Comparison of Acupuncture and Lumbar Traction for Lumbar Disc Herniation and Subsequent Risk of Lumbar Fusion or Discectomy: A Retrospective Cohort Study

Abstract

Background

Lumbar disc herniation (LDH) is commonly treated with conservative therapies, but comparative evidence on acupuncture versus lumbar traction and subsequent surgery is limited. We examined whether acupuncture was associated with lower 5-year risks of discectomy or spinal fusion than traction among adults with LDH.

Methods

We conducted a retrospective cohort study using deidentified electronic health records from the TriNetX US Collaborative Network from January 1, 2000, to December 31, 2023. Adults aged 20 years or older with LDH who received acupuncture or traction after diagnosis were included. We excluded patients with prior lumbar surgery, lumbar or pelvic fracture, vertebral malignancy, lumbar disc infection or treatment crossover. Propensity score matching was performed 1:1 for demographics, lifestyle factors, comorbidities, health care utilization and body mass index. Primary outcomes were 5-year discectomy and spinal fusion. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Results

After matching, 3045 patients were included in each group and baseline characteristics were balanced. Discectomy occurred in 77 patients receiving acupuncture and 175 receiving traction (2.5% v. 5.7%; hazard ratio [HR] 0.46, 95% confidence interval [CI] 0.35–0.60; absolute risk reduction [ARR] 3.2%; number needed to treat [NNT] 31). Spinal fusion occurred in 128 patients receiving acupuncture and 188 receiving traction (4.2% v. 6.2%; HR 0.73, 95% CI 0.58–0.91; ARR 2.0%; NNT 51).

Interpretation

Among adults with LDH, receipt of acupuncture was associated with lower observed 5-year risks of subsequent discectomy and spinal fusion compared with lumbar traction. These associations should be interpreted cautiously because residual confounding, including differences in concomitant treatments and unmeasured disease severity, cannot be excluded. Prospective studies are needed to determine whether acupuncture itself influences subsequent surgical utilization.

Plain Language Summary

Lumbar disc herniation can cause back pain and radiating leg symptoms. Most patients initially receive nonsurgical treatment, but some eventually undergo operations such as discectomy or spinal fusion. Acupuncture and lumbar traction are both conservative treatments, although little is known about whether patients receiving these treatments differ in their subsequent likelihood of undergoing surgery.

In this retrospective study, we analyzed deidentified electronic health record data from the TriNetX US Collaborative Network. After matching patients according to demographic and clinical characteristics, 3,045 patients receiving acupuncture were compared with 3,045 receiving lumbar traction. Patients in the acupuncture group had lower observed rates of discectomy and spinal fusion during the following five years.

Because this was an observational study, these results do not prove that acupuncture itself prevents surgery. Patients may have received other treatments, and differences in disease severity, treatment preferences, imaging findings, or physician recommendations could not be completely measured. Prospective studies are therefore needed to determine whether acupuncture itself influences subsequent surgical treatment.

Introduction

Lumbar disc herniation (LDH) is a common cause of low back pain and lumbar radiculopathy. Low back pain affects approximately 80% of individuals during their lifetime, and although many patients with LDH remain asymptomatic, about 1%–3% develop symptoms, most often between 30 and 50 years of age.¹ Symptomatic LDH commonly presents with low back pain and radiating leg symptoms and, in more severe cases, may cause sensory disturbance, motor weakness, or functional limitation. These symptoms can substantially impair quality of life and contribute to health care use and related expenditures.^2,3 Most patients are initially treated nonoperatively, including pharmacotherapy, physical therapy, acupuncture, traction, and other rehabilitation-based approaches. However, some patients have persistent or progressive symptoms and eventually require surgery, such as discectomy or spinal fusion.⁴ Although surgery can provide symptom relief in selected patients, it is associated with potential risks, including recurrent herniation, adjacent segment degeneration, and complications related to revision surgery.^5–7 Therefore, identifying effective conservative therapies that may reduce progression to surgical intervention remains clinically important.

Acupuncture is a traditional Chinese medicine–based intervention that includes manual acupuncture, warm needle acupuncture, and electroacupuncture. It has been increasingly studied in modern clinical research and has shown potential benefit for several pain-related conditions.^8,9 In LDH, several systematic reviews and meta-analyses have suggested that acupuncture may reduce pain and improve function compared with pharmacologic therapy or other conservative treatments.^10–12 For example, acupuncture has been associated with improved visual analog scale scores, higher overall effectiveness rates, and fewer adverse events than pharmacologic treatment in patients with LDH.¹² Lumbar traction is another commonly used noninvasive therapy that applies distractive force to the lumbar spine to reduce mechanical compression.¹³ However, evidence supporting traction remains mixed: some studies have reported short-term improvements in pain and function.^14,15 Whereas systematic reviews and clinical guidelines have questioned its routine use because of limited or inconsistent benefit.^16,17 In lumbar radiculopathy and LDH, traction may improve short-term pain and disability when added to physical therapy, but its long-term effects remain uncertain.^18–20 Few studies have directly compared acupuncture with traction therapy in relation to subsequent need for discectomy or spinal fusion. To address this evidence gap, we used real-world data from the TriNetX database to compare the risk of lumbar surgical intervention among patients with LDH treated with acupuncture versus traction therapy. Our objective was to determine whether acupuncture was associated with a lower subsequent risk of discectomy or spinal fusion compared with traction therapy in patients with LDH.

Lumbar traction was selected as the active comparator because, like acupuncture, it is a nonoperative, procedure-based intervention used in patients with LDH and can be identified through specific procedural codes within electronic health records. This provides a more clearly defined active-treatment comparison than broader and more heterogeneous treatment categories such as pharmacologic management or general physical therapy. Subsequent lumbar surgery was selected as a clinically consequential longitudinal outcome because standardized pain and functional measures are not consistently available across healthcare organizations in large electronic health record networks, whereas surgical procedures can be identified through procedural coding. However, surgical utilization should not be considered a direct measure of biological disease progression alone because it may also be influenced by symptom severity, patient preference, physician recommendations, access to care, and institutional practice.

Retrospective analyses of real-world electronic health record data provide an opportunity to evaluate treatment patterns and clinically relevant outcomes in large and heterogeneous patient populations. However, unlike randomized trials, treatment allocation is not random, and such studies remain vulnerable to treatment-selection bias, coding and measurement limitations, and residual confounding. Propensity score methods can improve comparability with respect to measured baseline characteristics but cannot account for clinical variables that are unavailable or incompletely captured in the database.

Methods

Data Source

This retrospective cohort study utilized data from the TriNetX research network, a continuously updated global health research platform. TriNetX provided access to deidentified electronic health records from more than 250 million participants contributed by its network of partner healthcare organizations (HCOs). We utilized data from the US Collaborative Network comprising 64 HCOs in the TriNetX research network between January 1, 2000, and December 31, 2023, and deidentified electronic health records of more than 800,000 participants. The platform provides aggregated patient counts and statistical summaries designed to maintain deidentification throughout data retrieval and analysis. The investigators had no access to protected health information or identifiable patient-level data. The local ethics determination for this study is reported in the Ethics Approval and Informed Consent section.

Study Population

Inclusion criteria for this study included participants who were aged ≥20 years at the index date, had at least one healthcare organization (HCO) visit, and received a diagnosis of LDH confirmed by ICD-9 and ICD-10 codes. These diagnoses included intervertebral disc displacement in the lumbar region (ICD-10: M51.26; ICD-9: 722.10), intervertebral disc disorders with radiculopathy in the lumbar region (ICD-10: M51.16), and other intervertebral disc disorders in the lumbar region (ICD-10: M51.86; ICD-9: 722.93). Participants were excluded if they had a history of lumbar spine surgery or diagnoses related to lumbar spine and pelvic fractures (ICD-10: S32), malignant vertebral neoplasms (ICD-10: C41.2), or lumbar intervertebral disc infections (ICD-10: M46.36).

Participants were divided into two distinct treatment groups. The acupuncture group consisted of participants who received acupuncture treatment on the index date, confirmed by procedural codes including CPT (97810, 97811, 97813, 97814), ICD-9-CM (99.92), or ICD-10-PCS (8E0H30Z). The traction therapy group consisted of participants who received traction therapy on the index date, identified through procedural codes CPT (97012) or ICD-10-PCS (2W65X0Z, 2W65XZZ). The index date for the acupuncture group was defined as the date of the first qualifying acupuncture session after LDH diagnosis, and similarly, the index date for the traction therapy group was the date of the first qualifying traction session after diagnosis. Participants were excluded if they underwent traction therapy after initiating acupuncture or received acupuncture following traction therapy, to minimize crossover between the two index treatments. Other conservative treatments, including pharmacologic and rehabilitation-based therapies, were not used as exclusion criteria. Therefore, the exposure groups represent real-world treatment cohorts rather than isolated acupuncture or traction monotherapy. Cohort assignment was based on documentation of a qualifying treatment procedure, while treatment frequency, cumulative number of sessions, and treatment duration were not standardized in the analytic exposure definition.

Outcome Measures and Covariates

The primary outcomes measured were the incidence of subsequent discectomy and spinal fusion surgeries, confirmed by relevant CPT, ICD-9, and ICD-10 codes. No prespecified minimum number of treatment sessions was required for cohort assignment. The first qualifying acupuncture or traction procedure after lumbar disc herniation diagnosis defined the index date. Additional covariates considered in the analysis were demographic characteristics such as age, sex, race, and healthcare utilization, lifestyle factors including tobacco use, alcohol-related disorders, nicotine dependence, and comorbidities such as diabetes mellitus (DM), osteoporosis, cerebrovascular diseases, chronic obstructive pulmonary disease (COPD), liver disease, and others detailed in Table 1.

Table 1.

Baseline Characteristics of the Study Participants Before and After Propensity Score Matching

Characteristic	Before Matching, No. (%)			After Matching, No. (%)
	Acupuncture (n = 3146)	Traction Therapy (n = 11,554)	Standardized Difference	Acupuncture (n = 3045)	Traction Therapy (n = 3045)	Standardized Difference
Age, mean (SD), ya	57.7 (14.3)	52.7 (14.0)	0.350	57.5 (14.3)	57.6 (13.5)	0.007
Sex
Female	2099 (66.7)	5592 (48.4)	0.377	2016 (66.2)	2034 (66.8)	0.013
Male	1026 (32.6)	5679 (49.2)	0.341	1008 (33.1)	988 (32.4)	0.014
Unknown	21 (0.7)	283 (2.4)	0.144	21 (0.7)	23 (0.8)	0.008
Race
White	2448 (77.8)	7925 (68.6)	0.209	2371 (77.9)	2420 (79.5)	0.039
Black or African American	223 (7.1)	1356 (11.7)	0.160	217 (7.1)	212 (7.0)	0.006
Asian	110 (3.5)	557 (4.8)	0.066	103 (3.4)	94 (3.1)	0.017
Other	148 (4.7)	382 (3.3)	0.071	139 (4.6)	119 (3.9)	0.033
Unknown	206 (6.5)	1162 (10.1)	0.127	204 (6.7)	186 (6.1)	0.024
Lifestyles
Nicotine dependence	193 (6.1)	1171 (10.1)	0.147	185 (6.1)	145 (4.8)	0.058
Tobacco use	33 (1.0)	177 (1.5)	0.043	31 (1.0)	26 (0.9)	0.017
Alcohol related disorders	49 (1.6)	243 (2.1)	0.041	45 (1.5)	34 (1.1)	0.032
Comorbidities
Diabetes mellitus	445 (14.1)	1572 (13.6)	0.016	423 (13.9)	398 (13.1)	0.024
Ischemic heart disease	254 (8.1)	894 (7.7)	0.013	246 (8.1)	235 (7.7)	0.013
Osteoporosis	265 (8.4)	363 (3.1)	0.228	238 (7.8)	230 (7.6)	0.010
Chronic kidney disease	213 (6.8)	456 (3.9)	0.126	199 (6.5)	178 (5.8)	0.029
Liver disease	232 (7.4)	521 (4.5)	0.121	196 (6.4)	178 (5.8)	0.025
COPD	125 (4.0)	417 (3.6)	0.019	117 (3.8)	93 (3.1)	0.043
Heart failure	120 (3.8)	272 (2.4)	0.085	106 (3.5)	90 (3.0)	0.030
Peripheral vascular disease	108 (3.4)	263 (2.3)	0.070	100 (3.3)	89 (2.9)	0.021
Connective tissue disease	107 (3.4)	151 (1.3)	0.139	94 (3.1)	93 (3.1)	0.002
Cerebrovascular disease	85 (2.7)	113 (1.0)	0.129	70 (2.3)	67 (2.2)	0.007
Transient ischemic attack	40 (1.3)	71 (0.6)	0.068	34 (1.1)	31 (1.0)	0.010
HIV	20 (0.6)	87 (0.8)	0.014	20 (0.7)	16 (0.5)	0.017
Lymphoid leukemia	26 (0.8)	28 (0.2)	0.080	18 (0.6)	16 (0.5)	0.009
Malignant (primary) neoplasm	71 (2.3)	16 (0.1)	0.196	17 (0.6)	16 (0.5)	0.005
Hemiplegia and hemiparesis	20 (0.6)	21 (0.2)	0.071	15 (0.5)	14 (0.5)	0.005
Dementia	13 (0.4)	17 (0.1)	0.050	11 (0.4)	13 (0.4)	0.011
Non-Hodgkin lymphoma	15 (0.5)	13 (0.1)	0.067	11 (0.4)	10 (0.3)	0.006
Peptic ulcer disease	10 (0.3)	39 (0.3)	0.003	10 (0.3)	10 (0.3)	0.000
Myeloid leukemia	14 (0.4)	10 (0.1)	0.070	10 (0.3)	10 (0.3)	0.000
Monocytic leukemia	10 (0.3)	10 (0.1)	0.052	10 (0.3)	10 (0.3)	0.000
Other leukemias of specified cell type	10 (0.3)	10 (0.1)	0.052	10 (0.3)	10 (0.3)	0.000
Leukemia of unspecified cell type	10 (0.3)	10 (0.1)	0.052	10 (0.3)	10 (0.3)	0.000
PPHH-SPSC	78 (2.5)	130 (1.1)	0.102	70 (2.3)	67 (2.2)	0.007
Medical utilization
Office or other outpatient services	1787 (56.8)	6157 (53.3)	0.071	1719 (56.5)	1704 (56.0)	0.010
Preventive medicine services	653 (20.8)	2340 (20.3)	0.013	646 (21.2)	626 (20.6)	0.016
Emergency department services	543 (17.3)	1810 (15.7)	0.043	498 (16.4)	426 (14.0)	0.066
Hospital inpatient and observation care services	347 (11.0)	603 (5.2)	0.214	290 (9.5)	249 (8.2)	0.047
Laboratory data
BMI ≥30 b	1187 (37.7)	4421 (38.3)	0.011	1144 (37.6)	1088 (35.7)	0.038

Notes: ^a Age at the index date. ^b Calculated as weight in kilograms divided by height in meters squared.

Abbreviations: BMI, body mass index; COPD, chronic obstructive pulmonary disease; HIV, human immunodeficiency virus; PPHH-SPSC, persons with potential health hazards related to socioeconomic and psychosocial circumstances; SD, standard deviation.

Statistical Analysis

Baseline characteristics, comorbidities, and other covariates were analyzed (Table 1). An absolute standardized difference greater than 0.1 was used as the threshold to indicate significant imbalance in baseline characteristics between the acupuncture and traction therapy groups. Propensity score matching (PSM) was performed at a 1:1 ratio for age, sex, race, lifestyles, comorbidities (DM, osteoporosis, cerebrovascular diseases, COPD, liver disease and others, listed in Table 1), and laboratory data such as body mass index (BMI), to improve comparability of measured baseline characteristics between the two groups. Survival analysis was conducted with Kaplan-Meier curves to compare time-to-event outcomes (discectomy and spinal fusion) between groups. Subgroup analyses were conducted to compare the effects on the incidence of discectomy and spinal fusion within each subgroup. The analyses were based on various baseline characteristics, including sex, age, race, lifestyles, and comorbidities. Continuous variables were expressed as mean and standard deviation (SD). Categorical variables were presented as frequency and percentage. For each outcome, hazard ratios (HRs) with 95% confidence intervals (CIs) were calculated using Cox proportional hazards regression. P <0.05 were considered statistically significant. All data management and HR calculations were performed using the TriNetX database.

Results

Patient Cohorts

A total of 880,063 patients aged ≥20 years with LDH were identified through the TriNetX database from January 1, 2000, to December 31, 2023. A total of 57,459 patients were excluded because of lumbar spine or pelvic fracture, malignant neoplasm of the vertebral column, or pyogenic infection of the lumbar intervertebral disc. The remaining 822,604 patients were eligible for treatment cohort assignment, and patients who had undergone discectomy or spinal fusion before their initial acupuncture or traction treatment were excluded. The acupuncture cohort included 3,146 patients who received acupuncture without documented crossover to traction, whereas the traction therapy cohort included 11,554 patients who received traction without documented crossover to acupuncture. Propensity score matching at a 1:1 ratio resulted in balanced cohorts of 3,045 patients each (Figure 1).

Figure 1.

Flow Diagram of Cohort Construction.

Abbreviations: LDH indicates lumbar disc herniation; PSM, propensity score matching.

Following matching, the mean ages were 57.5 and 57.6 years in the acupuncture and traction therapy cohorts, respectively. Sex distributions were 66.2% females and 33.1% males in the acupuncture cohort versus 66.8% females and 32.4% males in the traction therapy cohort. White patients represented the predominant race in both cohorts, comprising 77.9% and 79.5%, respectively (Table 1).

Baseline Characteristics

In the comparison between acupuncture and traction therapy, unbalanced distributions exist before PSM. Differences were observed in age (standardized difference = 0.350), sex (female: 0.377; male: 0.341; unknown: 0.144), and most racial groups (White: 0.209; Black or African American: 0.160; unknown race: 0.127). Additionally, variations were noted in lifestyle factors such as nicotine dependence (0.147), several comorbidities (osteoporosis: 0.228; chronic kidney disease: 0.126; liver disease: 0.121; connective tissue disease: 0.139; cerebrovascular disease: 0.129; malignant neoplasm: 0.196), potential health hazards related to socioeconomic and psychosocial circumstances (0.102), and medical utilization of hospital inpatient and observation care services (0.214). The standardized differences of these variables were all higher than 0.1, indicating significant differences between the two cohorts (Table 1).

After PSM was applied to those data, measured baseline covariate imbalance was reduced, and all standardized differences were below 0.1. The medical utilization of emergency department services has the highest standardized difference at 0.066. The most prevalent comorbidity was DM, with a standardized difference of 0.024, present in 13.9% in the acupuncture cohort versus 13.1% in the cohort receiving traction therapy.

Primary Outcomes

Table 2 summarizes the proportion of patients undergoing discectomy or spinal fusion within 5 years after the index treatment. Receipt of acupuncture was associated with lower observed risks of discectomy and spinal fusion compared with traction therapy (HR, 0.46; 95% CI, 0.35–0.60; P =0.02; and HR, 0.73; 95% CI, 0.58–0.91; P =0.006, respectively) (Table 2). Kaplan-Meier analysis showed separation of the curves for both outcomes (log-rank P <0.001) (Figure 2). These estimates represent associations within the matched observational cohorts and should not be interpreted as isolated causal treatment effects.

Table 2.

Incidence of Outcomes Over 5 Years After Propensity Score Matching by Treatment Group

	Patients with Outcome/Population at Risk, No./Total (%)
Outcome	Acupuncture	Traction therapy	Hazard ratio (95% CI)	P value
Discectomy	77/3045 (2.5)	175/3045 (5.7)	0.46 (0.35–0.60)	0.02
Spinal Fusion	128/3045 (4.2)	188/3045 (6.2)	0.73 (0.58–0.91)	0.006

Abbreviation: CI, confidence interval.

Figure 2.

Cumulative Incidence of Discectomy and Spinal Fusion in the Acupuncture Group vs Traction Therapy Group. (A) Discectomy. (B) Spinal fusion. Kaplan-Meier curves show the cumulative incidence (CI) over 5 years after the index date. The orange line represents the acupuncture therapy group, and the blue dashed line represents the traction therapy group. Shaded areas indicate 95% CIs. P values were calculated using the Log rank test.

Subgroup Analysis

Subgroup analyses were conducted as exploratory analyses. For discectomy, the overall association favored acupuncture (HR, 0.46; 95% CI, 0.35–0.60). Point estimates generally favored acupuncture across several subgroups; however, some estimates were imprecise and had 95% CIs crossing the null value. These included Black or African American patients (HR, 0.575; 95% CI, 0.168–1.965), Asian patients (HR, 0.444; 95% CI, 0.115–1.719), patients with alcohol-related disorders (HR, 1.224; 95% CI, 0.076–19.635), osteoporosis (HR, 1.357; 95% CI, 0.413–4.462), and COPD (HR, 0.502; 95% CI, 0.046–5.540). In these subgroups, fewer than 10 discectomy events occurred in each cohort, contributing to wide confidence intervals and substantial statistical uncertainty (Figure 3A).

Figure 3.

Subgroup Analyses for Discectomy and Spinal Fusion of Patients Treated with Acupuncture vs Traction Therapy. (A) Discectomy. (B) Spinal fusion. Forest plots show hazard ratios (HRs) and 95% confidence intervals (CIs) for discectomy and spinal fusion among patient subgroups comparing acupuncture with traction therapy. HRs were estimated using Cox proportional hazards regression. The vertical dashed line indicates the null value (HR = 1).

Abbreviations: HR indicates hazard ratio; CI, confidence interval; BMI, body mass index; COPD, chronic obstructive pulmonary disease.

For spinal fusion, several subgroup estimates also had wide 95% CIs crossing the null value, including males (HR, 0.942; 95% CI, 0.619–1.433), patients aged 20–64 years (HR, 0.822; 95% CI, 0.598–1.132) and ≥65 years (HR, 0.756; 95% CI, 0.497–1.151), Black or African American patients (HR, 1.538; 95% CI, 0.618–3.831), Asian patients (HR, 1.658; 95% CI, 0.276–9.959), patients with smoking-related factors (HR, 0.634; 95% CI, 0.280–1.439), alcohol-related disorders (HR, 0.744; 95% CI, 0.210–2.641), diabetes mellitus (HR, 1.071; 95% CI, 0.570–2.010), osteoporosis (HR, 0.682; 95% CI, 0.282–1.650), and COPD (HR, 0.269; 95% CI, 0.030–2.417) (Figure 3B). Because of sparse events and imprecision, these subgroup findings should be considered exploratory and should not be interpreted as demonstrating treatment-effect heterogeneity without adequate statistical power and formal interaction testing.

Discussion

This retrospective cohort study found that, among matched adults with LDH, receipt of acupuncture was associated with lower observed 5-year incidences of subsequent discectomy and spinal fusion compared with lumbar traction. These findings extend prior literature focused primarily on short-term symptom outcomes by evaluating a clinically consequential longitudinal endpoint. However, the associations should not be interpreted as evidence that acupuncture itself prevents surgery or alters the biological progression of LDH because treatment selection, concomitant therapies, disease severity, and surgical decision-making were not fully captured in the database.

Previous systematic reviews and clinical studies have reported improvements in pain and functional outcomes with acupuncture in patients with LDH and related conditions.^10,11,21 These studies primarily evaluated short-term symptomatic endpoints. Our analysis instead examined subsequent surgical utilization over five years, providing complementary real-world evidence regarding longer-term healthcare outcomes. The observed association should nevertheless be interpreted cautiously because surgery is influenced by clinical severity, patient preference, physician recommendation, healthcare access, and other treatment pathways in addition to symptom persistence.

The observed differences in subsequent surgical utilization should also be interpreted within the context of multimodal real-world management of LDH. Acupuncture and lumbar traction are rarely delivered in complete therapeutic isolation. Patients in either group may simultaneously receive medications, physical therapy, exercise-based rehabilitation, manual therapy, injections, or other conservative interventions. Patients selecting acupuncture may also differ from those receiving traction in treatment preferences, willingness to continue conservative care, healthcare-seeking behavior, and access to complementary services. Likewise, clinicians referring patients for acupuncture may follow different treatment pathways or surgical referral thresholds. Because these co-interventions and patient- or provider-level factors could not be comprehensively characterized, part of the observed association may reflect broader patterns of care rather than an isolated physiological effect of acupuncture.

Several mechanisms have been proposed for both treatment modalities. Lumbar traction has been reported to alter mechanical loading and intradiscal pressure and may influence paraspinal muscle activity.^14,22–27 Acupuncture has been proposed to modulate peripheral and central nociceptive pathways, endogenous opioid signaling, and inflammatory or neuroimmune pathways.^28–32 However, the present study did not measure intradiscal pressure, inflammatory biomarkers, neural signaling, or other mechanistic outcomes. Therefore, these mechanisms should be regarded as hypotheses derived from prior literature and cannot explain causally the observed differences in subsequent surgical utilization in the present analysis.

The present analysis also does not establish which patients are most likely to benefit from acupuncture in routine clinical practice. Although the subgroup analyses provide exploratory information, several estimates were imprecise because of limited event counts. Several subgroup estimates were imprecise because of small event counts and wide confidence intervals. Accordingly, these subgroup analyses should be considered exploratory and should not be interpreted as demonstrating treatment-effect heterogeneity without adequate statistical power and formal interaction testing. Prospective studies incorporating symptom severity, neurological findings, imaging characteristics, treatment dose, and patient-reported outcomes are needed to determine whether specific clinical phenotypes are more likely to benefit from acupuncture.

Limitations

This study has several limitations. First, as a retrospective cohort study using deidentified electronic health records, it is vulnerable to treatment-selection bias, coding limitations, and residual confounding. Clinically important determinants of treatment choice and surgery, including baseline pain severity, symptom duration, neurological deficits, functional status, and imaging characteristics such as herniation morphology, size and location, nerve-root compression, and concomitant spinal stenosis, were not comprehensively available. Second, propensity score matching improved balance across measured baseline characteristics but cannot account for unmeasured or incompletely measured variables. Third, acupuncture and traction were identified through procedural codes, which do not adequately characterize treatment protocols, treatment frequency or duration, cumulative treatment exposure, or practitioner training and experience. Treatment intensity and duration could not be standardized, and the exposure definition did not distinguish patients receiving a single treatment session from those undergoing repeated or complete treatment courses. The exposure definition therefore could not distinguish a single treatment encounter from a repeated or standardized treatment course. Fourth, other conservative treatments were not excluded. Patients may have received physical therapy, exercise-based rehabilitation, manual therapy, analgesic or anti-inflammatory medications, injections, or other interventions during follow-up, and differential use of these co-interventions may have contributed to the observed associations. Fifth, subsequent surgery is influenced not only by symptoms or disease progression but also by patient preferences, physician recommendations, institutional practice, access to care, and willingness to continue nonoperative management. These factors could introduce selection bias that cannot be fully addressed in the present design. Finally, several subgroup analyses included few events and wide confidence intervals; these estimates are exploratory and should not be interpreted as evidence of differential treatment effects. Several biological mechanisms have been proposed to explain the analgesic effects of acupuncture, including modulation of peripheral and central nociceptive pathways, endogenous opioid signaling, and inflammatory pathways. Lumbar traction has likewise been proposed to influence mechanical loading, intradiscal pressure, and paraspinal muscle activity. However, none of these mechanisms were directly evaluated in the present study. Therefore, mechanistic explanations for the observed differences in subsequent surgical utilization remain speculative and cannot be established from the current analysis.

The lower observed incidence of subsequent surgery among patients receiving acupuncture cannot necessarily be attributed to a direct therapeutic effect. Patients who select acupuncture may differ in their willingness to continue conservative management or their preferences regarding surgery, while clinicians referring patients for acupuncture may follow different treatment pathways or surgical referral thresholds. Patient- and provider-level treatment preferences may therefore influence subsequent surgical utilization independently of the biological effect of acupuncture. Prospective studies with standardized treatment protocols, detailed clinical and imaging information, treatment-dose assessment, patient-reported outcomes, and prespecified evaluation of co-interventions are needed to determine whether acupuncture itself influences subsequent surgical utilization in patients with LDH.

Conclusions

In this retrospective cohort study of adults with LDH, receipt of acupuncture was associated with lower observed 5-year incidences of subsequent discectomy and spinal fusion compared with lumbar traction. These associations should be interpreted cautiously because residual confounding related to concomitant treatments, clinical severity, treatment selection, patient preferences, and surgical decision-making could not be fully accounted for. The findings are hypothesis-generating and support prospective studies to determine whether acupuncture itself influences subsequent surgical utilization.

Funding Statement

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Ethics Approval and Informed Consent

This retrospective observational cohort study used de-identified electronic health record data from the TriNetX US Collaborative Network. No participants were recruited, and no intervention was assigned, administered, or modified by the investigators. The investigators had access only to de-identified data and aggregated statistical outputs and did not have access to identifiable patient-level information. Individual informed consent was therefore not required for this secondary analysis of de-identified data. The use of TriNetX in this research was approved by the Institutional Review Board of Chung Shan Medical University Hospital (CSMUH), under the approval number CS2-21176. Clinical trial registration was not applicable because this study did not involve prospective assignment of participants to an intervention.

Disclosure

Prof. Dr. Meng-Huang Wu reports stock or stock options from Omnisurgica, Jungu, outside the submitted work. The authors have no financial disclosures to declare.