Sage Journals: Discover world-class research

Abstract

Background:

The glymphatic system (GS) is critical for interstitial fluid clearance and has been implicated in various neurological disorders. However, its role in drug-resistant epilepsy (DRE) and response to vagus nerve stimulation (VNS) remains unclear.

Objectives:

This study aimed to assess GS function in patients with DRE using diffusion tensor image analysis along the perivascular space (DTI-ALPS) and to explore its association with VNS outcomes.

Design:

Data from patients with DRE who underwent VNS were retrospectively analyzed at a single center.

Methods:

Forty patients with DRE and 30 healthy controls were enrolled. ALPS indices were calculated for the left (ALPS-L), right (ALPS-R), and bilateral (ALPS-B) hemispheres. The clinical outcomes after VNS were classified using McHugh outcome classification. Statistical comparisons and logistic regression were used to examine group differences and the predictive value of ALPS indices.

Results:

Compared to controls, patients with DRE exhibited significantly reduced ALPS-L (p < 0.001) and ALPS-B (p = 0.005) indices. Among patients with DRE, responders to VNS had significantly higher ALPS-R (p = 0.021) and ALPS-B (p = 0.027) indices than nonresponders. ALPS-L exhibited a positive trend (p = 0.073). In logistic regression analyses, ALPS-R and ALPS-B were significantly associated with favorable VNS outcomes in univariate and multivariate models. After adjusting for age, sex, and disease duration, ALPS-L (p = 0.028, odds ratio (OR) = 2.739), ALPS-R (p = 0.020, OR = 2.802), and ALPS-B (p = 0.020, OR = 3.241) all emerged as independent predictors of VNS response.

Conclusion:

Glymphatic dysfunction is present in DRE and may influence responsiveness to VNS. Higher ALPS indices predicted better treatment outcomes, supporting ALPS potential as a noninvasive imaging biomarker for presurgical stratification in epilepsy.

Plain language summary

Glymphatic dysfunction in drug-resistant epilepsy

Why was the study done?

The study investigated glymphatic system function in drug-resistant epilepsy (DRE) and its potential role in predicting outcomes of vagus nerve stimulation (VNS) therapy, as no reliable biomarkers currently exist.

What did the researchers do?

Researchers compared DTI-ALPS indices (measuring glymphatic activity) between 40 DRE patients and 30 healthy controls, and assessed whether preoperative ALPS values could predict VNS treatment response using statistical models.

What did the researchers find?DRE patients showed significant glymphatic dysfunction compared to controls. Higher preoperative ALPS values, especially in the right and bilateral hemispheres, were strongly associated with better VNS treatment outcomes.

What do the findings mean?

The findings suggest that glymphatic function may influence VNS efficacy, and ALPS index could serve as a non-invasive biomarker to help identify DRE patients most likely to benefit from neuromodulation therapy.

Keywords

diffusion tensor image analysis along the perivascular space drug-resistant epilepsy glymphatic system vagus nerve stimulation

Introduction

The glymphatic system (GS) is a recently characterized perivascular fluid-transport network that drives cerebrospinal fluid (CSF) influx along periarterial spaces, enables its exchange with interstitial fluid (ISF), and promotes solute clearance via perivenous routes.¹ This process depends critically on the polarized expression of aquaporin-4 (AQP4) water channels on astrocytic endfeet, whose loss or mislocalization markedly slows the glymphatic flow and waste removal.² Efficient GS activity is crucial for maintaining cerebral homeostasis by facilitating neurotoxic metabolites clearance, most notably amyloid β, tau, and proinflammatory cytokines, from the brain parenchyma.³ Notably, glymphatic flux is maximized during deep sleep and is further propelled by intact neurovascular pulsation, underscoring the synergistic roles of sleep architecture and vascular dynamics in brain waste clearance.⁴

Since its first in vivo application via diffusion tensor imaging analysis along the perivascular space (DTI-ALPS), GS has been increasingly recognized as a critical factor in numerous neurological and neurodegenerative disorders. In alzheimer’s, ALPS index reduction is correlated with both Aβ deposition and cognitive decline.⁵ In parkinson’s, ALPS indices predict disease progression independent of nigrostriatal dopamine degeneration.^6,7 Besides, glymphatic dysfunction is associated with disability and lesion load in multiple sclerosis,⁸ white matter damage, and cognitive impairment in cerebral small vessel disease.⁹ These findings have positioned DTI-ALPS as a noninvasive, reproducible proxy of GS function with promising translational relevance.

Additionally, its established role in neurodegenerative disorders and glymphatic dysfunction has recently been implicated in chronic epilepsy. Multiple studies using the DTI-ALPS approach have demonstrated significantly reduced ALPS indices in patients with various epilepsy subtypes. For example, c and frontal lobe epilepsy (FLE) cohorts consistently exhibit glymphatic impairment compared to healthy controls (HC).^10,11 Similar findings have been reported in pediatric populations, including childhood absence epilepsy (CAE), Rolandic epilepsy (RE), and juvenile myoclonic epilepsy, suggesting that glymphatic compromise may occur early in disease evolution.^12
–14

Moreover, the degree of ALPS reduction in temporal lobe epilepsy (TLE) has been linked to lateralized seizure foci and associated cognitive deficits. Specifically, ALPS asymmetry is correlated with the dominant hemisphere involvement, whereas lower ALPS indices are correlated with poorer semantic fluency performance.^15,16 These observations support the notion that glymphatic dysfunction in epilepsy may reflect structural vulnerability and network-specific metabolic burden.

In pediatric and early-onset epilepsies, ALPS reduction appears even in patients without structural lesions or cognitive complaints, supporting the hypothesis that glymphatic impairment may precede overt neural damage.^12,13 Moreover, in focal cortical dysplasia (FCD), nonresponders to antiseizure medications (ASMs) exhibit more severe ALPS impairments than responders, implying a potential role of GS in drug resistance.¹⁷ Consistently, studies in newly diagnosed patients with focal epilepsy demonstrated early glymphatic compromise, with ALPS indices negatively correlated with disease duration.¹⁸

Despite growing recognition of GS disruption in epilepsy, its role in surgical or neuromodulatory outcome prediction remains poorly understood. Vagus nerve stimulation (VNS) is an FDA-approved palliative intervention for drug-resistant epilepsy (DRE) that modulates cortical excitability and autonomic tone.¹⁹ Beyond seizure control, VNS offers additional clinical benefits by promoting better cognitive outcomes, emotional health, and patient-reported life quality.²⁰ However, therapeutic response to VNS is highly variable, and no validated imaging biomarkers can predict individual outcomes.²¹ Some studies have suggested that ALPS impairment is correlated with poor ASM response,²² while others demonstrated partial GS recovery following resective epilepsy surgery.²³ These findings suggest that GS function, as measured by DTI-ALPS, may reflect underlying metabolic flexibility, network resilience, and neuroinflammatory burden, all of which may influence responsiveness to treatment. However, whether GS function can serve as a biomarker for VNS efficacy in DRE remains unknown.

This study compared ALPS-derived GS function between patients with DRE and HC and assessed the association between preoperative ALPS index and VNS treatment outcomes. We hypothesized that higher ALPS indices would be associated with better therapeutic responses, thereby establishing glymphatic integrity as a candidate biomarker for individualized neuromodulation planning.

Materials and methods

Participants

Forty patients with DRE who underwent VNS therapy at the First Affiliated Hospital of Anhui Medical University between June 2019 and April 2025 were retrospectively enrolled. The inclusion criteria were as follows: (1) Failure to achieve seizure control after at least 2 years of treatment with two or more appropriately selected and adequately dosed antiepileptic drugs, (2) presence of interictal epileptiform discharges confirmed by video-electroencephalography, (3) normal cognitive function indicated by mini-mental state examination scores before VNS implantation, and (4) all patients maintained a stable antiepileptic drug (AED) regimen for at least 4 weeks prior to MRI acquisition, in order to reduce the potential confounding influence of recent medication adjustments. Exclusion criteria included epilepsy due to treatable etiologies (e.g., metabolic, infectious, or structural abnormalities), bilateral vagus nerve damage, active local infection at or near the surgical site, intolerance to implanted devices, and comorbid systemic or neurological conditions that might interfere with treatment outcomes. In addition, individuals with overt structural brain abnormalities or significant white matter ischemic lesions on preoperative MRI were excluded to minimize potential confounding effects on DTI metrics. Age- and sex-matched HC data were derived from a previously established imaging database at our institution. These individuals had been recruited as healthy volunteers from the local community and hospital staff. All patient and control group participants provided written informed consent before participation. The enrollment and analysis workflow is illustrated in Figure 1(a).

Figure 1.

Schematic workflow and DTI-ALPS measurement framework. (a) The process included subject enrollment, MRI acquisition, VNS implantation, outcome assessment using the McHugh classification, and ALPS-based statistical modeling. (b) Representative DTI slice illustrating ROIs placement near the lateral ventricle. Inset: Magnified view identifying the projection ((1), blue), association ((2), green), and subcortical ((3), red) fiber ROIs. (c) Schematic diagram of perivascular space orientation and fiber tract alignment. Arrows represent the directionality of projection (blue), association (green), and subcortical (red) fibers relative to the perivascular space and principal diffusion axes (x, y, z).

MRI acquisition

All subjects underwent scanning using a Philips 3.0T scanner at the same center. The DTI sequence included 32 diffusion-encoding directions with a b-value of 800 s/mm², repetition time (TR) = 3500 ms, echo time (TE) = 83 ms, slice thickness = 2 mm, and field of view (FOV) = 160 × 160 mm. Besides DTI, high-resolution 3D T1-weighted imaging (3D-T1WI) and conventional axial T1WI, T2WI, and FLAIR sequences were acquired. Preoperative DTI data were collected within 1 week before VNS implantation. No major hardware or software upgrades occurred during the study period, ensuring methodological consistency across groups.

DTI processing

Raw diffusion data were preprocessed and conducted using MRtrix3 (Version 3.0.4; https://www.mrtrix.org) and FSL (Version 6.0.7; https://fsl.fmrib.ox.ac.uk/fsl).^24,25 The standard steps included denoising, Gibbs ringing artifact removal, eddy current and motion correction, and bias field correction. Diffusion tensor modeling was performed using FSL’s dtifit to generate maps of fractional anisotropy (FA), mean diffusivity, radial diffusivity, and axial diffusivity.

To enhance the reproducibility of ALPS index measurements and minimize head rotation artifacts, a reorientation technique was applied based on the method described by Tatekawa et al.²⁶ Briefly, FA map was linearly registered to the ICBM FA template using six degrees of freedom via FSL’s flirt, and the resulting transformation matrices were applied to the directionally encoded diffusivity maps (Dx, Dy, and Dz) using vecreg. This yielded reoriented diffusivity maps aligned to standard anatomical axes (ro-Dx, ro-Dy, and ro-Dz), improving intersubject comparability even in slight head misalignments during acquisition.

Regions of interest delineation and ALPS index calculation

Four spherical regions of interest (ROIs; diameter = 5 mm) were manually placed on reoriented color-FA maps using ITK-SNAP (Version 3.8.0; www.itksnap.org).²⁷ ROIs were bilaterally placed in the projection and association fiber regions adjacent to the lateral ventricles, based on standardized anatomical criteria. Placement was standardized at the level of the lateral ventricular body to ensure consistent visualization of periventricular white matter structures and stable anisotropy signals in the target fiber tracts. Projection fiber ROIs, corresponding primarily to the superior corona radiata, were positioned near the ventricular wall, with principal diffusivity oriented in the inferior to superior (z) direction. Association fiber ROIs, corresponding to the superior longitudinal fasciculus, were placed more laterally, with principal diffusivity along the anterior to posterior (y) direction. All ROIs were confined to homogeneous white matter areas to minimize partial volume effects and avoid inclusion of the ventricular cavity, gray matter, or regions with visible fiber crossings. A minimum distance of one voxel was maintained between the ROI boundary and the ventricular wall. Minor positional adjustments were permitted in cases of anatomical variation to preserve tract-specific localization and consistency of axial level across participants (Figure 1(b)). These ROIs extracted directional diffusivity values along the reoriented axes as x-, y-, and z-axes. Inter-rater reliability was assessed in a randomly selected subset of 20 participants using the intraclass correlation coefficient. All ALPS indices demonstrated good to excellent reliability, as detailed in Table S1. The reoriented ALPS (ro-ALPS) index was calculated using the following formula: ro-ALPS = (ro-Dx,proj &0x002B; ro-Dx,assoc)/(ro-Dy,proj &0x002B; ro-Dz,assoc). Where ro-Dx,proj, and ro-Dx,assoc denote diffusivity values in the right-left direction within the projection and association fibers, respectively; ro-Dy,proj, and ro-Dz,assoc denote diffusivity values in the anterior–posterior and inferior–superior directions, respectively (Figure 1(c)). ALPS indices were calculated separately for the left hemisphere (ALPS-L) and the right hemisphere (ALPS-R). Their average, referred to as the bilateral ALPS index (ALPS-B), was used in group comparisons and regression analyses. To assess hemispheric asymmetry in glymphatic function, we computed a lateralization index (LI) for each participant using the following formula: LI = (ALPS-R − ALPS-L)/(ALPS-R + ALPS-L).

Statistical analysis

All statistical analyses were conducted using the IBM SPSS Statistics for Windows, version 27.0 (IBM Corp., Armonk, NY, USA). A two-tailed p-value of <0.05 was considered indicative of statistical significance. Continuous variables with a normal distribution are expressed as mean ± standard deviation (SD) and compared using independent samples t tests, whereas categorical variables were evaluated using the Chi-square test. Between-group differences in demographic characteristics and ALPS indices between patients with DRE and HC were assessed using appropriate parametric or nonparametric tests, as dictated by data distribution. ALPS indices were further compared within the DRE cohort across McHugh outcome classifications I–V to explore potential gradational trends.²⁸ To assess the predictive value of ALPS indices for postoperative outcomes following VNS, we conducted both binary and ordinal logistic regression analyses. In the binary model, patients were categorized as responders (McHugh classes I–II) or nonresponders (classes III–V). In the ordinal model, the full five-level McHugh classification was used as the outcome variable to capture the gradation in treatment response. Both univariate and multivariate models were developed. The multivariate models included age, sex, and disease duration as covariates to adjust for potential confounding factors. Given the relatively small variance observed in raw ALPS indices, Z-score normalization was applied before regression modeling to enhance statistical stability and facilitate interindividual comparability. The standardization was performed using the formula: Z = (X − μ)/σ, where X represents the original ALPS index, μ denotes the mean, and σ represents the SD.

Results

Demographic characteristics

Forty patients with DRE and 30 HC were enrolled. There were nonsignificant differences in age (p = 0.949) or sex distribution (p = 0.441) between the two groups. Among patients with DRE, the mean number of AEDs at the time of VNS implantation was 2.650 ± 0.736. McHugh classification outcomes were distributed across classes I, II, III, IV, and V (Table 1).

Table 1.

Demographic and clinical characteristics of HC and DRE.

Variable	HC (n = 30)	DRE (n = 40)	t/χ²	p
Age (years)	26.233 ± 9.793	26.425 ± 14.122	0.064	0.949
Gender (male/female)	16/14	25/15	0.594	0.441
Disease duration (years)	/	9.250 ± 6.736	/	/
Number of AEDs (n)	/	2.650 ± 0.736	/	/
McHugh classification (n, %)
I		5 (12.5)
II		13 (32.5)
III		8 (20.0)
IV		11 (27.5)
V		3 (7.5)

p < 0.05 was considered statistically significant.

AEDs, antiepileptic drugs; DRE, drug-resistant epilepsy; HC, healthy control.

Comparison of ALPS metrics between DRE and HC groups

Compared with HC, patients with DRE exhibited significantly reduced ALPS-L (p < 0.001) and ALPS-B indices (p = 0.005). A nonsignificant difference was found in ALPS-R between groups (p = 0.386; Figure 2(a)). These findings indicate glymphatic dysfunction in DRE.

Figure 2.

ALPS indices in DRE and their association with VNS outcomes. (a) Violin plots illustrating differences between HC and DRE groups in ALPS-L, ALPS-R, and ALPS-B indices. ALPS-L and ALPS-B were significantly lower in the DRE group (p < 0.001 and p = 0.005, respectively), whereas nonsignificant difference was observed in ALPS-R (p = 0.386). (b) Violin plots comparing LI between HC and DRE groups. DRE group was significantly higher than HC group (p < 0.001). (c) Violin plots comparing ALPS indices between VNS responders and nonresponders among patients with DRE. Responders exhibited significantly higher ALPS-R (p = 0.021) and ALPS-B (p = 0.027) indices. ALPS-L exhibited a nonsignificant trend (p = 0.073). *p < 0.05, **p < 0.01, ***p < 0.001; ns indicates no statistical significance.

LI differences between groups

The LI was significantly greater in the DRE group than in HCs (p < 0.001), indicating a rightward asymmetry in glymphatic function. This finding suggests the presence of a hemispheric imbalance in patients with epilepsy (Figure 2(b)).

ALPS metrics between responders and nonresponders

To examine the relationship between preoperative ALPS indices and VNS efficacy, patients with DRE were classified as responders (McHugh classes I and II) or nonresponders (classes III–V). Independent samples t tests revealed that responders had significantly higher ALPS-R (p = 0.021) and ALPS-B indices (p = 0.027) than nonresponders. Although ALPS-L was also higher in responders, the difference did not reach statistical significance (p = 0.073). These findings suggest that increased glymphatic activity, particularly in the right and bilateral hemispheres, may be associated with improved VNS outcomes (Figure 2(c)).

Predictive value of ALPS metrics for VNS efficacy

Binary logistic regression was conducted using standardized (Z-scored) ALPS indices to further evaluate their prognostic value for VNS response. In univariate analysis (model 1), all three ALPS indices were positively associated with treatment efficacy. Although ALPS-L did not reach statistical significance (p = 0.082, odds ratio (OR) = 1.897), ALPS-R (p = 0.032, OR = 2.357) and ALPS-B (p = 0.039, OR = 2.284) were significantly associated with favorable outcomes. After adjusting for age, sex, and disease duration in the multivariate model (Model 2), all indices achieved statistical significance: ALPS-L (p = 0.028, OR = 2.739), ALPS-R (p = 0.020, OR = 2.802), and ALPS-B (p = 0.020, OR = 3.241; Table 2). These findings suggest that higher ALPS indices, particularly in the right and bilateral hemispheres, are independently associated with an increased likelihood of clinical response to VNS, supporting their potential utility as predictive imaging biomarkers.

Table 2.

Binary logistic regression analysis of DTI-ALPS index for predicting VNS response.

Projects	β	OR	p
Model 1
ALPS-L	0.640	1.897	0.082
ALPS-R	0.857	2.357	0.032
ALPS-B	0.826	2.284	0.039
Model 2
ALPS-L	1.008	2.739	0.028
ALPS-R	1.030	2.802	0.020
ALPS-B	1.176	3.241	0.020

p < 0.05 was considered statistically significant. Model 1, univariate analysis; Model 2, multivariate model adjusting for age, sex, and disease duration.

ALPS-B, ALPS Index of the bilateral hemisphere; ALPS-L, ALPS Index of the left hemisphere; ALPS-R, ALPS Index of the right hemisphere; DTI-ALPS, diffusion tensor image analysis along the perivascular space; OR, odds ratio; VNS, vagus nerve stimulation.

Ordinal logistic regression using the full five-level McHugh classification showed a consistent but nonsignificant association between higher ALPS-R values and better outcomes (OR = 0.58; p = 0.075). Although this may indicate a potential trend, ALPS-R, ALPS-L, and ALPS-B were all statistically nonsignificant. These results suggest that the effect of ALPS-R, if present, may reflect a threshold rather than a linear association across outcome levels (Table S2).

Discussion

This study assessed GS function in patients with DRE using DTI-ALPS method and explored its association with VNS outcomes. Compared to HC, patients with DRE exhibited significant reductions in ALPS indices, indicating glymphatic dysfunction. Significantly, higher preoperative ALPS indices, particularly in the right and bilateral hemispheres, were significantly associated with favorable VNS response. Both univariate and multivariate logistic regression analyses confirmed this association, suggesting that ALPS may serve as a noninvasive imaging biomarker to support individualized prediction of neuromodulatory efficacy in epilepsy.

This study found that ALPS index was significantly reduced in patients with DRE compared to HC, indicating widespread GS dysfunction in this population. This observation is consistent with findings in TLE, where reduced ALPS indices have been independently reported, suggesting altered perivascular clearance mechanisms in limbic regions vulnerable to epileptogenic activity.¹⁰ Moreover, similar impairments have been observed in patients with FLE, further supporting that GS disruption is not confined to mesial temporal structures but may extend to extratemporal cortical networks.¹¹ In pediatric populations, ALPS reductions have been documented in CAE¹² and RE,¹³ indicating that glymphatic dysfunction may emerge early in disease development, even without overt structural lesions. Together, these studies suggest that GS impairment is a shared pathophysiological hallmark across various epilepsy phenotypes. Rather than solely secondary to neuroanatomical damage, glymphatic disruption may reflect a broader vulnerability of interstitial clearance pathways intrinsic to epileptogenesis.²⁹ Mechanistically, this dysfunction may stem from seizure-related astrocytic activation, blood–brain barrier breakdown, and altered AQP4 expression, all of which can compromise CSF–ISF exchange and interstitial solute clearance.^30,31 From a clinical standpoint, impaired GS function may serve as an early biomarker of network instability and long-term structural vulnerability, particularly in patients with early-onset or pharmacoresistant epilepsy.

In this study, preoperative ALPS indices were significantly higher in patients who responded to VNS than in nonresponders, particularly in the right and bilateral hemispheres. These group differences reached statistical significance in independent t tests, indicating that glymphatic function may play a role in modulating neuromodulatory treatment outcomes. This finding strengthens prior observations in epilepsy research, where ALPS differences often exhibited directional trends across cognitive or surgical outcome measures despite not always reaching significance in smaller samples.^15,23 Our results suggest that glymphatic integrity may serve as a latent neurophysiological marker and a measurable predictor of VNS efficacy when assessed using appropriately powered binary frameworks.

Binary logistic regression analyses further confirmed the prognostic relevance of ALPS metrics. In univariate models, ALPS-R and ALPS-B exhibited significant associations with favorable VNS response, with OR exceeding 2.0. These associations became stronger and statistically significant in multivariate models adjusting for age, sex, and disease duration. ALPS-B demonstrated the highest predictive value (OR = 3.241, p = 0.020). Notably, the increase in OR in Model 2 suggests that ALPS exerts an independent predictive effect beyond clinical covariates influence. The suppression of the predictive value of ALPS in the unadjusted model likely reflects confounding factors such as age and disease duration, which impacts glymphatic function. Once these variables were controlled, the strength of ALPS-outcome association became more pronounced, reinforcing its value as a candidate biomarker for individual neuromodulatory responsiveness. Ordinal logistic regression using the full five-level McHugh classification showed a consistent negative association between ALPS-R and poorer outcome levels, although the result did not reach statistical significance. This may reflect limited statistical power due to small sample sizes in extreme outcome categories, which affected model stability. The observed trend may indicate a threshold effect, wherein glymphatic function above a certain level increases the likelihood of clinical benefit, rather than a continuous linear association across outcome strata.

Notably, although ALPS-R did not differ significantly between the DRE and HC groups, it emerged as a significant predictor of VNS responsiveness in both group comparisons and multivariate regression models. This apparent discrepancy reflects the distinct clinical questions underlying the analyses. The comparison between DRE and HC evaluates whether ALPS-R captures glymphatic dysfunction associated with disease presence, whereas the responder stratification assesses whether interindividual variability in ALPS-R relates to differential treatment outcomes. In this context, the predictive utility of ALPS-R may be more closely linked to preserved glymphatic reserve or compensatory functional integrity, rather than to the extent of pathological burden. Therefore, its lack of diagnostic group-level significance does not diminish its relevance as a biomarker of treatment responsiveness.

LI was analyzed to further clarify the predictive value of ALPS, as it reflects the relative distribution of glymphatic function between hemispheres. The DRE group exhibited significantly higher LI than HCs, indicating a rightward asymmetry in ALPS distribution. Rather than representing an isolated hemispheric imbalance, this asymmetry may help explain the specific association between ALPS-R and favorable VNS outcomes. The observed right-dominant LI could reflect a preserved glymphatic reserve in the right hemisphere that facilitates more effective neuromodulatory responsiveness, even in the presence of global dysfunction.

The known GS functions support the biological plausibility of these findings in metabolic clearance, ionic regulation, and neuroinflammatory control, all of which are critical for cortical excitability and neuromodulation. The amplified predictive effect after covariate adjustment reinforces prior evidence that glymphatic efficiency is shaped by aging and cumulative disease burden.^9,18 Consequently, while ALPS is not a standalone predictor, it appears to interact with patient-specific factors to influence treatment outcomes. This aligns with earlier findings in FCD, where poor ASM responders had lower ALPS indices,¹⁷ and in TLE, where ALPS indices correlated with seizure laterality and language impairment.^15,16 These data suggest that ALPS is a candidate biomarker of neurofunctional reserve relevant across pharmacological, surgical, and neuromodulatory epilepsy interventions.

Given that all three ALPS indices were significantly associated with VNS response in both group comparisons and multivariate regression analyses, our findings support the hypothesis that preserved glymphatic function may help identify patients more likely to benefit from neuromodulation. These results suggest that ALPS is a promising noninvasive imaging biomarker for presurgical risk stratification in DRE. Unlike conventional structural MRI or electrophysiological mapping, ALPS may reflect complementary aspects of brain health, including metabolic clearance and neuroinflammatory balance. If confirmed, integrating ALPS with clinical, structural, and cognitive markers could enhance precision in patient selection and treatment planning.

Limitations

This study has several limitations. First, the retrospective, single-center design may introduce selection bias and limit the generalizability of the findings. Second, the limited sample size, particularly within the DRE group (n = 40), reduces statistical power and may compromise the stability of regression model estimates, thereby increasing the risk of overfitting. Validation using larger, independent cohorts is essential to establish the robustness of these findings. Third, several physiological and molecular factors that may influence glymphatic function, including sleep quality, vascular pulsatility, and AQP-4 polarity, were not explicitly measured or statistically adjusted for in this study. Although all patients maintained stable AED regimens for at least 4 weeks prior to MRI, differences in drug type, dosage, and duration may have introduced residual effects on ALPS indices. Finally, while all DTI scans in this study were performed during standard daytime hours, previous research has shown that ALPS indices remain relatively stable throughout the day in awake individuals.^32,33 However, the exact scan time was not uniformly controlled. Future prospective studies incorporating multimodal imaging, sleep monitoring, and biomarker profiling are warranted to refine the physiological interpretation and clinical application of ALPS metrics in epilepsy care.

Conclusion

This study demonstrated that patients with DRE exhibit reduced ALPS indices, reflecting impaired glymphatic function. Higher ALPS values were significantly associated with favorable VNS outcomes, suggesting that glymphatic integrity may influence therapeutic efficacy. These findings support the potential of ALPS as a noninvasive imaging biomarker for presurgical evaluation and underscore the need for validation in larger, multimodal prospective studies.

Supplemental Material

sj-docx-1-tan-10.1177_17562864261436543 – Supplemental material for Glymphatic dysfunction in drug-resistant epilepsy and its association with vagus nerve stimulation outcomes: a DTI-ALPS study

Supplemental material, sj-docx-1-tan-10.1177_17562864261436543 for Glymphatic dysfunction in drug-resistant epilepsy and its association with vagus nerve stimulation outcomes: a DTI-ALPS study by Ziao Xu, Xiaofeng Wang, Bohang Liu, Xuefei Ji, Ming Shan, Hongwei Cheng and Lei Ye in Therapeutic Advances in Neurological Disorders

Footnotes

Acknowledgements

None.

Declarations

ORCID iDs

Ziao Xu

Lei Ye

Supplemental material

Supplemental material for this article is available online.

References

Holstein-Ronsbo

Gan

Giannetto

, et al. Glymphatic influx and clearance are accelerated by neurovascular coupling. Nat Neurosci 2023; 26: 1042–1053.

Murdock

Yang

Sun

, et al. Multisensory gamma stimulation promotes glymphatic clearance of amyloid. Nature 2024; 627: 149–156.

Ishida

Yamada

Nishiyama

, et al. Glymphatic system clears extracellular tau and protects from tau aggregation and neurodegeneration. J Exp Med 2022; 219: e20211275.

Chen

Liu

, et al. Effects of sleep on the glymphatic functioning and multimodal human brain network affecting memory in older adults. Mol Psychiatry 2025; 30: 1717–1729.

Kamagata

Andica

Takabayashi

, et al. Association of MRI indices of glymphatic system with amyloid deposition and cognition in mild cognitive impairment and Alzheimer disease. Neurology 2022; 99: e2648–e2660.

Wood

Nenert

Miften

, et al. Diffusion tensor imaging-along the perivascular-space index is associated with disease progression in Parkinson’s disease. Mov Disord 2024; 39: 1504–1513.

Bae

Kim

Choi

, et al. Glymphatic function assessment in Parkinson’s disease using diffusion tensor image analysis along the perivascular space. Parkinsonism Relat Disord 2023; 114: 105767.

Carotenuto

Cacciaguerra

Pagani

, et al. Glymphatic system impairment in multiple sclerosis: relation with brain damage and disability. Brain 2022; 145: 2785–2795.

Zhang

Zhou

Wang

, et al. Glymphatic clearance function in patients with cerebral small vessel disease. Neuroimage 2021; 238: 118257.

10.

Xie

Zhou

Jin

, et al. Bilateral glymphatic dysfunction and its association with disease duration in unilateral temporal lobe epilepsy patients with hippocampal sclerosis. Epilepsy Behav 2024; 155: 109777.

11.

Zhou

Peng

Yao

, et al. Association between glymphatic dysfunction and neurocognitive decline in patients with frontal lobe epilepsy. Quant Imaging Med Surg 2024; 14: 6745–6755.

12.

Wei

Qiu

, et al. Dysfunction of the glymphatic system in childhood absence epilepsy. Front Neurosci 2023; 17: 1312676.

13.

Yin

Wang

, et al. Unraveling the diffusion MRI-based glymphatic system alterations in children with Rolandic epilepsy. Acad Radiol 2025; 32: 1036–1042.

14.

Lee

Shin

, et al. Glymphatic system dysfunction in patients with juvenile myoclonic epilepsy. J Neurol 2022; 269: 2133–2139.

15.

Wang

Xia

Zhang

, et al. Association of glymphatic system dysfunction with cognitive impairment in temporal lobe epilepsy. Front Aging Neurosci 2024; 16: 1459580.

16.

Zhao

Zhou

, et al. The asymmetry of glymphatic system dysfunction in patients with temporal lobe epilepsy: a DTI-ALPS study. J Neuroradiol 2023; 50: 562–567.

17.

Jin

, et al. Glymphatic system dysfunction in epilepsy related to focal cortical dysplasia and its relationship with antiseizure medication response. Seizure 2025; 125: 31–36.

18.

Lee

Park

, et al. Glymphatic system function in patients with newly diagnosed focal epilepsy. Brain Behav 2022; 12: e2504.

19.

Gonzalez

HFJ

Yengo-Kahn

Englot

DJ.

Vagus nerve stimulation for the treatment of epilepsy. Neurosurg Clin N Am 2019; 30: 219–230.

20.

Tong

Wang

Qin

, et al. Vagus nerve stimulation for drug-resistant epilepsy induced by tuberous sclerosis complex. Epilepsy Behav 2022; 126: 108431.

21.

Clifford

Paranathala

Wang

, et al. Vagus nerve stimulation for epilepsy: a narrative review of factors predictive of response. Epilepsia 2024; 65: 3441–3456.

22.

Kim

Lee

, et al. Anti-seizure medication response and the glymphatic system in patients with focal epilepsy. Eur J Neurol 2024; 31: e16097.

23.

Zhang

, et al. Recovery of glymphatic system function in patients with temporal lobe epilepsy after surgery. Eur Radiol 2023; 33: 6116–6123.

24.

Tournier

Smith

Raffelt

, et al. MRtrix3: a fast, flexible and open software framework for medical image processing and visualisation. Neuroimage 2019; 202: 116137.

25.

Jenkinson

Beckmann

Behrens

, et al. FSL. Neuroimage 2012; 62: 782–790.

26.

Tatekawa

Matsushita

Ueda

, et al. Improved reproducibility of diffusion tensor image analysis along the perivascular space (DTI-ALPS) index: an analysis of reorientation technique of the OASIS-3 dataset. Jpn J Radiol 2023; 41: 393–400.

27.

Yushkevich

Piven

Hazlett

, et al. User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage 2006; 31: 1116–1128.

28.

McHugh

Singh

Phillips

, et al. Outcome measurement after vagal nerve stimulation therapy: proposal of a new classification. Epilepsia 2007; 48: 375–378.

29.

Sun

Peng

, et al. Enhancing glymphatic fluid transport by pan-adrenergic inhibition suppresses epileptogenesis in male mice. Nat Commun 2024; 15: 9600.

30.

Noe

Marchi

Central nervous system lymphatic unit, immunity, and epilepsy: is there a link?

Epilepsia Open 2019; 4: 30–39.

31.

Szu

Binder

DK.

Mechanisms underlying aquaporin-4 subcellular mislocalization in epilepsy. Front Cell Neurosci 2022; 16: 900588.

32.

Han

Zhou

Zhang

, et al. Age- and time-of-day dependence of glymphatic function in the human brain measured via two diffusion MRI methods. Front Aging Neurosci 2023; 15: 1173221.

33.

Taoka

Ito

Nakamichi

, et al. Diffusion tensor image analysis ALong the perivascular space (DTI-ALPS): revisiting the meaning and significance of the method. Magn Reson Med Sci 2024; 23: 268–290.