Alterations in white matter integrity in patients with postherpetic neuralgia evidenced by peak width of skeletonized mean diffusivity analysis

Abstract

Background

Postherpetic neuralgia (PHN) causes chronic neuropathic pain and reduced quality of life. The peak width of skeletonized mean diffusivity (PSMD), based on MD histogram along the fractional anisotropy skeleton, has shown robustness and clinical relevance in other neurological disorders.

Methods

This single-center cross-sectional study included 83 participants 42 patients with PHN and 41 age- and sex-matched healthy controls. All participants underwent diffusion tensor imaging (DTI), and data were analyzed using the FMRIB Software Library software. PSMD and conventional DTI metrics were computed and compared between groups.

Results

Patients with PHN exhibited marked increase in PSMD relative to healthy controls (2.862 ± 0.436 × 10^-4 vs. 2.369 ± 0.407 × 10^-4 mm²/s, P < 0.001). Additionally, MD and axial diffusivity (AD) were significantly elevated in the PHN group compared with controls (0.854 ± 0.035 vs. 0.836 ± 0.030, P = 0.012; 1.292 ± 0.036 vs. 1.264 ± 0.033, P < 0.001; respectively). Receiver operating characteristic analysis demonstrated that PSMD provided the greatest discriminative ability between groups (area under the curve = 0.816, P < 0.001). Furthermore, multivariable logistic regression including PSMD, MD, and AD identified PSMD as the sole independent predictor of PHN status (P = 0.002).

Conclusions

This study demonstrated that patients with PHN showed altered white matter integrity, with increased PSMD, MD, and AD values compared to healthy controls. PSMD showed superior diagnostic performance over most conventional DTI metrics and may serve as promising marker of widespread white matter disruption in PHN at the group level.

INTRODUCTION

Postherpetic neuralgia (PHN) is a chronic neuropathic pain syndrome that persists beyond the resolution of an acute herpes zoster rash and imposes a substantial burden on quality of life and healthcare utilization. PHN affects up to 30% patients with herpes zoster and represents one of the most common and debilitating chronic pain conditions in older adults [1–3]. Neuroimaging syntheses and primary studies have increasingly implicated central nervous system involvement in PHN, including alterations across structural and functional networks [4,5]. However, the specific loci and directionality of effects vary across cohorts and analytic pipelines, and no widely accepted imaging biomarker has been established for this condition [6–8]. Additionally, previous diffusion tensor imaging (DTI) studies in PHN have been limited by relatively small sample sizes and heterogeneous methodological approaches. This variability aligns with the broader uncertainties in PHN pathogenesis, which spans peripheral-central sensitization and neuroinflammatory mechanisms [9].

DTI is a standard method for quantifying white-matter (WM) microstructure; however, the conventional voxel-wise and region-of-interest approaches in DTI—most commonly tract-based spatial statistics (TBSS)—are limited by multiple-comparison burden and registration variability, which reduce its sensitivity to subtle, spatially diffuse abnormalities [10–12]. These limitations may partly explain the heterogeneous WM topographies reported in patients with PHN [6–8]. In other words, regional metrics alone can underrepresent the global burden of WM injury in this disorder [10,11,13].

The peak width of skeletonized mean diffusivity (PSMD) is a fully automated, histogram-based summary metric defined as the difference between the 95th and 5th percentiles of MD values sampled along the fractional anisotropy (FA) skeleton [14–16]. PSMD captures the dispersion of diffusivity across the brain’s principal WM tracts on a standardized skeleton, thereby indexing subtle and widespread injuries that regional analyses may overlook. Since its introduction, PSMD has shown robust associations with cerebral small-vessel disease burden and cognitive performance—particularly processing speed—in community-dwelling aging cohorts [15,16]. Moreover, PSMD has been instrumentally and biologically validated for reproducibility and cross-scanner standardization, supporting its suitability for multi-site clinical research [17–19]. Notably, the single-value summary of PSMD overcomes the multiple comparison limitation inherent in voxel-wise analyses and provides a standardized metric suitable for clinical decision-making and biomarker development. Taken together, these data support PSMD as a globally sensitive and cross-scanner-robust summary metric that mitigates analytic heterogeneity (e.g., regional dependence, multiple-comparison burden) by indexing diffuse WM injury on a single standardized scale [20]. This is particularly relevant to PHN, where microstructural abnormalities are spatially heterogeneous and variably localized.

Currently, PHN diagnosis relies primarily on clinical history and subjective pain reports, highlighting the potential value of objective imaging markers. A standardized imaging measure may facilitate characterization of white matter alterations and provide insights into the neural substrates of PHN, although further validation is required before clinical application.

Building on this rationale and representing the first application of the PSMD methodology in patients with PHN, this case-control study had three a priori aims: (1) to determine whether PSMD is elevated in patients with PHN relative to healthy controls; (2) to compare the diagnostic discrimination of PSMD with conventional DTI metrics (FA, MD, axial diffusivity [AD], and radial diffusivity [RD]); and (3) to examine the associations between PSMD and clinically relevant measures. Based on previous literature, the authors hypothesized that patients with PHN would exhibit increased PSMD accompanied by an MD/AD-predominant diffusion signature with relative preservation of FA/RD, and that PSMD would provide incremental value for case-control discrimination beyond conventional indices.

MATERIALS AND METHODS

The protocol was approved by the Institutional Review Board (IRB) of Inje University Haeundae Paik Hospital, approval number IRB-2023-10-030, and adhered to the Declaration of Helsinki and applicable regulations. All participants provided written informed consent. Fig. 1 illustrates the process used in this study.

Fig. 1.

Study workflow. The figure illustrates the overall workflow of this study. The icons used in this figure were generated with the assistance of ChatGPT (OpenAI). PHN: postherpetic neuralgia, NRS: numeric rating scale, DN4: Douleur Neuropathique en 4, BPI: brief pain inventory, MRI: magnetic resonance imaging, DTI: diffusion tensor imaging, PSMD: peak width of skeletonized mean diffusivity, FA: fractional anisotropy, MD: mean diffusivity, AD: axial diffusivity, RD: radial diffusivity, ROC: receiver operating characteristic.

1. Study participants

A total of 42 patients with PHN were recruited from the pain clinic of a university-affiliated hospital. A diagnosis of PHN was established when patients reported persistent pain localized to the dermatomal distribution of a prior herpes zoster rash, persisting for more than 3 months after rash onset [21]. Eligible patients were aged ≥ 18 years and capable of providing informed consent. The exclusion criteria were (1) a history of significant neurological or psychiatric disease (e.g., stroke, multiple sclerosis, or brain tumor), (2) uncontrolled systemic illness, which referred to severe or unstable medical conditions that could interfere with study participation or magnetic resonance imaging (MRI) acquisition; controlled hypertension or diabetes mellitus per se were not exclusion criteria, (3) contraindications to brain MRI scans (e.g., pacemaker, metallic implants, claustrophobia), and (4) concurrent chronic pain conditions unrelated to herpes zoster. The disease duration for patients with PHN was calculated as the time interval between herpes zoster rash onset and MRI acquisition (recorded in months). Additionally, 41 healthy control participants, matched for age and sex, were recruited from the surrounding community. They were confirmed to have no history of neuropathic pain or major neurological diseases, and no clinically significant structural brain lesions on MRI. Structural MRI scans were formally reviewed by a board-certified neuroradiologist with more than 10 years of clinical experience. Participants with focal structural brain lesions, including silent lacunar infarctions, tumors, or other major abnormalities, were excluded. Age-related WM changes such as WM hyperintensities were not used as exclusion criteria.

The severity of PHN symptoms for patients with PHN was assessed using the Douleur Neuropathique en 4 (DN4) score, a validated measure of neuropathic pain, consisting of 10 items assessing sensory descriptors and clinical examination findings [22]. In addition, the numeric rating scale (NRS) was used to assess current pain intensity, and the brief pain inventory (BPI) [23] was administered to evaluate pain-related functional interference.

2. DTI and T1-weighted images acquisition

Patients with PHN and healthy controls underwent DTI using a 3.0T MRI scanner equipped with a 32-channel head coil. DTI was performed using spin-echo single-shot echo-planar pulse sequences across 32 different diffusion directions (repetition time [TR]/echo time [TE] = 8,620/85 ms, FA = 90°, slice thickness = 2.25 mm, acquisition matrix = 120 × 120, field of view = 240 × 240 mm², and b value = 1,000 s/mm²). The three-dimensional T1-weighted images were scanned (inversion time, 1,300 ms; TR/TE, 8.6/3.96 ms; flip angle, 8°; and isotropic voxel size, 1 mm³). The image quality was assessed by visual inspection of motion artifacts, and datasets with excessive head movements were excluded from the analysis.

3. PSMD acquisition

PSMD values were derived from the DTI data using the FMRIB Software Library on a Linux platform, following a standardized four-step process [15,22];

1) Data was preprocessed including motion and eddy current correction, brain extraction, and tensor fitting of raw DTI data.

2) Skeletonization was performed, which aligns the FA maps to a standardized space and projects them onto a WM framework using TBSS. The same transformation matrices were then applied to the MD maps to generate skeletonized maps.

3) A custom mask with a threshold FA value of 0.3 was applied to the skeletonized MD maps to focus analyses on relevant WM tracts.

4) Histogram analysis was performed by generating histograms of the MD values from all voxels within the skeleton. The PSMD was calculated as the difference between the 95th and 5th percentiles of the histogram distribution.

PSMD values were calculated separately for the whole brain and right and left hemispheres.

4. FA, MD, AD, and RD analysis

Conventional DTI metrics, including FA, MD, AD, and RD, were computed using the DSI Studio software package. DTI data were imported in the DICOM format, and a brain mask was applied using thresholding, smoothing, and defragmentation steps. Reconstruction and fiber tracking were then performed. Whole-brain seeding was conducted with 10,000 seeds, using an angular threshold of 60°. Tracks with lengths < 30 mm were discarded. An Automated Anatomical Labeling template was used for brain parcellation, and every WM fiber was evaluated at the extreme points. Finally, conventional DTI measures including FA, MD, AD, and RD values were calculated.

5. White matter hypointensity volume analysis

To segment WM hypointensity from T1-weighted images and acquire WM hypointensity volume, the authors used White Matter Hyperintensities (WMH)-Synthseg, which provides segmentation for WM hypointensity from scans of any resolution and contrast without retraining, available as module in the development version of FreeSurfer. The WM hypointensity volume was corrected with total intracranial volume (WM hypointensity volume/total intracranial volume x 100).

6. Statistical analysis

Data normality was assessed using the Shapiro–Wilk test. Categorical variables were analyzed using the chi-square test or Fisher’s exact test, and continuous variables were tested using the independent samples t-test or Mann–Whitney U-test according to normal distribution. Correlations were assessed using Pearson’s test.

Receiver operating characteristic (ROC) curve analyses were performed for each parameter to evaluate the discriminatory ability of PSMD and conventional DTI metrics (FA, MD, AD, and RD) to differentiate patients with PHN from healthy controls. The area under the curve (AUC) was then calculated to quantify the diagnostic performance, with higher AUC values indicating better discriminatory power. Binary logistic regression analysis was then performed, with group status as the dependent variable, to further evaluate the discriminatory power of the diffusion metrics. Only parameters showing significant between-group differences in the univariate analysis were included as independent variables. The model was constructed using the entry method and model fit was assessed by the –2 log-likelihood statistic, chi-square test, and pseudo-R² values (Cox & Snell and Nagelkerke). The Hosmer–Lemeshow goodness-of-fit test was used to examine the calibration of the model. To account for the potential influence of WM hypointensity volume on PSMD, an analysis of covariance (ANCOVA) was performed with PSMD as the dependent variable, group (PHN vs. healthy controls) as the fixed factor, and WM hypointensity volume as a covariate. This analysis was conducted to determine whether the between-group difference in PSMD remained significant after adjusting for differences in WM hypointensity volume. Adjusted mean PSMD values were estimated for each group while controlling for the covariate. All statistical analyses were performed using MedCalc® Statistical Software version 23.2.7 (2025, MedCalc Software, Ltd.; https://www.medcalc.org). A P value less than 0.05 was considered statistically significant. Categorical variables are presented as frequency (percentage), and continuous variables as mean value ± standard deviation or median (interquartile range, IQR) depending on distribution.

7. Sample size

A formal a priori sample size calculation was performed using G*Power (v3.1.9.4) [24]. The authors assumed α = 0.05, power = 0.80, allocation ratio = 1:1, and a conservative medium effect size (Cohen’s d = 0.60) for the primary analysis, a two-sided independent-samples t-test comparing PSMD between groups. The resulting target size was n = 45 per group (total n = 90), allowing approximately 10% attrition; therefore, they planned to enroll n = 50 per group (total n = 100). The analytical sample consisted of 42 patients with PHN and 41 controls. Accordingly, the authors emphasized effect sizes with 95% confidence intervals (CIs) rather than post-hoc power calculations. ROC analyses were prespecified as complementary assessments of diagnostic discrimination.

RESULTS

1. Clinical characteristics

This study included 42 patients with PHN and 41 healthy controls. No significant between-group differences were observed in age (70.4 ± 10.3 vs. 69.3 ± 6.5 years, P = 0.595) or sex distribution (male: 76.2% vs. 78.0%, P = 0.841). Among patients with PHN, the median NRS score was 45 (IQR: 30–60). The involved nerve areas were located on the right and left sides in 42.9% and 57.1% patients, respectively. The thoracic region was most commonly involved (50.0%), followed by the cranial (35.7%), cervical (11.9%), and lumbar (2.4%) regions. The median DN4 and BPI scores for patients with PHN were 5 (IQR: 4–6) and 25 (IQR: 17–33), respectively. All patients were prescribed gamma-aminobutyric acid (GABA)ergic drugs. In addition, 45.2% of the patients received antidepressants, 54.8% received opioids, and 21.4% were treated with non-steroidal anti-inflammatory drugs (Table 1).

Table 1.

Clinical characteristics of patients with PHN and healthy controls

	Patients with PHN (n = 42)	Healthy control (n = 41)	P value
Age (yr)	70.4 ± 10.3	69.3 ± 6.5	0.595
Male	32 (76.2)	32 (78.0)	0.841
Disease duration (mo)	19.1 [5.2–53.1]	-	-
NRS	45 [30-60]	-	-
Side of involvement	Right: 18 (42.9)/ Left: 24 (57.1)	-	-
Involved nerve area			-
Cranial area	15 (35.7)	-
Cervical area	5 (11.9)	-
Thoracic area	21 (50.0)	-
Lumbar area	1 (2.4)	-
DN4	5 [4–6]	-	-
BPI	25 [17–33]	-	-
Medication			-
GABAergic drug	42 (100.0)	-
Antidepressant	19 (45.2)	-
Opioid	23 (54.8)	-
NSAID	9 (21.4)	-

Values are presented as mean ± standard deviation, number (%), or median [interquartile range Q1, Q3].

PHN: postherpetic neuralgia, NRS: numeric rating scale, DN4: Douleur Neuropathique en 4, BPI: brief pain inventory, NSAID: non-steroidal anti-inflammatory drug, -: none.

2. Group difference in PSMD

The PSMD was significantly higher in patients with PHN compared to healthy controls (2.862 ± 0.436×10^-4 vs. 2.369 ± 0.407×10^-4 mm²/s, P < 0.001) (Fig. 2). However, no significant difference was observed between the ipsilateral and the contralateral side of the prior herpes zoster rash (2.809 ± 0.464 × 10^-4 vs. 2.804 ± 0.462 × 10^-4 mm²/s, P = 0.961).

Fig. 2.

Comparison of PSMD between patients with PHN and healthy controls. The figure shows that PSMD is significantly higher in patients with PHN compared to that in healthy controls (2.862 ± 0.436 × 10^-4 vs. 2.369 ± 0.407 × 10^-4 mm²/s, P < 0.001). The box plots show the median (horizontal line), interquartile range (box), and minimum–maximum values (whiskers). Individual data points are overlaid as circles. PSMD: peak width of skeletonized mean diffusivity, PHN: postherpetic neuralgia.

After adjusting for WM hypointensity volume using ANCOVA, the group difference in PSMD remained significant (F[1,80] = 17.10, P < 0.001). The adjusted mean PSMD was 2.78 ×10^-4 mm²/s in patients with PHN and 2.46 ×10^-4 mm²/s in healthy controls. The effect size was moderate to large (partial η² = 0.176), indicating that PHN status accounted for a substantial proportion of variance in PSMD independent of WM hypointensity volume. To evaluate the potential influence of cranial herpes zoster, PSMD values were compared between patients with cranial and non-cranial zoster involvement. No significant difference was observed between the two groups (2.866 ×10^-4 vs. 2.859 ×10^-4 mm²/s, P = 0.961).

To examine the potential influence of medications, subgroup analyses were performed within the PHN group according to opioid and antidepressant use. No significant differences in PSMD were observed between patients receiving opioids and those not receiving opioids (2.767 ×10^-4 vs. 2.976 ×10^-4 mm²/s, P = 0.124). Similarly, PSMD did not differ between patients treated with antidepressants and those not receiving antidepressants (2.777 ×10^-4 vs. 2.932 ×10^-4 mm²/s, P = 0.255).

3. Group difference in conventional DTI metrics

Significant differences were observed between the MD and AD subgroups. Both MD and AD were higher in patients with PHN compared to healthy controls (0.854 ± 0.035 vs. 0.836 ± 0.030, P = 0.012; 1.292 ± 0.036 vs. 1.264 ± 0.033, P < 0.001; respectively). However, no significant difference was observed in the FA (0.449 ± 0.021 vs. 0.444 ± 0.014, P = 0.215) and RD (0.635 ± 0.037 vs. 0.622 ± 0.030, P = 0.079) between patients with PHN and healthy controls.

4. Group difference in white matter hypointensity volumes

The corrected WM hypointensity volumes were higher in patients with PHN than in the healthy controls (0.218% ± 0.206% vs. 0.094% ± 0.164%, P = 0.003).

5. ROC curve analysis and Logistic regression

ROC curve analysis was performed to evaluate the discriminatory power of diffusion metrics (Fig. 3). PSMD showed the highest diagnostic performance with an AUC of 0.816 (95% CI: 0.716–0.893, P < 0.001). In contrast, FA (AUC = 0.562, 95% CI: 0.448–0.670, P = 0.347) demonstrated poor discriminative ability, while MD (AUC = 0.685, 95% CI: 0.573–0.782, P = 0.001), AD (AUC = 0.754, 95% CI: 0.648–0.842, P < 0.001), and RD (AUC = 0.621, 95% CI: 0.508–0.726, P = 0.053) showed moderate performance. Pairwise comparisons revealed that PSMD performed significantly better than FA (P = 0.005), MD (P = 0.006), and RD (P < 0.001), whereas the difference between PSMD and AD was not statistically significant (P = 0.157). Among the other indices, AD outperformed MD (P = 0.007) and RD (P = 0.001). These findings suggest that PSMD provides superior diagnostic accuracy, compared with most conventional diffusion parameters, in differentiating patients with PHN from controls. The optimal cutoff values derived from the ROC analyses were 3 × 10^–4 for PSMD, 0.465 for FA, 0.844 for MD, 1.272 for AD, and 0.643 for RD.

Fig. 3.

ROC curve analysis. The figure indicates that PSMD showed the highest diagnostic performance, with an AUC of 0.816 (95% CI: 0.716–0.893, P < 0.001). In contrast, FA (AUC = 0.562, 95% CI: 0.448–0.670, P = 0.347) demonstrates poor discriminative ability, while MD (AUC = 0.685, 95% CI: 0.573–0.782, P = 0.001), AD (AUC = 0.754, 95% CI: 0.648–0.842, P < 0.001), and RD (AUC = 0.621, 95% CI: 0.508–0.726, P = 0.053) show moderate performance. ROC: receiver operating characteristic, PSMD: peak width of skeletonized mean diffusivity, AUC: area under the curve, CI: confidence interval, FA: fractional anisotropy, MD: mean diffusivity, AD: axial diffusivity, RD: radial diffusivity.

Subsequently, logistic regression analysis was performed to assess the predictive value of the diffusion parameters for differentiating patients with PHN from controls. The overall model significantly improved fit over the null model (χ² = 28.3, df = 3, P < 0.001), with Cox & Snell R² = 0.289 and Nagelkerke R² = 0.385, indicating a moderate explanatory power. Among the predictors, PSMD emerged as a significant independent factor (β = –28,972.53, Wald = 9.00, P = 0.002), while MD showed a trend toward significance (β = 39.27, P = 0.066). However, AD did not reach statistical significance (β = –32.81, P = 0.136).

6. Correlations between PSMD and clinical characteristics

In patients with PHN, PSMD was positively correlated with age (r = 0.721, P < 0.001; Fig. 4). However, PSMD was not significantly correlated with the DN4 score (r = –0.156, P = 0.324), NRS score (r = –0.091, P = 0.567), BPI score (r = –0.153, P = 0.334), or disease duration (r = 0.058, P = 0.713).

Fig. 4.

The results of the correlations analysis. A positive correlation is evident between age and PSMD in patients with PHN (r = 0.721, P < 0.001). PSMD: peak width of skeletonized mean diffusivity, PHN: postherpetic neuralgia.

DISCUSSION

This age- and sex-adjusted case-control study provides convergent evidence that PHN is associated with a diffuse WM burden, as summarized by a single global index. Notably, the PSMD in patients with PHN was higher than that in matched healthy controls. Additionally, across the ROC curve analyses, PSMD showed the strongest case–control discrimination among the tested metrics and increased with age, as expected. Furthermore, hemispheric PSMD (symptomatic vs. contralateral) in patients with PHN did not differ significantly, consistent with non-lateralized diffuse WM involvement.

These observations are concordant with previous PHN neuroimaging studies, which have documented heterogeneous spatial patterns of structural and functional abnormalities across cohorts and analytic pipelines with no consensus biomarker being identified [6–8,25]. Instead of convergent tract-specific effects, the data of the present study indicate a global signature indexed by PSMD—an automated skeleton-based histogram metric that quantifies diffuse WM injury on a standardized scale [16,26]. Beyond PHN, PSMD has also been linked to cerebral small-vessel disease burden and cognitive performance in community-dwelling older adults, thus serving as a marker of widespread WM vulnerability [16,26,27–30]. Moreover, PSMD has been validated as a vascular contributor to cognitive impairment and a dementia biomarker and demonstrated cross-scanner reproducibility and standardization, supporting its use in multi-site studies [19,20,30,31]. Evidence from cerebral amyloid angiopathy further reinforces the sensitivity of PSMD to diffuse injury across cerebrovascular phenotypes [28]. Collectively, these data suggest that the sensitivity of PSMD stems from indexing the dispersion of diffusivity along core WM pathways—an advantageous feature when abnormalities are widespread but individually subtle.

Several design and analytical features strengthen the internal validity of this interpretation. First, skeleton-based projections reduce partial-volume effects and registration variability inherent to voxel-wise or tractography summaries, limiting spurious dispersion signals [10,11,32,33]. Second, previous evidence of cross-scanner reproducibility for PSMD supports the view that the observed group contrast is not specific to a single site or instrument [19,20]. Third, the agreement between non-lateralized PSMD increases and previous reports of spatially heterogeneous abnormalities is compatible with a diffuse, network-level burden rather than a tract-specific effect, although a non-specific global process cannot be excluded [6–8].

Mechanistically, a nonlateralized increase in PSMD suggests a broad microstructural compromise rather than lesion-side-restricted degeneration. The absence of hemispheric asymmetry in PSMD should be interpreted with caution. While this finding may reflect diffuse WM involvement in PHN, alternative explanations should be considered. Chronic neuropathic pain has been associated with large-scale brain network reorganization rather than focal changes limited to specific nociceptive pathways [6,34]. In addition, systemic neuroinflammatory responses following herpes zoster infection may contribute to widespread microstructural alterations in the brain [9,34,35]. Finally, vascular mechanisms may also partly contribute to the global PSMD signal, as PSMD has been validated as a marker of diffuse WM injury related to cerebral small vessel disease [16,26,27]. Together, these mechanisms may contribute to the non-lateralized PSMD increase observed in patients with PHN. In addition, the exploratory subgroup analysis by zoster location did not show evidence that cranial involvement disproportionately drove the PSMD increase, although more subtle vascular contributions cannot be excluded. Potential contributors to PHN include deafferentation-related central reorganization, neuroinflammatory processes, and microvascular stress [9,34,35], all of which can also increase the diffusivity variance across the skeleton without focal FA loss. While the present study’s design cannot adjudicate mechanisms, the observed MD/AD-predominant changes with relative preservation of FA/RD are consistent with diffuse, non-tract-specific alterations reported in neuropathic pain states. These interpretations remain provisional and motivate multimodal studies that combine diffusion, perfusion/vascular markers, and neuroinflammatory assays [6–8].

Clinically, the superiority of PSMD over most conventional diffusion metrics supports its use as a concise screening signal for trial enrichment (selecting patients with PHN and a greater global WM burden) rather than as a stand-alone diagnostic. Methodologically, the cross-scanner standardization and high repeatability of PSMD make it a practical endpoint for multisite designs [19,20]. Additionally, the absence of hemispheric asymmetry favors whole-brain summary measures, when effects are expected to be diffuse, over symptom-side analyses alone. Finally, the spatial sampling strategy, contrasting PSMD (skeleton-based) with tractography-derived summaries, highlights how masking and aggregation choices affect sensitivity, and harmonizing or reporting both improves interpretability across studies. In practice, age-standardized PSMD z-scores and pre-specified enrichment thresholds (e.g., percentile cut-offs) facilitate trial screening, while paired reporting of discrimination and calibration (AUC with CIs and calibration curves) clarifies clinical utility [36].

One notable finding of the present study was the absence of significant correlations between PSMD and clinical measures of neuropathic pain severity (DN4, NRS, and BPI) as well as disease duration. This observation suggests that PSMD may not directly reflect the subjective intensity of pain or the temporal progression of PHN. Instead, PSMD may represent a more stable index of global WM vulnerability rather than a state-dependent marker of symptom severity; however, because of the cross-sectional design, the temporal ordering of these alterations cannot be determined. This interpretation is consistent with previous studies investigating PSMD in aging and cerebrovascular disease. PSMD has been shown to reflect diffuse microstructural WM damage across major tracts and is strongly associated with overall small vessel disease burden and cognitive processing speed rather than specific neurological symptoms [16,26,27]. In community-dwelling older adults, PSMD has been proposed as a robust marker of global WM integrity and brain frailty, reflecting cumulative microstructural alterations that may predispose individuals to neurological dysfunction [28,30]. Similarly, PSMD elevations have been reported across various neurological conditions characterized by widespread WM vulnerability, including cerebral amyloid angiopathy and REM sleep behavior disorder [17,28,29]. From this perspective, elevated PSMD in PHN may reflect underlying WM vulnerability or diffuse microstructural alterations associated with chronic pain, rather than pain severity itself. However, the cross-sectional design precludes determining whether these alterations preceded PHN or resulted from persistent neuropathic pain. Thus, PSMD should be interpreted as a marker of global WM burden associated with PHN rather than a direct biomarker of pain severity or disease progression. Longitudinal studies beginning in the acute herpes zoster phase are needed to clarify the temporal relationship between WM alterations and PHN.

Nevertheless, this study has some limitations. First, the single-center sampling and modest sample size of this study constrain precision and limit the stability of the null findings for conventional metrics. Additionally, the cross-sectional design precludes inferences regarding temporal dynamics or treatment responses. Second, the authors did not comprehensively assess cognition, quantify WM hyperintensities, or capture detailed lifestyle/vascular risk profiles. In addition, formal WMH grading, lacunes, microbleeds, and clinical vascular risk factors such as hypertension, diabetes mellitus, and dyslipidemia were not systematically modeled or directly compared between groups. These factors could influence diffusion metrics; therefore, residual vascular confounding remains possible and further prospective studies are warranted. Third, although exploratory subgroup analyses according to opioid and antidepressant use did not show significant differences in PSMD within the PHN group, medication-related confounding cannot be fully excluded because dose, duration, and cumulative exposure were not modeled, and the potential influence of universal GABAergic drug use could not be examined. Fourth, differences in spatial sampling between PSMD (skeleton-based) and conventional metrics (atlas/tractography-based) may also contribute to apparent sensitivity gaps. Lastly, another limitation of this study is the absence of a comparison group of patients with acute herpes zoster who did not develop PHN. Therefore, it remains unclear whether the elevated PSMD observed in the PHN group reflects mechanisms specific to persistent neuropathic pain or more general effects related to the herpes zoster event itself, such as systemic inflammation, neuroinflammation, or vascular stress [37]. Future studies including patients with acute herpes zoster and prospective follow-up will be required to determine whether PSMD alterations are specific to PHN development.

Consequently, future studies should: (1) establish test–retest reliability and minimal detectable change for PSMD in patients with PHN; (2) evaluate longitudinal trajectories relative to pain persistence, flare, and recovery; (3) integrate PSMD with clinical measures and blood-based biomarkers of neuroinflammation or microvascular injury; and (4) compare PSMD-guided trial enrichment against traditional eligibility to test whether it improves power or predicts response to mechanism-based therapies.

In conclusion, this study demonstrated that patients with PHN exhibited significant alterations in WM integrity, as evidenced by increased PSMD, MD, and AD values relative to healthy controls. PSMD appears to be a useful group-level imaging marker reflecting global WM alterations associated with PHN.