Impacts of Pain Characteristics on the Outcomes of Short-Term Spinal Cord Stimulation for the Treatment of Post-Herpetic Neuralgia: A Prospective Self-Controlled Study

Abstract

Background

Herpes zoster and postherpetic neuralgia are increasingly prevalent and associated with substantial pain burden and reduced quality of life. Despite the increasing use of neuromodulation strategies such as stSCS, evidence on their effects across pain phenotypes remains limited.

Aim

To investigate the heterogeneous pain characteristics of postherpetic neuralgia and to explore how different pain phenotypes respond to stSCS.

Design

A prospective self-controlled study.

Methods

LCA was used to identify different pain phenotypes, followed by LMM to evaluate differential responses after stSCS in pain, anxiety, ADL and sleep. The report was guided by the STROBE checklist.

Results

A total of 207 patients were included. StSCS was effective in improving pain, anxiety and sleep status in patients with PHN, although a reduction in ADL was also observed. LCA identified two distinct pain phenotypes: Typical irritative phenotype (Class 1) and Atypical high-burden phenotype (Class 2). Differential treatment responses were observed between the two phenotypes, with Class 1 showing greater pain relief and sleep improvement, whereas Class 2 exhibited a higher symptom burden and a comparatively weaker response to stSCS.

Conclusion

This study confirms the clinical efficacy of stSCS in patients with PHN and demonstrates heterogeneity in treatment response across distinct pain phenotypes. These findings underscore the importance of pre-treatment pain phenotype stratification to inform individualized stSCS-based management strategies and support the integration of a multidisciplinary approach for high-burden phenotypes to optimize therapeutic outcomes.

Background

Herpes zoster is a common infectious disease caused by reactivation of the varicella-zoster virus (Varicella-zoster virus, VZV), with a global incidence that has been steadily increasing in recent years. It affects approximately 3–5 per 1000 individuals annually, rising to 8–12 per 1000 among those aged over 50 years.¹ In China, the prevalence is approximately 7.7%, and nearly 29.8% of affected patients subsequently develop postherpetic neuralgia (PHN).² PHN is typically defined as pain persisting for more than 1 months after the resolution of the herpes zoster rash. Its pathophysiology is primarily associated with peripheral and central nervous system injury following viral reactivation. Damaged nerves can lead to ectopic discharges, repetitive stimulation, and increased neuronal excitability, thereby promoting the development and maintenance of peripheral and central sensitization, ultimately resulting in chronic neuropathic pain. Clinically, PHN is characterized by substantial heterogeneity, including persistent spontaneous pain, allodynia, paroxysmal severe pain, and various sensory abnormalities.³ Moreover, symptom burden varies considerably among individuals and is often accompanied by depression, anxiety, sleep disturbances, and reduced physical function, which collectively lead to a significant impairment in quality of life.⁴

Although numerous publications address PHN, evidence for the treatment of PHN remains limited, and therapeutic strategies primarily aim at symptom relief.⁵ Advances in neuromodulation have provided novel approaches for managing neuropathic pain. Spinal cord stimulation (SCS) is a neuromodulation technique in which electrical pulses generated by an implanted pulse generator are delivered to epidural electrodes, selectively stimulating the dorsal columns of the spinal cord. This process modulates pain signal transmission and produces paresthesia in the painful area, and has been demonstrated to provide analgesic effects in various chronic neuropathic pain conditions.⁶ Based on stimulation duration and implantation strategy, SCS can be classified into short-term spinal cord stimulation (stSCS) and permanent spinal cord stimulation. StSCS is typically administered for 7–14 days with subsequent electrode removal, whereas permanent SCS involves long-term implantation to achieve sustained analgesia.⁷ Although permanent SCS offers durable pain relief, its clinical application is limited by relatively strict indications and high costs. In contrast, stSCS is a minimally invasive, easily performed, and repeatable neuromodulation approach that has been increasingly applied for rapid control of acute and subacute pain. It has shown promising clinical potential and is recommended for the management of PHN.^8,9

StSCS has been shown to achieve favorable outcomes in PHN treatment. For example, 77.8% of PHN patients receiving stSCS experienced an average pain reduction of 57%.¹⁰ In one-month follow-up, stSCS provided significant pain relief, with quality-of-life improvements persisting at 6–12 months, especially in sleep, mood, and life enjoyment.¹¹ Additionally, stSCS has been found to improve overall health, social functioning, role emotional, mental health, bodily pain, physical functioning, role physical, and vitality.¹² However, most existing studies have focused on overall treatment efficacy, with limited investigation into whether patients with different pain phenotypes respond differently to stSCS. Pain characteristics may influence the effectiveness of electrical stimulation, as patients without hyperalgesia or with pain hypersensitivity benefit less from stSCS.¹³ Considering the individualized variability in pain presentation, it is necessary to explore the effects of stSCS across different pain phenotypes to provide a more targeted basis for differential treatment.

Latent class analysis (LCA), an unsupervised statistical method, has been widely applied in studies of chronic pain, mental health, and sleep to identify hidden subgroups with similar symptom patterns.¹⁴ Compared with traditional stratification methods relying on a single variable, LCA can automatically identify latent classes based on multidimensional self-reported data, making it more suitable for studies of complex and co-occurring symptom phenotypes. Recently, LCA has been increasingly used to identify differential effects in intervention studies.¹⁵

The Study

Based on the above, the present study aimed to investigate the heterogeneous pain characteristics of PHN and to explore how different pain phenotypes respond to stSCS in terms of pain, anxiety, ADL, and sleep.

Methods

Study Design

This study was a prospective self-controlled study and was guided by the STROBE checklist.

Participant

Convenience sampling was used to recruit patients with PHN who were hospitalized at Changde First People’s Hospital between January and November 2025.

Inclusion criteria: (1) The diagnosis of PHN meets the following criteria (pain persisting for more than 30 days after the resolution of rash). The diagnosis must be confirmed through a combination of patient history, clinical examination, and laboratory test results;¹⁶ (2) Age ≥18 years; (3) Undergoing stSCS treatment with a treatment duration of at least 7 days; (4) Able to provide informed consent, understand the purpose and procedures of the study, and cooperate with follow-up questionnaires and laboratory assessments, with clear verbal communication. Exclusion criteria: (1) Presence of severe systemic diseases, including cardiovascular, neurological, or endocrine disorders, or malignant tumors; (2) Presence of psychiatric disorders or intellectual disabilities that preclude cooperation; (3) Those who are concurrently participating in other interventional clinical studies.

Preoperatively, patients received individualized pharmacological treatment according to the Expert Consensus on Full-Course Management of Herpes Zoster-Associated Pain,¹⁷ primarily including pregabalin 150 mg or 300 mg, alone or in combination with acetaminophen and dihydrocodeine 40 mg. Regarding spinal cord stimulation, according to the Expert Consensus on Spinal Cord Stimulation for Chronic Pain,¹⁸ conventional low-frequency SCS uses a pulse frequency below 1200 Hz, and high-frequency SCS uses 5–10 kHz.

In this study, stSCS electrode implantation was performed in the operating room by pain physicians. The epidural electrode was inserted at the target spinal level via a percutaneous puncture under local anesthesia. The electrodes used were 8-contact spinal cord stimulation needle electrodes or nerve stimulation test electrodes (Model 3873) manufactured by Changzhou Ruiwansheng Medical Technology Co., Ltd. After the procedure, patients were instructed to remain on strict bed rest for 2 days, and to keep the puncture site clean and dry with sterile dressing to prevent infection. During stSCS treatment, low-frequency spinal cord stimulation was applied, with stimulation parameters typically set at a frequency of 50–300 Hz, pulse width of 60–360 μs, and current intensity of 0.6–15 mA, which adjusted gradually to achieve optimal analgesic effect. The following presents representative case reports of cervical and thoracic stSCS procedures.

Case 1: Cervical Segment stSCS

A 72-year-old male presented with a 5-month history of PHN involving the right neck, shoulder, and upper limb. On September 23, 2025, DSA was used in the operating room to localize the target intervertebral foramen and plan the optimal puncture trajectory. After marking the entry point and routine sterilization, percutaneous puncture was performed at the C7/T1 interspace under local anesthesia with DSA guidance. An 8-contact spinal cord stimulation electrode (Changzhou Ruiwansheng Medical Technology Co., Ltd.) was advanced into the epidural space, and its position was adjusted so that contacts 0–1 were placed near the right C2 dorsal root ganglion. Intraoperative stimulation parameters were adjusted based on patient feedback to ensure coverage of the painful area. The electrode was then secured and connected to an external pulse generator, Figure 1. Postoperatively, the patient remained on strict bed rest for 2 days and was instructed to keep the cervical dressing clean and dry and avoid excessive neck movement. Stimulation parameters were dynamically adjusted during follow-up. On October 10, 2025, the patient reported significant pain relief, and the electrode was removed prior to discharge.

Figure 1.

Intraoperative imaging of stSCS (Cervical).

Case 2: Thoracic Segment stSCS

A 69-year-old female presented with a 2-year history of PHN involving the left thoracic back, lumbar region, and abdomen. On August 11, 2025, DSA thin-slice imaging was used in the operating room to localize the target intervertebral foramen and plan the optimal puncture trajectory. After marking the entry point and routine sterilization, percutaneous puncture was performed at the T11/12 interspace under DSA guidance with local anesthesia. An 8-contact spinal cord stimulation electrode (Changzhou Ruiwansheng Medical Technology Co., Ltd.) was advanced into the epidural space, and its position was adjusted so that contacts 2–3 were placed near the dorsal root ganglion, with the electrode tip positioned at approximately the C4 level. Intraoperative stimulation parameters were adjusted based on patient feedback to ensure adequate pain coverage. The electrode was then secured and connected to an external pulse generator, Figure 2. Postoperatively, the patient remained on strict bed rest for 2 days and was instructed to avoid spinal movement and keep the dressing clean and dry. Stimulation parameters were dynamically adjusted. On August 22, 2025, significant pain relief was observed, and the electrode was removed prior to discharge.

Figure 2.

Intraoperative imaging of stSCS (Thoracic).

Data Collection

Instruments

• (1)Demographic questionnaire: Based on relevant literature, a questionnaire was developed to collect nine indicators, including sex, age, education level, smoking history, body mass index, and pain location, etc.
• (2)Self-reported pain assessment: Based on repeated discussions within the research team, PHN was characterized in a multidimensional manner encompassing pain quality, temporal dynamics, attack features, and severity. Accordingly, four self-reported indicators were included: pain type (dull/sharp), pain pattern (intermittent/continuous), breakthrough pain (BTP) frequency (<5/≥5), and pain intensity (VAS score).
• (3)Generalized Anxiety Disorder-7 (GAD-7): Developed by Spitzer et al, the GAD-7 is a seven-item scale used to screen for generalized anxiety disorder and assess symptom severity. Each item is rated on a 4-point Likert scale (0 = not at all, 1 = several days, 2 = more than half the days, 3 = nearly every day), with total scores ranging from 0 to 21, where higher scores indicate greater anxiety.¹⁹ In Chinese general hospital outpatients, the GAD-7 demonstrated good reliability and validity, with a cutoff of 10 yielding a sensitivity of 86.2%, specificity of 95.5%, and kappa = 0.825.²⁰
• (4)Activities of Daily Living (ADL) scale: Originally developed by Mahoney and Barthel and adapted for the Chinese population in 2007, the ADL scale includes 10 items, with a total score ranging from 0 to 100, where higher scores indicate better self-care ability.^21,22
• (5)Pittsburgh Sleep Quality Index (PSQI): Developed by Buysse et al, the PSQI evaluates sleep quality over the past month, encompassing 18 items across seven dimensions, including subjective sleep quality, sleep duration, and sleep latency. Higher total scores indicate poorer sleep quality.²³ For the Chinese population, a PSQI total score >7 differentiates good and The test-retest reliability of PSQI was 0.994. The split-half reliability coefficient of PSQI was 0.824, and the overall Cronbach’s α coefficient was 0.845.²⁴

Procedure

Data were collected before and after the stSCS procedure during hospitalization. Baseline data (T0) were collected within 24 hours of admission, and post-treatment data (T1) were collected on the day of discharge. Demographic information was extracted from the hospital information system, while self-reported pain, GAD-7, ADL, and PSQI data were collected by the responsible nurse.

Data Analysis

Statistical analyses were performed using SPSS 26.0 and Mplus 8.3. Categorical variables were described using counts and percentages (%), and continuous variables were summarized as mean ± standard deviation (Mean ± SD). The data analysis included four steps. ①The paired t-test was employed to verify differences before and after stSCS treatment. ②The four items of self-reported pain were recoded as categorical variable for statistical convenience (Pain scores were categorized into mild, moderate, and severe groups based on the following ranges: 0–3, 4–7, and 8–10, respectively). Latent class models with 1 to 5 classes were iteratively constructed, and model fit was evaluated. Model fit was assessed using Akaike’s Information Criterion (AIC), Bayesian Information Criterion (BIC), and adjusted BIC (aBIC), where lower values indicate better fit, and entropy (0–1), where higher values indicate greater classification accuracy. The Lo-Mendell-Rubin Likelihood Ratio Test (LMRT) and the Bootstrapped Likelihood Ratio Test (BLRT) were used to compare model performance, with p < 0.05 indicating that the K-class model was significantly better than the K-1 class model.²⁵ ③Based on the identified latent classes, Linear Mixed Models (LMM) were employed as the primary analytical approach to investigate the longitudinal trajectories of outcome variables across different classes. ④To test the robustness of the primary findings, Generalized Estimating Equations (GEE) with robust standard errors were further conducted as a sensitivity analysis. Consistency between the results derived from the LMMs and GEEs would indicate robust associations between the latent classes and longitudinal outcomes. All statistical tests were two-sided, and a P-value of < 0.05 was considered.

Quality Control

During data collection, all responsible nurses involved in the study were trained to ensure the use of standardized instructions, questionnaire explanations, and data recording procedures. After participants completed the questionnaires, researchers immediately checked the data for completeness and anomalies, such as out-of-range values or inconsistent responses. Any issues identified were addressed on-site by politely confirming or completing the information with the participant. During data entry and analysis, all questionnaire data were independently double-checked by two researchers. Questionnaires exhibiting patterned responses, logical inconsistencies, or more than 20% missing data were excluded. Additionally, a statistical expert was consulted to review and refine the analytical methods employed in the study.

Ethical Considerations

This study was approved by the Ethics Committee of Changde First People’s Hospital (No. 2025-123-01) and was conducted in accordance with the Declaration of Helsinki. Prior to data collection, researchers verbally explained the study objectives and participants’ right to withdraw at any time. Written informed consent was obtained before the initial data collection. To protect participants’ privacy, all subjects were assigned numeric codes, and all data were securely stored by designated personnel to ensure confidentiality and data protection.

Result

Participant Characteristics

Between January and November 2025, a total of 250 patients were initially enrolled in the study. Among them, 6 patients withdrew during the study, 10 patients refused to participate in the second data collection, and 27 datasets were excluded due to missing data, Figure 3. Ultimately, 207 patients were included in the final analysis, Table 1. The cohort was predominantly aged 70–79 years, with primary education or below, married, non-smokers, normal body mass index (BMI), and mainly experiencing trunk pain. Prior antiviral treatment and combination therapy were commonly used. The duration from rash onset to hospital admission was 55.97±83.43 days. The stSCS treatment duration was 10.23±2.52 days and the hospital duration was 12.55±2.69 days.

Figure 3.

Study Participant Flow Diagram.

Table 1.

Baseline Characteristics of Patients (n = 207)

Item	Quantity (Number)	Proportion (%)
Age (years)
50–59	24	11.59
60–69	61	29.47
70–79	86	41.55
80–89	36	17.39
Gender
Male	101	48.79
Female	106	51.21
Education level
Primary school or below	77	37.20
Junior high school	58	28.02
Senior high school	52	25.12
College or above	20	9.66
Marital status
Unmarried	2	0.97
Married	130	62.80
Divorced	5	2.42
Widowed	70	33.81
Smoking history
No	114	55.07
Yes	93	44.93
BMI
Underweight	72	34.71
Normal	101	48.79
Overweight	30	14.49
Obese	4	1.93
Pain location
Trunk	196	94.69
Limbs	3	1.45
Mix	8	3.86
Pain Management Pattern
None	23	11.1
Monotherapy	43	20.8
Combination Therapy	141	68.1
Prior antiviral treatment
No	66	31.88
Yes	141	68.12
Duration from rash onset to hospital admission	55.97±83.43	–
StSCS treatment duration	10.23±2.52	–
Hospital duration	12.55±2.69	–

Notes: Pain management pattern was classified as None, Monotherapy, and Combination Therapy. Combination Therapy primarily included pregabalin (150–300 mg), alone or in combination with acetaminophen and dihydrocodeine (40 mg).

Pre-Post Changes

After stSCS treatment, patients exhibited reduced pain (t=22.446, p<0.001), alleviated anxiety symptoms (t=13.445, p<0.001), and improved sleep quality (t=11.295, p < 0.001). However, activities of daily living (ADL) scores showed a decrease (t=2.090, p= 0.038), Table 2.

Table 2.

Changes of VAS, GAD-7, ADL and PSQI After stSCS (T0-T1, N = 207)

Item	Mean±SD	95% CI	t	p
VAS	−2.94±1.88	2.679, 3.195	22.446	< 0.001**
GAD-7	−2.85±3.05	2.432, 3.268	13.445	< 0.001**
ADL	−3.42±23.55	0.194, 6.647	2.090	0.038*
PSQI	−3.56±4.54	2.939, 4.182	11.295	< 0.001**

Notes: *p<0.05, **p<0.001.

Abbreviations: SD, standard deviation; CI, confidence interval; T0, pre-treatment; T1, post-treatment.

LCA Results

The model fitting results and LCA stratification results are shown in Table 3 and Figure 4. Based on the following criteria, the two-class model was selected as the optimal solution: (a) The two-class model demonstrated the lowest AIC and aBIC values compared with the other models (AIC=1014.148, aBIC=1015.626); (b) The two-class model demonstrates a high classification accuracy rate (Entropy = 1.000), and both the LMRT and BLRT were statistically significant (p < 0.001), indicating a significantly better fit than the one-class model; (c) The lowest class probability was 20.8%, confirming the stability of the class assignment.

Table 3.

LCA Model Results

Model	LL	AIC	BIC	ABIC	Entropy	LMRT	BLRT	Lowest Class Probability (%)
						p	p
1	−508.565	1025.130	1038.461	1025.787	–	–	–	–
2	−498.074	1014.148	1044.142	1015.626	1.000	< 0.001**	< 0.001**	20.8
3	−493.075	1014.150	1060.808	1016.449	0.597	0.0859	0.096	21.9
4	−491.102	1020.203	1083.525	1023.325	0.755	0.2602	0.275	10.4
5	−491.011	1030.022	1110.007	1033.965	0.637	0.6565	0.660	11.1

Notes: Bold values indicate the retained 2-class solution. **P < 0.01.

Figure 4.

LCA Results Plot.

Class 1 (n=164, 79.2%, Typical irritative phenotype) was characterized by a high probability of irritative pain quality, a pronounced temporal pattern, and frequent breakthrough pain episodes, whereas the probability of severe pain intensity was relatively low. This class included a relatively balanced sex distribution and was predominantly composed of patients aged 70–79 years, with primary school education or below, married status, a history of smoking, normal BMI, trunk pain distribution, and receipt of combination analgesic and antiviral therapy. Class 2 (n=43, 20.8%, Atypical high-burden phenotype) showed lower probabilities of pain pattern and BTP frequency and was not characterized by typical irritative pain features, but demonstrated a higher likelihood of severe pain intensity. Similar demographic and clinical characteristics were observed in this class, including balanced sex distribution, predominance of patients aged 70–79 years, primary school education or below, married status, smoking history, normal BMI, trunk pain distribution, and use of combination analgesic and antiviral therapy, Table 4.

Table 4.

Demographic Distribution of the Two Classes

Items	Class 1 (164, 79.2%)	Class 2 (43, 20.8%)	χ2/F	p
Age (years)			4.503a	0.212
50–59	18 (11.0)	6 (14.0)
60–69	52 (31.7)	9 (20.9)
70–79	63 (38.4)	23 (53.5)
80–89	31 (18.9)	5 (11.6)
Gender			0.000a	0.995
Male	80 (48.8)	21 (48.8)
Female	84 (51.2)	22 (51.2)
Education level			0.746a	0.862
Primary school or below	62 (37.8)	15 (34.9)
Junior high school	45 (27.4)	13 (30.2)
Senior high school	40 (24.4)	12 (27.9)
College or above	17 (10.4)	3 (7.0)
Marital status			3.810a	0.283
Unmarried	1 (0.6)	1 (2.3)
Married,	106 (64.6)	24 (55.8)
Divorced	5 (3.0)	0 (0)
Widowed	52 (31.8)	18 (41.9)
Smoking history			0.055a	0.815
No	91 (55.5)	23 (53.5)
Yes	73 (44.5)	20 (46.5)
BMI			1.318a	0.725
Underweight	56 (34.2)	16 (37.2)
Normal	81 (49.4)	20 (46.5)
Overweight	23 (14.0)	7 (16.3)
Obese	4 (2.4)	0 (0)
Pain location			0.390a	0.823
Trunk	156 (95.1)	40 (93.0)
Limbs	2 (1.2)	1 (2.3)
Mix	6 (3.7)	2 (4.7)
Pain Management Pattern			6.920a	0.031*
None	18 (10.9)	5 (11.6)
Monotherapy	28 (17.1)	15 (34.9)
Combination Therapy	118 (72.0)	23 (53.5)
Prior antiviral treatment			0.709a	0.400
No	50 (30.5)	16 (37.2)
Yes	114 (69.5)	27 (62.8)
Duration from rash onset to hospital admission	53.26±73.40	66.33±114.40	0.835b	0.362
StSCS treatment duration	10.11±2.47	10.67±2.71	1.716b	0.192
Hospital duration	12.65±2.64	12.16±2.91	1.098b	0.296
VAS	7.30±1.45	6.74±1.29	5.182b	0.024*
GAD-7	9.65±1.92	8.56±1.88	11.12b	<0.001**
ADL	83.88±15.87	85.05±14.02	0.193b	0.661
PSQI	8.62±3.60	9.26±5.91	0.785b	0.377

Notes: ^achi-square test; ^banalysis of variance. Pain management pattern was classified as None, Monotherapy, and Combination Therapy. Combination Therapy primarily included pregabalin (150–300 mg), alone or in combination with acetaminophen and dihydrocodeine (40 mg). Class 1: Typical irritative phenotype (n=164); Class 2: Atypical high-burden phenotype (n=43). *p<0.05, **p<0.001.

Class Differences in Outcomes

LMM were constructed to compare changes in pain, anxiety, sleep quality, and ADL from pre‑surgery (T0) to post‑surgery (T1) between the two latent classes, with adjustment for baseline pain, baseline anxiety, and pain management pattern. Both classes showed improvements in pain (Class 1: −3.091, SE=0.145, p<0.001; Class 2: −2.349, SE=0.284, p<0.001) and anxiety (Class 1: −3.146, SE=0.234, p<0.001; Class 2: −1.721, SE=0.458, p<0.001), with Class 1 greater reductions in both outcomes (between-group differences: 0.742 and 1.425, respectively; class × time: both p<0.001). For sleep, both classes showed improvement (Class 1:-3.549, SE=0.355, p<0.001; Class 2: −3.605, SE=0.693, p<0.001), the between-group difference was small (0.056, SE=0.780), which may indicate limited clinical significance. No significant changes were observed for ADL in either class (class × time: p=0.086), Table 5.

Table 5.

LMM Results for VAS, GAD-7, ADL, and PSQI After stSCS Treatment

Items	Class	T0 (Mean ± SD)	T1 (Mean ± SD)	T0→T1 Estimate (SE)	Between-Group Estimate (SE)	Class × Time
						F	p
VAS	Class 1	7.30±1.45	4.21±1.63	−3.091 (0.145)**	0.742 (0.148)*	260.013	<0.001
	Class 2	6.74±1.296	4.39±1.56	−2.349 (0.284)**
GAD-7	Class 1	9.65±1.92	6.51±3.17	−3.146 (0.234)**	1.425 (0.514)*	97.157	<0.001
	Class 2	8.56±1.888	6.84±2.79	−1.721 (0.458**
ADL	Class 1	83.88±15.87	81.11±18.16	−2.768 (1.840)**	3.139 (4.039)	2.482	0.086
	Class 2	85.05±14.02	79.13±17.61	−5.907 (3.594)
PSQI	Class 1	8.62±3.60	5.07±2.73	−3.549 (0.355)**	0.056 (0.780)	63.483	<0.001
	Class 2	9.26±5.91	5.65±2.76	−3.605 (0.693)**

Note: Data are presented as mean ± SD or estimate (SE). Within-group change and between-group difference were derived from linear mixed models adjusted for baseline pain, baseline anxiety, and pain management pattern. Negative values indicate a decrease in scores (improvement for pain and anxiety; deterioration for ADL). Class 1: Typical irritative phenotype (n=164); Class 2: Atypical high-burden phenotype (n=43). *p<0.05, **p<0.001.

Abbreviations: SD, standard deviation; SE, standard error.

Sensitivity Analysis

The results from the GEE and LMM models were consistent, showing that Class 1 achieved significantly greater improvements in pain and anxiety than Class 2. This finding was stable across different analytical methods, indicating the robustness of the results. The statistical significance of sleep differed between LMM (p<0.001) and GEE (p<0.947), likely due to differences in the assumptions underlying each method: LMM requires normal distribution of residuals, whereas GEE does not impose this stringent requirement, Table 6.

Table 6.

GEE Results for VAS, GAD-7, ADL, and PSQI After stSCS Treatment

Items	B (SE)	Wald χ2	p	95% CI
VAS
Intercept	4.395 (0.235)	349.250	<0.001**	3.934, 4.856
Group (Class 1 vs Class 2)	−0.188 (0.267)	0.496	0.481	−0.712, 0.335
Time (T0 vs T1)	2.349 (0.281)	69.981	<0.001**	1.799, 2.899
Group × Time	0.743 (0.316)	5.522	0.019*	0.123, 1.362
GAD-7
Intercept	6.837 (0.420)	265.255	<0.001**	6.014, 7.660
Group (Class 1 vs Class 2)	−0.331 (0.487)	0.462	0.497	−1.286, 0.624
Time (T0 vs T1)	1.721 (0.422)	16.662	<0.001**	0.895, 2.547
Group × Time	1.425 (0.484)	8.675	0.003*	0.477, 2.374
ADL
Intercept	79.140 (2.654)	888.943	<0.001**	73.937, 84.342
Group (Class 1 vs Class 2)	1.970 (3.007)	0.429	0.512	−3.924, 7.864
Time (T0 vs T1)	5.907 (3.059)	3.730	0.053	−0.088, 11.902
Group × Time	−3.139 (3.598)	0.761	0.383	−10.191, 3.913
PSQI
Intercept	5.651 (0.416)	184.662	<0.001**	4.836, 6.466
Group (Class 1 vs Class 2)	−0.578 (0.467)	1.530	0.216	−1.494, 0.338
Time (T0 vs T1)	3.605 (0.767)	22.117	<0.001**	2.102, 5.107
Group × Time	−0.056 (0.840)	0.004	0.947	−1.701, 1.589

Notes: Data are presented as B (SE) with 95% CI. Models were adjusted for baseline pain, baseline anxiety, and pain management pattern. The interaction term Group × Time represents the between-group difference in change from T0 to T1. Class 1: Typical irritative phenotype (n=164); Class 2: Atypical high-burden phenotype (n=43). *p<0.05, **p<0.001.

Abbreviations: GEE, generalized estimating equations; SE, standard error; CI, confidence interval.

Discussion

Principal Findings

This prospective self-controlled study suggests that stSCS can alleviate pain, anxiety, and sleep disturbances, but reduce ADL during hospitalization. LCA identified two distinct subgroups, revealing the heterogeneity and underlying patterns of PHN pain phenotypes: a typical irritative phenotype (Class 1) and an atypical high-burden phenotype (Class 2). StSCS demonstrated differential modulatory effects across pain phenotypes, with more pronounced improvements in pain and anxiety observed in Class 1 patients, whereas its effects on sleep and activities of daily living were relatively limited.

Interpretation of Findings

After stSCS treatment, patients experienced improvements in pain, anxiety, and sleep, consistent with previous studies.²⁶ Pain relief is a well-established clinical effect of stSCS. Spinal cord stimulation can reduce pain sensitivity and suppress abnormal discharge by stabilizing aberrant sodium channel activity, decreasing ectopic firing, and modulating the dorsal root ganglion.²⁷ By alleviating pain, stSCS also reduces pain-induced emotional burden. Research has shown that patients with chronic pain have a threefold higher risk of being diagnosed with anxiety or depression compared to the general population.²⁸ Therefore, stSCS can improve patients’ anxiety levels by attenuating pain-triggered emotional reactivity. In addition to reducing pain sensitivity, stSCS improves anxiety by decreasing pain-related threat perception. This study also observed improvements in sleep quality, likely due to the combined effects of pain reduction and anxiety alleviation. Previous research has demonstrated a bidirectional relationship among pain, anxiety, and sleep disturbances: pain can cause difficulty falling asleep, anxiety further fragments sleep, and poor sleep can exacerbate pain.^29,30 Consequently, when stSCS alleviates pain and anxiety, patients achieve more stable sleep in the short term. Moreover, our findings also indicate that stSCS was associated with an reduction in ADL scores in patients with PHN, suggesting a transient decline in patients’ self-care ability in the short term. This may be related to increased dependency resulting from the requirement for strict bed rest or restricted spinal mobility following stSCS implantation.

Although no significant differences were observed in baseline characteristics between the two groups, results showed that stSCS produced greater analgesic and anxiolytic effects in Class 1 patients. This suggests that treatment response may be driven mainly by underlying pain phenotypes. Neuropathic pain is highly heterogeneous. Different PHN phenotypes show different contributions of peripheral and central sensitization, which may affect neuromodulation response.^31,32 Class 1 was characterized by irritative pain, temporal fluctuation, and breakthrough pain. This pattern is consistent with peripheral sensitization and increased dorsal horn excitability. Peripheral sensitization involves reduced nociceptor thresholds and enhanced peripheral input. Spinal synaptic plasticity further amplifies stimulus-dependent pain. This phenotype shows limited evidence of widespread central reorganization.^33,34 Therefore, pain is more dependent on peripheral input and is more variable. In this context, stSCS may reduce pain by modulating dorsal horn excitability and enhancing Aβ-mediated inhibitory gating. This reduces abnormal afferent amplification and stabilizes pain fluctuations.^35,36 Reduced pain variability may also decrease unpredictability, thereby lowering threat appraisal and improving anxiety.³⁷ In contrast, Class 2 showed persistent high-intensity pain, suggesting dominant central sensitization. Central sensitization involves sustained hyperexcitability in spinal and supraspinal networks. It leads to pain spread, allodynia, and dissociation from peripheral stimuli.^34,38 This phenotype may be less responsive to stSCS due to stronger central network maintenance mechanisms.

Clinical Implication

These findings suggest that pain phenotype stratification should be performed prior to stSCS treatment in patients with PHN to optimize clinical decision-making. In patients with a typical irritative phenotype, stSCS may serve as an effective primary analgesic intervention. In contrast, for patients with an atypical high-burden phenotype, neuromodulation alone may be insufficient to address the complexity of their symptom profile. A multidisciplinary approach should therefore be integrated during hospitalization, including optimization of analgesic pharmacotherapy, sleep management, and psychological support, in order to reduce overall symptom burden and improve treatment experience.

Limitation

This study has several limitations. First, individual differences in analgesic use, including drug type, dosage, and administration frequency, may have introduced confounding effects on pain, sleep, and anxiety outcomes. Although sensitivity analyses were conducted to examine potential biases from medication use, the individualized and dynamically adjusted nature of clinical analgesic management makes it impossible to fully exclude residual confounding. Future research should employ standardized and quantified analgesic protocols in randomized controlled trials (RCTs) to further validate the findings. Second, the characterization of pain phenotypes was based on only four self-reported pain indicators, which may not fully capture the complexity and multidimensional nature of PHN pain. This limited feature set may reduce the granularity of phenotypic classification. Future studies should incorporate additional pain-related dimensions to enrich phenotype characterization and improve clinical interpretability. Third, this study focused on the acute effects of stSCS during hospitalization and only compared short-term outcomes before and after the procedure, without long-term follow-up on treatment responses across different pain phenotypes. Future studies are needed to evaluate its long-term efficacy and stability.

Conclusion

This prospective self-controlled study suggests that stSCS yields differential therapeutic responses across distinct PHN pain phenotypes. The results demonstrated improvements in pain, anxiety, and sleep status during stSCS treatment, accompanied by a reduction in ADL. Based on the two pain phenotypes identified by LCA, typical irritative (Class 1) and atypical high-burden phenotypes (Class 2), marked heterogeneity in treatment response was observed. Patients with the typical irritative phenotype experienced greater pain relief, whereas those with the atypical high-burden phenotype exhibited a higher overall symptom burden and a poorer response to stSCS. These findings highlight the importance of pre-treatment pain phenotype stratification to guide individualized stSCS-based management strategies in PHN, and support a multidisciplinary approach for high-burden phenotypes to optimize overall therapeutic outcomes.

Funding Statement

This study was supported by The science and technology innovation Program of Changde city (No. 2024-110-02) and Hunan Provincial Natural Science Foundation Of China (No. 2026JJ81685).

Data Sharing Statement

Participants were guaranteed confidentiality regarding any personally identifiable information collected during the research process. The raw data can be obtained from the corresponding author, Yalan Bu.

Author Contributions

Rong Yuan: Data curation; Formal analysis; Investigation; Methodology; Software; Writing - original draft; Writing - review & editing. Yan Jiang: Data curation; Formal analysis; Investigation; Methodology; Software; Writing - original draft; Writing - review & editing. Hong Hu: Data curation; Formal analysis; Funding acquisition; Investigation; Project administration. Pei Liu: Data curation; Investigation. Zhonghui Deng: Conceptualization; Funding acquisition; Project administration; Supervision; Writing - review & editing. Yalan Bu: Conceptualization; Funding acquisition; Project administration; Supervision; Writing - review & editing. All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work.