, and correction for susceptibility-induced distortions, head motion, and eddy currents (FSL top-up and eddy). Bias field inhomogeneity was corrected using the N4 algorithm. Structural data were processed using the FreeSurfer recon-all stream to generate subject-specific cortical reconstructions and volumetric segmentations (aseg/wmparc), which served as anatomical masks for subsequent ROI analyses.
DTI-ALPS Index
The DTI-ALPS index was calculated separately for the left and right hemispheres, and the bilateral mean was derived as their average. Four 5-mm spherical ROIs were placed in the JHU-ICBM-FA-1-mm template space at the level of the lateral ventricle body: projection fibres (left: 116,110,99; right: 64,110,99) and association fibres (left: 128,110,99; right: 51,110,99), following established coordinates [26, 27]. Individual FA maps were linearly registered to this template using FLIRT, and the resulting transformation matrices were applied to the directional diffusivity maps (Dxx, Dyy, Dzz). The hemisphere-specific DTI-ALPS index was computed as the ratio of mean x-axis diffusivity in projection and association fibre regions to the mean of y-axis diffusivity in projection fibres and z-axis diffusivity in association fibres (Fig. 1).
Fig. 1.

Neuroimaging processing pipeline. a PVSVF: T1-weighted images were preprocessed and processed using vesselness filtering to segment perivascular spaces within white matter. b FW: Diffusion tensor imaging data underwent preprocessing and regularised bi-tensor model fitting to generate free water maps. c DTI-ALPS: FA maps were registered to standard space using FSL FLIRT, followed by extraction of directional diffusivity maps (Dxx, Dyy, Dzz) and ROI placement on projection and association fibres. Bottom panel: Representative ROIs used in this study, including WM, BAS, CSO, HP, NAcc, OFC. PVSVF perivascular space volume fraction, FW free water, DTI-ALPS Diffusion Tensor Imaging Along the Perivascular Space, WM white matter, BAS basal ganglia, CSO centrum semiovale, HP hippocampus, NAcc nucleus accumbens, OFC orbitofrontal cortex
PVS Volume Fraction and Volumetric Analysis
Perivascular space segmentation employed a fully automated pipeline integrating FreeSurfer (v7.4.1), FSL, and custom algorithms in Python. T1-weighted images were first denoised using non-local means filtering (QIT toolbox). A Frangi vesselness filter was then applied in dark-object mode to detect hypointense tubular structures within white matter, generating vesselness probability maps. The Frangi filter enhances structures matching a tubular morphology across multiple scales by analysing the eigenvalues of the Hessian matrix at each voxel; the scale range was set from 0.1 to 9 voxels. The resulting vesselness maps were standardised to z-scores across the whole brain. Voxels exceeding a z-score threshold of 2.3 were classified as candidate PVS, and a minimum cluster size of 5 voxels was applied to remove isolated noise artefacts [28].
Six bilateral ROIs were defined from individual FreeSurfer parcellations: cerebral white matter, centrum semiovale, basal ganglia, nucleus accumbens, hippocampus, and orbitofrontal cortex white matter. ROI volumes were extracted, and perivascular space volume fraction (PVSVF) was calculated as the ratio of segmented PVS volume to ROI volume (Fig. 1).
Free water (FW) values
FW volume fraction maps were generated using a regularised bi-tensor model implemented in the Dipy library. This model separates the diffusion signal into a tissue compartment and an isotropic FW compartment, estimating the fractional volume of extracellular water at each voxel [29]. FreeSurfer-derived ROI masks were registered to native diffusion space using FLIRT with nearest-neighbour interpolation, and mean FW values were extracted from each region (Fig. 1).
Statistical Analysis
Statistical analyses were performed using SPSS (version 25.0; IBM, Armonk, NY, USA) and Python (version 3.9). Normality was assessed using the Shapiro–Wilk test and visual inspection of Q–Q plots. Continuous variables are presented as mean ± standard deviation (SD) for normally distributed data, or median (interquartile range, IQR) for non-normally distributed data. Categorical variables are expressed as frequencies and percentages (n, %). Homogeneity of variance was assessed using Levene’s test. For each imaging metric, a parametric analysis of covariance (ANCOVA) was used when all four groups satisfied normality; otherwise a rank-based ANCOVA on rank-transformed values was applied. Group differences in demographic and clinical characteristics were evaluated using one-way analysis of variance (ANOVA) or the Kruskal–Wallis test for continuous variables, and the Chi-square test for categorical variables. For neuroimaging metrics, intergroup differences were assessed using ANCOVA. Age, sex and intracranial volume (ICV) were included as covariates for all imaging metrics. For hippocampal volume, years of education was pre-specified as an additional covariate, given the established association between educational attainment and hippocampal structure within the cognitive-reserve framework [30]. Education was not applied to perivascular-space, DTI-ALPS or free-water metrics, for which no comparable mechanism is established. Cognitive performance (MoCA) was not used as a covariate, as it is a probable downstream consequence of hippocampal change rather than a confounder, and adjusting for it would over-adjust the effect of interest. Sensitivity analyses using alternative covariate specifications were performed to assess the robustness of findings (Supplementary Material Table S12). The centrum semiovale PVSVF and hippocampal volume were defined a priori as primary regions of interest, and the remaining regions were exploratory. Because the central hypothesis concerned progressive change across the exposure continuum, an a priori linear trend test across the ordered groups (HC < MN < MM < MOH), using orthogonal polynomial contrasts, served as the primary inferential analysis. As a pre-specified single-degree-of-freedom contrast, this trend test was evaluated directly and did not require a significant omnibus effect [31]. Four-group omnibus ANCOVAs across regions were treated as a secondary screen and corrected for multiple comparisons using the Benjamini–Hochberg false-discovery-rate procedure both within each imaging modality and across the full battery of regional ANCOVAs. Post hoc pairwise comparisons were performed for significant effects. Because the principal imaging metrics were non-normally distributed, effect sizes were computed on rank-transformed values, consistent with the rank-based ANCOVA, with 95% confidence intervals from 2000 bootstrap resamples, and are reported as Cohen’s d.
To investigate the relationships among neuroimaging markers and between imaging metrics and clinical phenotypes, partial correlation analyses were conducted within multiple patient subgroups: the MOH group (n = 29), the MM group (n = 34), the MN group (n = 46), the medication-exposed group (MOH + MM, n = 63), and the total migraine cohort (n = 109). Age, sex, years of education, and ICV were included as covariates in all analyses. The normality of each variable was assessed using the Shapiro–Wilk test, Pearson partial correlation was applied when both variables satisfied normality assumptions (p > 0.05), while Spearman partial correlation was employed otherwise. False discovery rate (FDR) correction (Benjamini–Hochberg) was applied within each subgroup, separately for imaging-imaging and imaging-clinical correlations.
Receiver operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic value of imaging markers. To assess the combined diagnostic utility of perivascular-space burden and hippocampal volume, we fitted logistic-regression models including both markers as main effects: a right-hemispheric model with right PVSVF-CSO and right hippocampal volume, and a global model with the corresponding total measures. Whether a multiplicative term improved fit beyond the additive model was tested with a likelihood-ratio test. The area under the curve (AUC) with 95% confidence intervals (estimated via bootstrap resampling, 2000 iterations), sensitivity, and specificity were calculated for each model. Internal validation was performed using tenfold stratified cross-validation, and the optimal classification threshold was determined using the Youden index.
To explore potential non-linear relationships between continuous variables and neuroimaging markers, polynomial regression was employed. Both linear models were fitted, and F tests were used to assess whether the quadratic term significantly improved model fit (indicating non-linearity). For age-related analyses across all participants (n = 149), sex and ICV were included as covariates. For disease duration analyses among migraine patients (n = 109), age, sex, ICV, and education were controlled. Inflection points were estimated from the quadratic regression coefficients. Spearman correlation coefficients were reported for bivariate associations given the non-normal distribution of disease duration.
Results
Demographics and Clinical Characteristics
Between April 2024 and November 2025, 166 participants were initially recruited, of whom 17 were excluded due to a history of using rimegepant or triptans (n = 5), incomplete clinical or MRI data (n = 11), or sleep disorders (n = 1). The final analysis included 149 participants: 29 MOH, 34 MM, 46 MN, and 40 HCs (Fig. 2).
Fig. 2.

Participant recruitment and selection flowchart. HC healthy control, MM patients with migraine and moderate medication use, MN medication-naive migraine patients, MOH medication-overuse headache
Demographic and clinical characteristics are summarised in Table 1. Four groups were well-matched in sex distribution χ2 = 0.02, p = 0.999) and ICV (H = 0.99, p = 0.804), but differed significantly in age (F [3,145] = 6.92, p < 0.001) and educational attainment (H = 15.81, p = 0.001). Post hoc analysis showed that the MOH and HC groups were older than MM and MN groups, and that the MOH group had significantly fewer years of education compared to the MN and HC groups.
Table 1.
Demographic, and clinical parameters of patients with MOH, MM, MN and HCs
| MOH (n = 29) | MM (n = 34) | MN (n = 46) | HC (n = 40) | χ2/F/H | p | |
|---|---|---|---|---|---|---|
| Demographics | ||||||
| Age(years) | 48.6 ± 12.4 | 41.0 ± 9.5 | 38.9 ± 10.9 | 46.5 ± 9.1 | 6.92 | < 0.001 |
| Sex (female) | 20 (69.0%) | 24 (70.6%) | 32 (69.6%) | 28 (70.0%) | 0.02 | 0.999 |
| Education (years) | 4.0 (3.0, 7.0) | 5.0 (3.0,12.0) | 8.0 (5.0, 15.0) | 8.0 (5.0, 10.5) | 15.81 | 0.001 |
| Clinical scores | ||||||
| MoCA | 17.0 (15.0, 20.0) | 23.5 (20.0,25.0) | 24.0 (23.0,26.0) | 26.0 (25.0,27.0) | 70.19 | < 0.001 |
| HAMA | 7.0 (7.0, 9.0) | 6.0 (5.0, 7.8) | 5.0 (4.0, 7.0) | 3.0 (3.0, 4.0) | 66.03 | < 0.001 |
| HAMD | 6.0 (5.0, 7.0) | 5.0 (3.0, 7.0) | 5.0 (4.0, 6.0) | 3.0 (2.0, 3.0) | 56.01 | < 0.001 |
| PHQ-9 | 3.0 (3.0, 4.0) | 2.0 (1.2, 3.0) | 2.0 (1.2, 3.0) | 2.0 (2.0, 2.0) | 24.91 | < 0.001 |
| PSQI | 7.0 (6.0, 9.0) | 7.0 (5.0, 9.0) | 6.0 (5.0, 8.0) | 4.0 (3.0, 4.0) | 69.03 | < 0.001 |
| FSS | 36.0 (26.0, 45.0) | 32.5 (30.0,37.5) | 35.5 (30.0,39.8) | 17.0 (15.5,19.2) | 79.01 | < 0.001 |
| Migraine burden | ||||||
| Frequency | 23.0 (19.0, 28.0) | 5.0 (4.0, 6.0) | 2.0 (1.0, 5.0) | 59.82 | < 0.001 | |
| Duration (years) | 11.0 (6.0, 18.0) | 10.0 (5.2, 15.0) | 5.5 (3.0, 10.0) | 10.94 | 0.004 | |
| HIT-6 | 62.0 (60.0, 66.0) | 58.0 (50.0,60.0) | 50.0 (46.0,55.8) | 43.91 | < 0.001 | |
| MIDAS | 66.0 (61.0, 81.0) | 12.0 (11.0,13.0) | 5.0 (2.0, 7.0) | 63.08 | < 0.001 | |
| MSQ | 62.0 (59.0, 68.0) | 69.0 (67.0,71.0) | 79.0 (78.0,82.0) | 24.30 | < 0.001 | |
| VAS | 7.0 (6.0, 7.0) | 6.0 (5.0, 7.0) | 5.0 (5.0, 6.0) | 17.81 | < 0.001 | |
| Medication days/month | 24 (20.0, 27.0) | 5 (3.0,8.0) | ||||
| Medication type, n (%) | ||||||
| NSAIDs only | 4.0 (13.8%) | 27.0 (79.4%) | ||||
| Caffeine-containing compounds | 25.0 (86.2%) | 7.0 (20.6%) | ||||
MM patients with migraine and moderate medication use, MN medication-naive migraine patients, MOH medication-overuse headache
Clinically, the MOH group exhibited the most severe disease burden with higher headache frequency, longer disease duration, and greater headache-related disability (HIT-6 and MIDAS scores; all p < 0.001). Additionally, this group demonstrated worse cognitive function (MoCA) and a higher burden of psychiatric comorbidities, including anxiety (HAMA), depression (HAMD, PHQ-9), fatigue (FSS), and sleep disturbances (PSQI), relative to all other groups (all p < 0.001).
Neuroimaging Findings
Glymphatic Function and Microstructural Integrity
Analysis of the DTI-ALPS index (left, right, and mean) showed no significant intergroup differences in either unadjusted or adjusted models (all padj > 0.05). Similarly, FW fractions across all regions of interest (including the WM, CSO, BAS, Nacc, HP, OFC) revealed no significant differences among the four groups, regardless of covariate adjustment (all padj > 0.05) (Fig. 3).
Fig. 3.

a DTI-ALPS index across patient groups. Violin plots with embedded box plots showing left, right, and mean DTI-ALPS indices. No significant intergroup differences were observed (all padj > 0.05). Individual data points are displayed. b, c, d PVSVF, FW, and brain volumes, respectively across six region groups. Significant group differences in PVSVF-CSO (padj = 0.025); regional FW content; no significant differences (all padj > 0.05). Regional brain volumes; hippocampal volume significant when education included as covariate (padj = 0.014). Covariates: age, sex, ICV (and education for hippocampus). *p < 0.05. CSO Centrum semiovale, DTI-ALPS diffusion tensor imaging along the perivascular space, FW free water, HC healthy control, MM patients with migraine and moderate medication use, MN medication-naive migraine patients, MOH medication-overuse headache, PVSVF perivascular space volume fraction, vol volume
Perivascular Space Burden
Centrum semiovale PVSVF was a pre-specified primary region. The a priori trend analysis showed a progressive increase in right CSO PVSVF across the continuum (HC < MN < MM < MOH; ptrend = 0.003, FDR adjusted q = 0.006; Fig. 4), which remained significant when education was added as a covariate (ptrend = 0.007). Consistent with this, the four-group ANCOVA was significant for CSO PVSVF (padj = 0.025; Fig. 3); CSO was the only region to reach nominal significance, although this regional effect did not survive FDR correction, either within the six PVS regions (q = 0.150) or across the full battery of imaging ANCOVAs (q = 0.430); the finding therefore rests on the a priori trend test and the convergent associations reported below. The effect was lateralised to the right hemisphere (padj = 0.018) and absent on the left (padj = 0.255). Relative to controls, right CSO PVSVF was elevated in MOH (padj = 0.014, d = 0.59, 95% CI: 0.12 to 1.14) and in MM (padj = 0.030, d = 0.47, 95% CI: −0.02 to 0.99), with no difference in MN (d = 0.00) (Supplementary Material Table S9).
Fig. 4.

Linear trend analysis of right perivascular space burden and right hippocampal volume. a Right PVSVF-CSO (%) showed a significant progressive increase across the disease spectrum (HC < MN < MM < MOH; ptrend = 0.003). b Right hippocampal volume demonstrated corresponding decline (ptrend = 0.003). Box plots show median and interquartile range; individual data points are displayed. CSO centrum semiovale, HC healthy control, MM patients with migraine and moderate medication use, MN medication-naive migraine patients, MOH medication-overuse headache, PVSVF perivascular space volume fraction
Volume Alterations
With education pre-specified as a covariate (see Methods), hippocampal volume differed across groups (padj = 0.014; Fig. 3). For total hippocampal volume. the group difference was in the same direction but did not reach significance without education (padj = 0.052; Supplementary Material Table S12), whereas the right hippocampus, on which the lateralised interpretation rests, was already significant under the base model before education was added (padj = 0.009; Supplementary Material Table S13). The progressive decline across the continuum was significant in the primary trend analysis (ptrend = 0.001, q = 0.004; Fig. 4) and held with or without education. Volume reduction relative to controls was significant in all migraine subgroups (Supplementary Material Table S9), most markedly in MOH (padj = 0.001, d = − 0.61, 95% CI: − 1.16 to − 0.12), with MN (padj = 0.002, d = − 0.49, 95% CI: − 0.97 to − 0.06) and MM (padj = 0.021, d = − 0.32, 95% CI: − 0.84 to 0.15) also reduced.
Age-Related Associations
The DTI-ALPS index showed a significant negative correlation with age (Spearman ρ = − 0.306, 95% CI: − 0.445 to − 0.153, p < 0.001) (Fig. 5). Polynomial regression revealed significant non-linearity for the age-ALPS relationship (F = 4.49, p = 0.036, ΔR2 = 0.026). Piecewise regression identified an optimal breakpoint at 55.3 years, after which the decline rate increased approximately 4.5-fold (− 0.0075/year vs − 0.0018/year for ages ≤ 55 years). Similarly, PVSVF-CSO demonstrated a non-linear age relationship (F = 12.83, p < 0.001), with PVS burden nearly doubling from the 20–35 to the 55 + age group.
Fig. 5.

a, b Age-related trajectories of glymphatic markers. Age-related trajectories of glymphatic markers (n = 149). a Non-linear relationship between age and DTI-ALPS index (mean). Polynomial regression revealed significant non-linearity (F = 4.49, p = 0.036), with accelerated decline after approximately 55 years. Spearman ρ = − 0.306, p < 0.001. b Non-linear relationship between age and PVSVF-CSO (F = 12.83, p < 0.001). Linear correlation was not significant (ρ = 0.112, p = 0.174), but PVS burden demonstrated accelerated accumulation during mid-life. Black curves: quadratic regression fit; grey shading: 95% confidence intervals. Individual data points colour-coded by group: HC (green), MN (blue), MM (orange), MOH (red). Covariates: sex, ICV. c Age-stratified comparison of glymphatic markers. Bar chart comparing mean DTI-ALPS index (blue, left y-axis) and PVSVF-CSO (red, right y-axis) across four age groups. DTI-ALPS showed gradual decline with age (9.2% reduction in 55 + vs 20–35 group), while PVSVF-CSO nearly doubled from youngest to oldest age group. Error bars: standard error of the mean. Sample sizes indicated at bottom of each bar. CSO centrum semiovale, DTI-ALPS diffusion tensor imaging along the perivascular space, HC healthy control, MM patients with migraine and moderate medication use, MN medication-naive migraine patients, MOH medication-overuse headache, PVSVF perivascular space volume fraction
Correlation Analysis
Partial correlation analyses, controlling for age, sex, education, and ICV, examined associations among neuroimaging markers and between imaging metrics and clinical phenotypes across patient subgroups.
Associations Among Imaging Markers
The DTI-ALPS index demonstrated negative correlations with PVSVF-CSO across medication-exposed groups. In the medication-exposed group, mean ALPS was negatively associated with total PVSVF-CSO (r = − 0.442, p < 0.001, pFDR = 0.005), and right ALPS with right PVSVF-CSO (r = − 0.397, p = 0.001, pFDR = 0.011). This pattern was also significant in the MM group (mean ALPS vs total PVSVF-CSO: r = − 0.533, p = 0.001, pFDR = 0.020; right ALPS vs right PVSVF-CSO: r = − 0.469, p = 0.005, pFDR = 0.044) and the total migraine cohort (r = − 0.244, p = 0.011, pFDR = 0.037), but did not reach significance in the MOH group alone (r = − 0.267, p = 0.161, pFDR = 0.296) or MN group (p > 0.05) (Fig. 6).
Fig. 6.

Correlation heatmaps of imaging markers across patient subgroups. Partial correlation matrices of imaging markers across patient subgroups. a MOH group (n = 29). b MM group (n = 34). c MN group (n = 46). d All migraine patients (n = 109). Colour intensity represents partial correlation coefficients. Red indicates positive correlations; blue indicates negative correlations. Significance levels: *p < 0.05, **p < 0.01, ***p < 0.001. DTI-ALPS diffusion tensor imaging along the perivascular space, FW free water, HP hippocampus, L left, MM patients with migraine and moderate medication use, MN medication-naive migraine patients, MOH medication-overuse headache, pVS perivascular space volume fraction, R right, Vol volume
In the MOH group, right ALPS was significantly associated with right hippocampal FW (r = − 0.539, p = 0.003, pFDR = 0.043). This association was replicated in the total migraine cohort (r = − 0.243, p = 0.011, pFDR = 0.037) and showed a trend in the medication-exposed group (r = − 0.276, p = 0.029, pFDR = 0.070), but was not significant in MM (p > 0.05) or MN groups (p > 0.05).
Right PVSVF-CSO was positively correlated with right hippocampal FW across multiple groups: total migraine cohort (r = 0.306, p = 0.001, pFDR = 0.020), the medication-exposed group (r = 0.317, p = 0.011, pFDR = 0.048), and MN group (r = 0.406, p = 0.005, pFDR = 0.044). In the MOH group, this association showed a trend (r = 0.426, p = 0.021, pFDR = 0.091) but did not survive FDR correction.
Higher hippocampal FW was associated with smaller hippocampal volume in the total migraine cohort (r = − 0.248, p = 0.009, pFDR = 0.037) and, at the hemispheric level, on the left in the MM group (r = − 0.448, pFDR = 0.045). Unlike the upstream markers, this final step was not right-lateralised: the right-hemisphere correlation did not reach significance in any group (all p > 0.05) and was absent in MOH (right: r = 0.08).
Clinical Correlations
The DTI-ALPS index exhibited consistent negative correlations with clinical burden across all migraine subgroups. In the total migraine cohort (n = 109), mean ALPS was negatively associated with sleep disturbance (PSQI: ρ = − 0.515, p < 0.001, pFDR < 0.001), depression (HAMD: ρ = − 0.495, p < 0.001, pFDR < 0.001), fatigue (FSS: ρ = − 0.463, p < 0.001, pFDR < 0.001), anxiety (HAMA: ρ = − 0.448, p < 0.001, pFDR < 0.001), and pain intensity (VAS: ρ = − 0.317, p = 0.001, pFDR = 0.005) (Fig. 7). Similar patterns were observed in the medication-exposed group, MM, and MN subgroups, with all associations surviving FDR correction. In the MOH group, these correlations showed consistent direction but did not survive FDR correction, possibly reflecting limited statistical power.
Fig. 7.

Correlations between glymphatic markers and clinical indicators in the total migraine cohort (n = 109). a–e DTI-ALPS index (mean) showed significant negative correlations with PSQI (ρ = − 0.515, p < 0.001), HAMD (ρ = − 0.495, p < 0.001), HAMA (ρ = − 0.448, p < 0.001), FSS (ρ = − 0.463, p < 0.001), and VAS (ρ = − 0.317, p = 0.001). (f–h) PVSVF-CSO demonstrated positive correlations with PSQI (ρ = 0.358, p < 0.001), HAMA (ρ = 0.344, p = 0.001), and HAMD (ρ = 0.305, p = 0.001). (I) Right hippocampal volume was negatively associated with disease duration in medication-exposed patients (MOH + MM, n = 63; ρ = − 0.363, p = 0.003). Red lines: linear regression fit;
shaded areas: 95% confidence intervals. All correlations shown remained significant after Benjamini–Hochberg FDR correction (pFDR < 0.05). CSO centrum semiovale, DTI-ALPS diffusion tensor imaging along the perivascular space, FSS Fatigue Severity Scale, HAMA Hamilton Anxiety Rating Scale, HAMD17-item-Hamilton Depression Rating Scale, HP hippocampus, MM patients with migraine and moderate medication use, MOH medication-overuse headache, PSQI Pittsburgh Sleep Quality Index, PVSVF perivascular space volume fraction, VAS visual analogue scale
Total PVSVF-CSO burden was positively correlated with clinical symptoms in the total migraine cohort: sleep disturbance (PSQI: ρ = 0.358, p < 0.001, pFDR = 0.001), anxiety (HAMA: ρ = 0.344, p < 0.001, pFDR = 0.002), and depression (HAMD: ρ = 0.305, p = 0.001, pFDR = 0.007) (Fig. 7). These associations were also significant in the medication-exposed group (all pFDR < 0.05) but did not survive correction in the MOH group alone.
Right hippocampal volume demonstrated a negative association with disease duration in medication-exposed patients (the medication-exposed group: ρ = − 0.363, p = 0.003, pFDR = 0.022) (Fig. 7). This association was not significant in individual MOH or MM subgroups.
Traditional linear correlation between disease duration and hippocampal volume was non-significant (all p > 0.05). However, polynomial regression, controlling for age, sex, ICV, and education, revealed significant non-linear relationships for right hippocampal volume (F = 6.49, p = 0.012, ΔR2 = 0.030) and total hippocampal volume (F = 6.91, p = 0.010, ΔR2 = 0.030), with consistent inflection points at approximately 10.9–11.2 years of disease duration. These non-linear effects were robust to the exclusion of education (right: p = 0.009; total: p = 0.006), with unchanged inflection points, suggesting accelerated volume decline beyond this threshold (Fig. 8).
Fig. 8.

Disease duration-related changes in hippocampal volume. Non-linear relationship between disease duration and hippocampal volume in migraine patients (n = 109). a Right hippocampal volume: polynomial regression demonstrated significant non-linearity (F = 6.49, p = 0.012) with inflection point at approximately 11.2 years. b Total hippocampal volume: F = 6.91, p = 0.010; inflection point ~ 10.9 years. Traditional linear correlations were non-significant (all p > 0.05). Black curves: quadratic fit after adjustment for age, sex, ICV, and education; grey shading: 95% confidence intervals; red dashed lines: estimated inflection points. Individual data points colour-coded: MN (blue), MM (orange), MOH (red). Results suggest accelerated hippocampal volume decline beyond approximately 11 years of disease duration. MM patients with migraine and moderate medication use, MN medication-naive migraine patients, MOH medication-overuse headache
Diagnostic Value of Multimodal Markers
ROC analysis assessed the discriminative capacity of a combined model including CSO PVSVF and hippocampal volume. A multiplicative term did not improve fit beyond the additive combination (likelihood-ratio test, p = 0.960), so an additive main-effects model was retained. In differentiating MOH from HCs, the right-hemispheric model (right PVSVF-CSO and right hippocampal volume) achieved an AUC of 0.75 (95% CI: 0.65–0.87; tenfold cross-validation AUC: 0.72 ± 0.14), with a sensitivity of 79.3% and specificity of 65.0%, consistent with the right-hemispheric pattern. For differentiating MOH from MM, the global combined model (total PVSVF-CSO and total hippocampal volume) achieved an AUC of 0.69 (95% CI: 0.56–0.81; tenfold cross-validation AUC: 0.68 ± 0.15), with a sensitivity of 96.6% and specificity of 37.0% (Fig. 9).
Fig. 9.

ROC curves for diagnostic discrimination. ROC analysis evaluated the discriminative capacity of the combined model combining PVSVF-CSO and hippocampal volume. Top row: comparisons involving MOH (MOH vs HC, MOH vs MN, MOH vs MM). Bottom row: comparisons involving MM and MN (MM vs HC, MM vs MN, MN vs HC). Red, blue, and green curves represent right-hemispheric, left-hemispheric, and total combined models, respectively. Area under the curve (AUC) values are shown in the legend. The right-hemispheric model achieved AUC = 0.75 (95% CI: 0.65–0.87) for MOH vs. HC discrimination. MM patients with migraine and moderate medication use, MN medication-naive migraine patients, MOH medication-overuse headache, ROC receiver operating characteristic
Discussion
This multi-modal neuroimaging study provides evidence for glymphatic-related alterations across the migraine-medication exposure continuum. We found a graded, cross-sectional increase in centrum semiovale PVS burden and a reduction in hippocampal volume across the medication exposure continuum, both with right-hemispheric predominance. Correlation analyses revealed a lateralised association between right-hemispheric PVS burden and ipsilateral hippocampal microstructural changes. The combined assessment of these structural markers achieved moderate diagnostic accuracy for MOH identification.
Increased CSO Perivascular Space Burden and Glymphatic Impairment
The observed increase in CSO PVS burden in MOH patients aligns with the emerging understanding of perivascular space morphology as a biomarker of impaired waste clearance [32, 33]. Perivascular spaces serve as conduits for CSF-ISF exchange, and their dilation typically reflects compromised glymphatic drainage capacity [34]. The selective involvement of the centrum semiovale warrants consideration. Unlike basal ganglia PVS, which predominantly reflects hypertensive arteriopathy, CSO-PVS burden has been specifically linked to subcortical white matter drainage impairment and migraine pathophysiology. Neuroimaging studies using 3T MRI have demonstrated that CSO-EPVS significantly correlated with deep white matter hyperintensity severity in migraine patients, suggesting that perivascular dysfunction may precede or contribute to white matter injury [35]. The predominance of PVS changes in medication-using groups suggests that pharmacological exposure may be a contributing factor to these structural alterations. In the context of MOH, this may arise from a dual metabolic burden: the accumulation of endogenous neuropeptides, particularly CGRP, alongside exogenous drug metabolites requiring hepatic biotransformation and subsequent central nervous system clearance [36, 37]. Because the four-group difference did not survive correction across regions, this interpretation rests on the pre-specified trend and the convergent associations, and is advanced as a hypothesis.
Right-Hemispheric Vulnerability and a Proposed Mechanistic Sequence
The observed right-hemispheric predominance of structural alterations aligns with established neurobiological asymmetries. Vestibular and autonomic processing demonstrates right-hemisphere dominance in right-handed individuals, and migraine-related grey matter changes preferentially affect right-sided structures [38, 39]. Additionally, norepinephrine-mediated slow vasomotion during sleep drives glymphatic clearance through rhythmic arterial pulsations [40]. Chronic analgesic exposure, particularly caffeine-containing compounds common among MOH patients, may disrupt this physiological pump mechanism through sustained vasoconstriction, preferentially affecting the autonomically dominant right hemisphere [41].
This lateralised vulnerability was further evidenced by the graded, cross-sectional association observed across the medication exposure gradient. Right PVSVF-CSO demonstrated a stepwise increase from healthy controls through medication-naive migraineurs to medication-exposed groups, whereas left-sided changes showed no comparable progression. The selective right-hemispheric gradient suggests that medication-related metabolic burden accumulates preferentially in regions with higher baseline autonomic drive, consistent with the proposed noradrenergic mechanism [40, 42, 43].
Correlation analyses were consistent with a lateralised, right-hemisphere link between glymphatic and limbic measures. In MOH patients, lower right DTI-ALPS was associated with higher right hippocampal FW, and right PVSVF-CSO correlated with right hippocampal FW across several subgroups. These associations were clearest in the pooled and medication-exposed cohorts; within MOH alone the individual links were mostly at trend level, given the small sample. We therefore regard this right-hemisphere sequence of glymphatic-to-limbic change as a hypothesis for longitudinal testing rather than an established pathway.
Hippocampal Volume Reduction and a Possible Disease Duration Threshold
Hippocampal volume reduction in MOH patients extends prior observations of grey matter alterations in chronic headache disorders [44, 45]. The stepwise pattern of hippocampal volume reduction across the disease spectrum, with the most pronounced changes observed in MOH followed by chronic migraine with medication use and medication-naive migraineurs, suggests a cumulative pathological process. This gradient may reflect the neurotoxic consequences of recurrent trigeminal activation, whereby inflammatory mediators propagate through the trigeminal–parabrachial–amygdala–hippocampus circuit [46]. The hippocampus, owing to its high metabolic demands and dense glutamatergic innervation, exhibits particular vulnerability to excitotoxic stress associated with chronic pain states [47]. Consistent with this, elevated hippocampal FW content was negatively correlated with hippocampal volume in the total migraine cohort and MM group, though this association did not reach significance in the MOH group alone and, unlike the upstream markers, was not right-lateralised. Right hippocampal volume demonstrated a significant negative association with disease duration in medication-exposed patients. The nonlinear trajectory, with accelerated volume loss beyond approximately 11 years of disease duration, raises the possibility of a critical window during which structural changes may become more pronounced. If confirmed longitudinally, this finding could inform early intervention strategies.
Integrating Null Findings: DTI-ALPS and Free Water
The absence of significant group differences in DTI-ALPS index contrasts with our structural findings. Recent studies have reported both elevated and reduced ALPS indices across migraine subtypes, reflecting complex stage-dependent glymphatic regulation. Wu et al. demonstrated reduced DTI-ALPS in chronic migraine patients with medication overuse [14], while Zhang et al. observed elevated indices during migraine chronification [48]. The present null finding at the group level may reflect several methodological and biological factors.
Age exerted a substantial confounding influence on DTI-ALPS in our cohort. The index showed accelerated decline after approximately 55 years of age, and, when controlled in ANCOVA, age remained a strong independent predictor whereas the between-group effect became non-significant [49]. Given the significant age disparity between groups in our sample, age-related glymphatic decline may have overshadowed disease-specific effects. This observation underscores the necessity of age-stratified analyses in future studies examining migraine chronification.
Additionally, the present null finding may represent a compensatory plateau wherein glymphatic upregulation observed in earlier disease stages has reached functional limits. The robust individual-level correlations between ALPS index and clinical measures, as well as the significant association with PVS burden, support the physiological relevance of this metric despite apparent group-level normalisation [50]. The modest sample size also limited statistical power for detecting medium-sized effects.
The absence of group differences in regional FW indices suggests that gross tissue oedema has not yet developed at this disease stage. These null results should not be taken as evidence against glymphatic involvement, since the coarse diffusion acquisition and likely partial-volume contamination would have reduced sensitivity to any true difference. Free water quantification reflects extracellular fluid accumulation, which typically represents a more advanced pathological state than the PVS morphological changes observed in our cohort [51]. However, the significant correlations between FW content and PVS burden at the individual level indicate that more subtle microstructural changes may precede overt oedematous transformation [52]. This pattern is consistent with a staged model wherein perivascular drainage impairment manifests initially as PVS dilation, with parenchymal FW accumulation representing a subsequent downstream consequence.
Diagnostic Implications
A combined model of right PVSVF-CSO and right hippocampal volume achieved an AUC of 0.75 for distinguishing MOH from healthy controls (sensitivity 79.3%, specificity 65.0%), with comparable performance for MOH versus MN discrimination (global model: AUC = 0.69). The moderate discriminative accuracy indicates these markers should be considered as potential adjuncts to established clinical criteria rather than standalone diagnostic tools. The low specificity of the global model (37%) for MOH versus MM discrimination indicates substantial overlap in glymphatic–structural features between these medication-exposed groups, consistent with shared underlying pathology along the medication exposure continuum.
Limitations
Several limitations should be acknowledged. The cross-sectional design precludes causal inference and cannot separate medication-specific effects from headache chronification, as the medication-naive comparators were largely episodic; a chronic-migraine group without medication overuse is therefore needed in future work. The MM group also spanned a broad medication-use range (one to fourteen days per month) without recorded cumulative duration, adding heterogeneity to the intermediate category. Because MOH patients seldom achieve prolonged headache-free periods, residual subclinical activity may also have affected the imaging measures. Several potentially influential variables were not recorded, including aura status and preventive treatment (both linked to hippocampal morphometry), white-matter hyperintensity burden (associated with CSO-PVS enlargement [34]), and headache laterality (precluding structure-symptom analysis). Technically, the diffusion acquisition was suboptimal for DTI-ALPS and bi-tensor free-water modelling (20 directions, single-shell b = 1000 s/mm2, 4 mm slices), and projecting 1-mm regions onto these coarse data may have introduced partial-volume contamination, together reducing sensitivity to diffusion-based differences; the automated perivascular-space pipeline was also not validated against manual ratings. Finally, the sample was restricted to right-handed participants, so the right-sided predominance may partly reflect this selection; the seventeen exclusions may have shifted the sample toward milder disease; and the modest MOH group (n = 29) limited power. Accordingly, the regional CSO PVSVF effect did not survive correction across all regions, rests on the a priori trend analysis, and should be regarded as hypothesis-generating pending replication.
Conclusions
This study characterises neuro-glymphatic alterations across the migraine-medication exposure continuum, showing a graded, cross-sectional increase in centrum semiovale perivascular space burden and hippocampal volume reduction. Correlation analyses revealed a lateralised association between right-hemispheric glymphatic function and ipsilateral limbic change, with a tentative inflection near 11 years of disease duration beyond which hippocampal volume decline appeared to accelerate, a pattern requiring longitudinal confirmation. If replicated longitudinally, these findings suggest that glymphatic-related structural markers might aid in early detection before full MOH criteria are met. Our results extend current understanding of MOH pathophysiology beyond traditional neurovascular frameworks, raising the possibility that impaired waste clearance mechanisms may represent a modifiable therapeutic target (Fig. 10).
Fig. 10.

Proposed mechanistic model of glymphatic dysfunction across the migraine-medication exposure continuum. Left panel (Normal Glymphatic Clearance): under physiological CSF flows along periarterial spaces, facilitated by AQP4 channels on astrocytic endfeet, enabling efficient clearance of metabolic waste through PVS. Middle panel (Clearance Failure): progressive medication exposure imposes a cumulative metabolic burden, leading to impaired glymphatic clearance. This manifests as PVS enlargement due to fluid stagnation and accumulation of waste products. Right panel (Downstream Damage): prolonged clearance failure is proposed to increase FW content and drive hippocampal volume loss, the latter showing right-hemispheric predominance. AQP4 aquaporin-4, CSF cerebrospinal fluid, PVS perivascular space
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgments
Medical Writing/Editorial Assistance
The authors used DeepSeek (DeepSeek AI, Hangzhou, China), an artificial-intelligence-based large language model, to assist with language editing during preparation of this manuscript. All text was reviewed, verified, and approved by the authors, who take full responsibility for the integrity and accuracy of the work. No funding was received for this assistance.
Author Contributions
Conceptualization: Ange Li, Yunqin Wu; Data curation: Li Li, Weinv Fan; Methodology: Ange Li, Yunqin Wu, Guoping Chen; Formal analysis: Bin Chen, Zhenyi Fan; Resources: Qiongfeng Guan, Guoping Chen; Writing – original draft: Ange Li; Writing – review & editing: Yunqin Wu, Qiongfeng Guan. All authors reviewed the manuscript.
Funding
This work was supported by the Zhejiang Medical and Health Program, Zhejiang, China (No.2025KY1385); the Hwa Mei Foundation, Ningbo, Zhejiang, China (No. 2023HMKY45); AI-Enabled design synthesis biological evaluation, and clinical development of innovative therapeutics for ischemic stroke, Zhejiang, China (No. SIML-202306); the Ningbo Municipal Health and Wellness Science and Technology Program, Ningbo, Zhejiang, China (No. 2023Y18); and the DCMST.NHC Clinical Research Project (No. WKZX2025CZ0205). The journal’s Rapid Service Fee was funded by the authors.
Data Availability
The datasets generated and/or analysed during the current study are available from the corresponding author upon reasonable request, subject to institutional and ethical restrictions on sharing patient data.
Declarations
Conflict of Interest
Ange Li, Qiongfeng Guan, Guoping Chen, Bin Chen, Li Li, Weinv Fan, Zhenyi Fan, and Yunqin Wu declare that they have no competing interests. The authors would like to express their sincere gratitude to the Department of Radiology at Ningbo No. 2 Hospital for their support and collaboration. We are also deeply thankful to all the participants who dedicated their time and effort to this study.
Ethical Approval
This study was approved by the Ethics Committee of Ningbo No. 2 Hospital (approval numbers YJ-NBEY-KY-2023-199-01 and SL-NBEY-KY-2026-239-01) and was performed in accordance with the Helsinki Declaration of 1964 and its later amendments. All participants provided written informed consent prior to enrolment. No individually identifiable patient information is included in this article.
References
- 1.Ashina S, Terwindt GM, Steiner TJ, et al. Medication overuse headache. Nat Rev Dis Primers. 2023;9(1):5. [DOI] [PubMed] [Google Scholar]
- 2.Gosalia H, Moreno-Ajona D, Goadsby PJ. Medication-overuse headache: a narrative review. J Headache Pain. 2024;25(1):89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Micieli A, Robblee J. Medication-overuse headache. CMAJ. 2018;190(10):E296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Diener HC, Holle D, Solbach K, Gaul C. Medication-overuse headache: risk factors, pathophysiology and management. Nat Rev Neurol. 2016;12(10):575–83. [DOI] [PubMed] [Google Scholar]
- 5.Messina R, Castaldo M, et al. Breaking the medication overuse headache cycle: from nociplastic pain mechanisms to patient-centered interventions. J Headache Pain. 2026;27(1):108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Raggi A, Leonardi M, Arruda M, et al. Hallmarks of primary headache: part 1 - migraine. J Headache Pain. 2024;25(1):189. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.García-Azorín D, Perez-de-la-Parte A, Tana C, et al. Hallmarks of headache disorders: part 5 - secondary headaches. J Headache Pain. 2026;27(1):62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Peng S, Liu J, Liang C, et al. Aquaporin-4 in glymphatic system, and its implication for central nervous system disorders. Neurobiol Dis. 2023;179:106035. [DOI] [PubMed] [Google Scholar]
- 9.Yamamoto EA, Bagley JH, Geltzeiler M, et al. The perivascular space is a conduit for cerebrospinal fluid flow in humans: a proof-of-principle report. Proc Natl Acad Sci U S A. 2024;121:e2407246121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Barisano G, Lynch KM, Sibilia F, et al. Imaging perivascular space structure and function using brain MRI. Neuroimage. 2022;257:119329. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.He J, Zhou M, Zhao F, et al. FGF-21 and GDF-15 are increased in migraine and associated with the severity of migraine-related disability. J Headache Pain. 2023;24(1):28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Vandenbussche N, Laterza D, Lisicki M, et al. Medication-overuse headache: a widely recognized entity amidst ongoing debate. J Headache Pain. 2018;19(1):50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Vittorini MG, Sahin A, Trojan A, et al. The glymphatic system in migraine and other headaches. J Headache Pain. 2024;25(1):34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Wu CH, Chang FC, Wang YF, et al. Impaired glymphatic and meningeal lymphatic functions in patients with chronic migraine. Ann Neurol. 2024;95(3):583–95. [DOI] [PubMed] [Google Scholar]
- 15.Cao Y, Huang M, Fu F, et al. Abnormally glymphatic system functional in patients with migraine: a diffusion kurtosis imaging study. J Headache Pain. 2024;25(1):118. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Headache Classification Committee of the International Headache Society (IHS). The International Classification of Headache Disorders, 3rd edition. Cephalalgia. 2018;38(1):1–211. [DOI] [PubMed]
- 17.Islam N, Hashem R, Gad M, et al. Accuracy of the Montreal Cognitive Assessment tool for detecting mild cognitive impairment: a systematic review and meta-analysis. Alzheimers Dement. 2023;19(7):3235–43. [DOI] [PubMed] [Google Scholar]
- 18.Hamilton M. The assessment of anxiety states by rating. Br J Med Psychol. 1959;32(1):50–5. [DOI] [PubMed] [Google Scholar]
- 19.Hamilton M. A rating scale for depression. J Neurol Neurosurg Psychiatry. 1960;23(1):56–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Levis B, Benedetti A, Thombs BD. Accuracy of Patient Health Questionnaire-9 (PHQ-9) for screening to detect major depression: individual participant data meta-analysis. BMJ. 2019;365:l1476. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Buysse DJ, Reynolds CF 3rd, Monk TH, et al. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Res. 1989;28(2):193–213. [DOI] [PubMed] [Google Scholar]
- 22.Krupp LB, LaRocca NG, Muir-Nash J, et al. The fatigue severity scale: application to patients with multiple sclerosis and systemic lupus erythematosus. Arch Neurol. 1989;46(10):1121–3. [DOI] [PubMed] [Google Scholar]
- 23.Stewart WF, Lipton RB, Dowson AJ, et al. Development and testing of the Migraine Disability Assessment (MIDAS) questionnaire to assess headache-related disability. Neurology. 2001;56(6 Suppl 1):S20–8. [DOI] [PubMed] [Google Scholar]
- 24.Martin BC, Pathak DS, Sharfman MI, et al. Validity and reliability of the migraine-specific quality of life questionnaire (MSQ version 2.1). Headache. 2000;40(3):204–15. [DOI] [PubMed] [Google Scholar]
- 25.Kosinski M, Bayliss MS, Bjorner JB, et al. A six-item short-form survey for measuring headache impact: the HIT-6. Qual Life Res. 2003;12(8):963–74. [DOI] [PubMed] [Google Scholar]
- 26.Taoka T, Masutani Y, Kawai H, et al. Evaluation of glymphatic system activity with the diffusion MR technique: diffusion tensor image analysis along the perivascular space (DTI-ALPS) in Alzheimer’s disease cases. Jpn J Radiol. 2017;35(4):172–8. [DOI] [PubMed] [Google Scholar]
- 27.Zhang W, Zhou Y, Wang J, et al. Glymphatic clearance function in patients with cerebral small vessel disease. Neuroimage. 2021;238:118257. [DOI] [PubMed] [Google Scholar]
- 28.Kamagata K, Andica C, Takabayashi K, et al. Association of MRI indices of glymphatic system with amyloid deposition and cognition in mild cognitive impairment and Alzheimer disease. Neurology. 2022;99(24):e2648–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Garyfallidis E, Brett M, Amirbekian B, et al. Dipy, a library for the analysis of diffusion MRI data. Front Neuroinform. 2014;8:8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Stern Y. Cognitive reserve in ageing and Alzheimer’s disease. Lancet Neurol. 2012;11(11):1006–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Maxwell SE, Delaney HD, Kelley K. Designing experiments and analyzing data: a model comparison perspective. 3rd ed. New York: Routledge; 2017. [Google Scholar]
- 32.Wardlaw JM, Benveniste H, Nedergaard M, et al. Perivascular spaces in the brain: anatomy, physiology and pathology. Nat Rev Neurol. 2020;16(3):137–53. [DOI] [PubMed] [Google Scholar]
- 33.Duering M, Biessels GJ, Brodtmann A, et al. Neuroimaging standards for research into small vessel disease-advances since 2013. Lancet Neurol. 2013;22(7):602–18. [DOI] [PubMed] [Google Scholar]
- 34.Rasmussen MK, Mestre H, Nedergaard M. The glymphatic pathway in neurological disorders. Lancet Neurol. 2018;17(11):1016–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Yuan Z, Li W, Tang H, et al. Enlarged perivascular spaces in patients with migraine: a case-control study based on 3T MRI. Ann Clin Transl Neurol. 2023;10(7):1160–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Messlinger K. The big CGRP flood - sources, sinks and signalling sites in the trigeminovascular system. J Headache Pain. 2018;19(1):22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Schain AJ, Melo-Carrillo A, Strassman AM, et al. Cortical spreading depression closes paravascular space and impairs glymphatic flow: implications for migraine headache. J Neurosci. 2017;37(11):2904–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Hougaard A, Amin FM, Hoffmann MB, et al. Structural gray matter abnormalities in migraine relate to headache lateralization, but not aura. Cephalalgia. 2015;35(1):3–9. [DOI] [PubMed] [Google Scholar]
- 39.Messina R, Rocca MA, Colombo B, et al. Gray matter volume modifications in migraine: a cross-sectional and longitudinal study. Neurology. 2018;91(3):e280–92. [DOI] [PubMed] [Google Scholar]
- 40.Hauglund NL, Andersen M, Tokarska K, et al. Norepinephrine-mediated slow vasomotion drives glymphatic clearance during sleep. Cell. 2025;188(3):606-622.e17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Addicott MA, Yang LL, Peiffer AM, et al. The effect of daily caffeine use on cerebral blood flow: how much caffeine can we tolerate? Hum Brain Mapp. 2009;30(10):3102–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Fontes MAP, Dos Santos Machado LR, Viana ACR, et al. The insular cortex, autonomic asymmetry and cardiovascular control: looking at the right side of stroke. Clin Auton Res. 2024;34(6):549–60. [DOI] [PubMed] [Google Scholar]
- 43.Ferraro S, Klugah-Brown B, Tench CR, et al. The central autonomic system revisited - convergent evidence for a regulatory role of the insular and midcingulate cortex from neuroimaging meta-analyses. Neurosci Biobehav Rev. 2022;142:104915. [DOI] [PubMed] [Google Scholar]
- 44.Planchuelo-Gomez A, Garcia-Azorin D, Guerrero AL, et al. Gray matter structural alterations in chronic and episodic migraine: a morphometric magnetic resonance imaging study. Pain Med. 2020;21(11):2997–3011. [DOI] [PubMed] [Google Scholar]
- 45.Luo W, Teng Y, Chen X, et al. Gray matter alterations in medication overuse headache: a voxel-based morphometry meta-analysis. Pain Physician. 2025;28(2):E115–27. [PubMed] [Google Scholar]
- 46.Goadsby PJ, Holland PR, Martins-Oliveira M, et al. Pathophysiology of migraine: a disorder of sensory processing. Physiol Rev. 2017;97(2):553–622. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Popoli M, Yan Z, McEwen BS, Sanacora G. The stressed synapse: the impact of stress and glucocorticoids on glutamate transmission. Nat Rev Neurosci. 2011;13(1):22–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Zhang X, Wang W, Bai X, et al. Increased glymphatic system activity in migraine chronification by diffusion tensor image analysis along the perivascular space. J Headache Pain. 2023;24(1):147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Wang J, Zhou Y, Zhang K, et al. Glymphatic function plays a protective role in ageing-related cognitive decline. Age Ageing. 2023;52(7):afad107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Ringstad G. Glymphatic imaging: a critical look at the DTI-ALPS index. Neuroradiology. 2024;66(2):157–60. [DOI] [PubMed] [Google Scholar]
- 51.Hong H, Tozer DJ, Markus HS. Relationship of perivascular space markers with incident dementia in cerebral small vessel disease. Stroke. 2024;55(4):1032–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Li H, Jacob MA, Cai M, et al. Perivascular spaces, diffusivity along perivascular spaces, and free water in cerebral small vessel disease. Neurology. 2024;102(8):e209306. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated and/or analysed during the current study are available from the corresponding author upon reasonable request, subject to institutional and ethical restrictions on sharing patient data.
