Effectiveness and Safety of Electroacupuncture in Patients with Treatment-Resistant Insomnia: A Randomized, Assessor-Blinded, Waitlist-Controlled Pilot Trial

Introduction

Insomnia is a common sleep disorder¹ that burdens individuals’ well-being and public health.² Although clinical practice guidelines recommend cognitive behavioral therapy for insomnia (CBT-I) as the first-line treatment,^3–6 pharmacotherapy remains the predominant treatment approach in real-world clinical settings.^7–9 Although over half of patients with chronic insomnia use sleep medications for 6 months or longer, only 28% experience symptom relief.¹⁰ Clinical interventions and studies for patients dissatisfied with sleep despite long-term medication use remain limited.^6,11 Persistent insomnia despite long-term pharmacotherapy is concerning, as chronic insomnia is associated with higher risks of mood and anxiety disorders, cognitive decline, and cardiovascular morbidity.¹²

Despite these concerns, long-term pharmacotherapy is often continued in clinical practice because access to alternative non-pharmacological treatments, such as CBT-I, remains limited, and discontinuation of hypnotics can be challenging due to dependence and withdrawal concerns.^6,7,10,11

Definitions of treatment-resistant insomnia (TRI) vary across clinical studies,^13–16 but it is generally described as insomnia unresponsive to standard treatments or lacking significant improvement in sleep quality or quantity despite appropriate treatment for 3 months or longer. In this study, TRI was operationally defined as persistent moderate-to-severe insomnia despite at least three months of stable pharmacotherapy. Although CBT-I is recommended as a first-line treatment, access to CBT-I remains limited in routine clinical practice in South Korea, where insomnia management is mostly focused on pharmacological treatment. Therefore, the current definition mainly addresses pharmacotherapy-resistant insomnia in a real-world clinical setting.

The number of patients seeking complementary and alternative insomnia treatments has increased over time. In Korea, this trend is reflected in the growing use of Korean medicine for insomnia.¹⁷ According to surveys conducted by Korean medicine practitioners, most patients with insomnia visiting Korean medicine clinics are already taking sleep medications, with reductions in side effects and dosage being the primary motivations.¹⁸ This indicates the growing demand for treatments that address the limitations of pharmacotherapy.

Practitioner surveys identify acupuncture as the most common non-pharmacological intervention for insomnia in Korean medical institutions.¹⁸ It is safe for patients on medications due to its low risk of drug interactions. Recent systematic reviews and meta-analyses have reported that acupuncture treatment for 3 weeks or longer is associated with improvements in sleep quality, with effects that may be comparable to those observed with pharmacotherapy in some studies.^19,20 A large network meta-analysis further compared different acupuncture-based interventions for primary insomnia and similarly reported improvements in subjective sleep outcomes across treatments modalities.²¹ However, these evidence syntheses primarily included heterogeneous insomnia populations and focused on comparative efficacy across intervention types, without specifically addressing treatment resistance or ongoing long-term hypnotic use. Consequently, their findings may not fully reflect the clinical characteristics of patients who seek Korean traditional medical treatment in real-world settings.

Electroacupuncture (EA), a variant of acupuncture, has emerged as a promising treatment option to address existing clinical needs. EA provides more consistent and enhanced stimulation through controlled intensity, frequency, and duration. Previous studies have reported potential benefits for sleep-related physiological mechanisms, including changes in brain functional connectivity—particularly within networks related to emotional regulation and self-referential processing, such as the default mode network and limbic regions²²—as well as autonomic nervous system balance modulation.²³

Studies show EA significantly improves Pittsburgh Sleep Quality Index (PSQI) and Insomnia Severity Index (ISI) scores in patients with insomnia compared with other acupuncture methods.²⁴ A multicenter study reported that EA is an effective and safe intervention for improving sleep quality, insomnia severity, and anxiety-depressive symptoms in patients with insomnia.²⁵ EA also reduces medication use in patients with long-term benzodiazepine dependence.^26,27

Most existing acupuncture-based studies, including those evaluating EA, have not specifically examined patients with persistent insomnia symptoms despite ongoing long-term hypnotic use. Therefore, this study focused on patients with TRI who continued long-term hypnotic use without meaningful symptom relief, reflecting common clinical practice in South Korea, where access to CBT-I remains limited.

Materials and Methods

Study Design

This multicenter, randomized, waitlist-controlled, assessor-blinded pilot study was conducted at two Korean medicine hospitals: Pusan National University Korean Medicine Hospital and Dongshin University Korean Medicine Hospital (located in Gwangju). The trial was also prospectively registered with the Clinical Research Information Service (CRIS) under the identifier KCT0003235.

Participant Recruitment and Study Schedule

A total of 50 individuals diagnosed with insomnia disorder were enrolled through a competitive recruitment process. Prior to participation, all participants received a comprehensive explanation of the study aims and procedures, and written informed consent was obtained. A qualified investigator confirmed eligibility based on predetermined criteria. Participants were then randomly allocated to either the experimental or control group at the initial visit (baseline) using a 1:1 allocation ratio. Study visits were scheduled over 10 weeks according to group allocation. The experimental group completed 14 scheduled visits: 12 treatment sessions (twice weekly for 6 weeks), post-intervention evaluation (Week 6), and 4-week follow-up assessment (Week 10). The control group only attended assessment visits: baseline and Week 2, 4, 6, and 10. The overall study schedule is summarized in Table 1.

Table 1 Schedule of Enrollment, Intervention, and Assessments

Inclusion and Exclusion Criteria

Inclusion Criteria

Eligibility for participation required fulfillment of all the following criteria:

1. Age 19 to 80 years.
2. Continuous physician-prescribed insomnia medications for ≥ 3 months, with no change in medication type or dosage within the two-week period prior to enrollment.
3. ISI ≥ 15.
4. Meeting the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) diagnostic criteria for insomnia disorder.
5. Provision of written informed consent.

Exclusion Criteria

Individuals were excluded from the study if they met any of the following criteria:

1. Recent Korean medicine treatments for insomnia within the two-week period.
2. Initiation of, or plans to initiate dietary supplements or non-pharmacological therapies for insomnia (eg, CBT-I, meditation) within 2 weeks before or during the trial.
3. Unstable schizophrenia, mania, or bipolar disorder within 6 months, or Hospital Anxiety and Depression Scale (HADS) anxiety or depression subscale score ≥ 11.
4. Diagnosed substance abuse/dependence within 6 months.
5. History of suicide/homicide attempt, or self-injury.
6. Currently engaged in shift work or circadian rhythm-disrupting schedules.
7. Presence of severe pain or any medical condition that causes significant sleep disturbance.
8. Current use of hemostatic agents.
9. Abnormal thyroid function (free T4 or TSH < 0.1 or > 5.1 μIU/mL).
10. Clinically significant abnormalities on laboratory testing, including but not limited to: total bilirubin > 2× institutional upper limit of normal (ULN); AST or ALT > 2.5× ULN; serum creatinine > 2.5× ULN; white blood cell count < 1.5 × 10⁹/L or ≥ 10.0 × 10⁹/L; absolute neutrophil count (ANC) < 1000/μL; platelet count < 75 × 10⁹/L; positive urine hCG; or any other findings deemed clinically significant.
11. Diagnosis of a severe chronic or terminal illness.
12. History of hypersensitivity reactions or intolerance to acupuncture.
13. Presence of implanted devices that could interfere with EA or cause electrical stimulation hypersensitivity.
14. Pregnancy, lactation, or intention to become pregnant during the study period.
15. Participated in another clinical trial within the previous four weeks, or currently involved in another study.
16. Anticipated difficulty adhering to the study protocol, including treatment attendance or completion of questionnaires.
17. Any other condition judged by the investigators to render the participant unsuitable for the trial.

Sample Size

The effectiveness and safety of EA for TRI have not been previously evaluated. Sample size determination was based on a comparable study investigating EA for insomnia, which used ISI as the primary outcome.²⁸ Yeung et al²⁸ reported mean changes in ISI scores of 5.8 (Standard deviation [SD] = 3.2) and 2.1 (SD = 4.7) for the experimental and control groups respectively. With an assumed effect size of 0.92, 20 participants per group were calculated (two-sided α = 0.05, power = 80%). Participants were allocated in a 1:1 ratio. Allowing for 20% dropout over 10 weeks, the total sample size was set at 50 participants, 25 per group.

Recruitment

Participants were enrolled through internal and external recruitment strategies, including institutional bulletin boards, official websites, local newspapers, and subway advertisements.

Randomization, Allocation Concealment, and Assessor Blinding

An independent statistician generated the randomization sequence via site-stratified block randomization using SAS version 9.4 (SAS Institute Inc., Cary, NC). Allocation lists were sealed in opaque envelopes, sent to each site, and stored in double-locked cabinets. Investigators opened the envelopes sequentially and recorded the assigned numbers in the electronic medical records. Since a waitlist control design was employed, both practitioners and participants were aware of group allocation. To reduce assessment bias, outcome assessors who were not involved in treatment delivery or group allocation were blinded and conducted evaluations in separate rooms.

Intervention

All participants received a brochure containing educational information on sleep hygiene. Participants were instructed to maintain the same type and dosage of their prescribed ongoing hypnotic medications throughout the study period. Any change in hypnotic medication was considered a protocol violation and predefined as a criterion for discontinuation.

Experimental Group: Electroacupuncture

Participants allocated to the experimental group received treatment using disposable, sterilized filiform needles (0.25 × 30 mm; Dong-bang Acupuncture, Seoul, Korea) inserted at the following acupoints (Figure 1): Yintang (EX-HN3), Baihui (GV20), bilateral Shenmen (HT7), Neiguan (PC6), Dazhong (KI4), and Jinmen (BL63). The selection of acupoints was based on Yeung et al,²⁸ Yeung et al,²⁹ Kim et al,³⁰ and Kim et al³¹ After achieving deqi, an EA device (ES-160, Ito Co., Tokyo, Japan) was connected to five pairs of acupoints (EX-HN3–GV20; left and right PC6–HT7; left and right KI4–BL63) and delivered 4 Hz stimulation at a noticeable but comfortable intensity for 30 min. The treatment was administered by licensed doctors of Korean medicine with a minimum of 10 years of clinical experience. Detailed EA treatment protocols are provided in Supplement 1.

Figure 1 Locations of acupoints used for electroacupuncture. Notes: Red dots indicate positive and blue dots indicate negative electrode placement for electroacupuncture stimulation.

Control Group: Waitlist-Control

Participants in the waitlist-control group did not receive EA treatment. They maintained the type and dosage of their regular physician-prescribed medications for insomnia throughout the study period. Participants could continue existing self-care practices, but initiating new medications or supplementary treatments for insomnia was prohibited during the trial.

Concomitant Treatment

Traditional Korean medicine interventions for insomnia (eg, acupuncture, cupping, moxibustion, or herbal medicine) were prohibited during the trial. Participants who changed the type or dosage of their medications for insomnia were withdrawn.

Non-pharmacological treatments (eg, CBT-I or dietary supplements) were permitted, if initiated at least two weeks prior to the screening and unchanged during the study.

Participants were instructed to report any new treatment, which was recorded in the case report form (CRF). Those receiving prohibited treatments were excluded from the trial.

Criteria for Discontinuation and Withdrawal

Participants were withdrawn under any of the following conditions: (1) voluntary withdrawal or discontinuation of consent; (2) violation of the inclusion or exclusion criteria during the trial; (3) occurrence of a serious adverse event; (4) substantial protocol violation; (5) failure to complete follow-up assessments; (6) any alteration in the type or dosage of prescribed sleeping medication; (7) missing more than three consecutive sessions or attending fewer than a total nine sessions; or (8) any other situation deemed inappropriate for continued participation by the investigator.

Outcome Measurements

Primary Outcome

The primary outcome measure was the between-group difference in the mean change in ISI scores from baseline to post-intervention (Week 6). The ISI was administered at five time points: baseline, Week 2, 4, 6, and 4-week follow-up (Week 10).

Secondary Outcomes

Secondary outcomes included changes in ISI and PSQI, along with sleep parameters measured through sleep diaries and actigraphy. ISI, PSQI, sleep diary information, and actigraphy data obtained using ActiGraph wGT3X-BT (ActiGraph Corp., Pensacola, FL) were collected at baseline, Week 2, 4, 6, and 10.

Upon waking, participants completed daily sleep diaries documenting bedtime, wake time, sleep onset latency, total sleep duration, number of awakenings, use of sleep medications, and morning refreshment which was defined as the participant’s perceived level of refreshment or fatigue immediately upon waking and was assessed using a single-item, three-point scale (1 = refreshed, 2 = somewhat refreshed, 3 = fatigued). These diary entries were used to assess sleep onset latency (SOL), total sleep time (TST), sleep efficiency (SE), number of awakenings, and morning refreshment. Actigraphy-derived sleep variables included TST, SE, wake after sleep onset (WASO), average awakening length, and number of awakenings. Actigraphy was worn on the non-dominant wrist. Activity counts were collected in 60-second epochs, and sleep periods were determined by cross-referencing with sleep diary entries. Data were analyzed using the Cole-Kripke algorithm³² in ActiLife software (version 6; ActiGraph Corp., Pensacola, FL), which is commonly used for sleep-wake estimation in adult populations.

To evaluate the short-term effects of EA on insomnia, participants completed the ISI and PSQI after one week with two sessions, and were assessed using a visual analog scale (VAS) for physical symptoms accompanying insomnia including headache, dyspepsia, and fatigue, and other related symptoms.

To investigate participants’ pattern identification, the instrument on Pattern Identification Tool for Insomnia (PIT-Insomnia) was administered at baseline. This self-reported questionnaire consisted of 47 items, and one of five pattern identification types was determined based on the final score.³³

Economic Evaluation

Outcomes including quality of life (QoL) and productivity loss, and costs were measured at baseline and post-intervention (Week 6). QoL was assessed using the EuroQoL five-dimension three-level questionnaire (EQ-5D-3L), the EuroQoL visual analogue scale (EQ-VAS), and the 36-Item Short Form Health Survey (SF-36). EQ-5D-3L scores were converted to utility values using Korean tariff values, and quality-adjusted life years (QALYs) were calculated using the area under the curve approach. Work productivity loss was measured using the Work Productivity and Activity Impairment (WPAI) (0 = no impairment, 1 = maximum impairment). Medical costs included Korean medicine therapy and conventional medicine services during the 6-week period, reported in 2020 Korean won (KRW). For international comparability, costs were also expressed in U.S. dollars (USD), converted from KRW at an exchange rate of 1 USD = 1500 KRW as of January 19, 2026. This economic evaluation adopted a healthcare system perspective with a 6-week time horizon, reporting incremental cost-effectiveness ratios (ICERs) for EA versus usual care.

Feasibility Outcomes

To explore clinical trial feasibility, recruitment, adherence, and completion rates were evaluated. Recruitment rate was defined as the proportion of participants enrolled among all individuals screened. Adherence rate was defined as the proportion of participants in the experimental group completing at least 75% of scheduled EA sessions. Completion rate was defined as the proportion of participants in each group completing the trial through the final follow-up assessment without withdrawal. All indicators were expressed as percentages.

Safety Assessment

Safety evaluations comprised laboratory testing performed at the screening and post-intervention phases. Throughout the study period, investigators assessed vital signs and conducted interviews at each visit to document any adverse events (AEs) reported by participants. The severity of AEs was graded as mild, moderate, or severe in accordance with Spilker’s criteria. Causality was determined using the World Health Organization–Uppsala Monitoring Centre classification system, with events categorized as definitely related, probably related, possibly related, definitely unrelated, or unknown. All reported AEs and any abnormal laboratory findings were recorded and analyzed.

Data Collection and Management

Following written informed consent, participant data were recorded in CRFs by trained clinical research coordinators. Outcome assessments were conducted by evaluators blinded to group allocation. The principal investigator established a standard operating procedure (SOP) specifying CRF completion, data entry, and study conduct to ensure data integrity and trial reliability. Automated data queries identified out-of-range values and corrected discrepancies, after which the dataset was verified, cleaned, and locked. Access to the electronic CRF system was restricted to the investigator responsible for data management.

Data Monitoring and Auditing

Data monitoring and auditing procedures were implemented to ensure study quality and integrity by the Clinical Research Center of the Korea Institute of Oriental Medicine, an independent organization with no conflicts of interest. An initiation meeting was convened prior to the study commencement. A SOP manual was developed and provided to investigators at all sites to maintain uniformity of research procedures across sites.

Monitoring personnel visited participating institutions at key time points during the trial, including participant enrollment, study midpoint, and completion. These visits verified consistency between the CRF data and source documents, ensured adherence to the approved protocol, SOPs, Good Clinical Practice, and regulatory guidelines, assessed participant safety, including laboratory results, and AEs, and addressed protocol-related issues with investigators.

Statistical Analysis

Analyses were performed based on the full analysis set (FAS). Descriptive statistics summarized demographic and baseline characteristics of each group using means and SDs for continuous variables, and frequencies and percentages for categorical variables. For continuous variables, normality was assessed using the Shapiro–Wilk test. Between-group differences were analyzed using the independent t-test or Wilcoxon rank-sum test as appropriate. Categorical variables were compared using the chi-square test or Fisher’s exact test. Paired categorical variables were evaluated using McNemar’s exact test. For repeatedly measured outcomes, linear mixed models (LMM) analyzed time effects, group effects, and time × group interactions with random intercept for participants. Non-normally distributed or categorical outcomes used generalized linear mixed models (GLMM) or cumulative link mixed models (CLMM). For variables demonstrating significant interactions, Tukey’s post hoc tests were performed to adjust for multiple comparisons.

All analyses were two-sided with a significance set at 0.05, conducted using R software (version 4.4.3; R Foundation for Statistical Computing, Vienna, Austria).

Results

Study Flow and Sample Characteristics

The trial was conducted from January 2019 to December 2020, recruiting 70 volunteers from two hospitals. After excluding 20 individuals (11 ineligible, 9 withdrew consent), 50 participants were enrolled and randomized to the experimental group (n = 25) or the waitlist-control group (n = 25). During the trial, 1 participant in the experimental group and 7 in the control group discontinued the trial due to consent withdrawal, protocol violation, or unrelated serious adverse event (SAE). Ultimately, 24 participants in the experimental group and 18 in the control group completed the intervention following the study protocol (Figure 2).

Figure 2 Flow Diagram of participants through the study.

Baseline demographic and clinical characteristics did not differ significantly between groups (Tables 2 and 3), or between completers and dropouts within each group (Supplement 2).

Table 2 Characteristics of Participants at Baseline

Table 3 Baseline Characteristics of Insomnia Between Groups

Changes in Sleep-Related Assessment Scales

Changes in ISI Scores Over Time

All main effects for group, time, and the group × time interaction in ISI scores were statistically significant (p < 0.005). In between-group comparisons at each time point, the experimental group exhibited significantly lower mean ISI scores than the control group at all assessments (p < 0.02) (Figure 3 and Supplement 3).

Figure 3 Changes in Insomnia Severity Index scores. Notes: Data are presented as adjusted means ± standard error (SE) calculated using linear mixed-effects models. Within-group comparison with baseline * p < 0.05; ** p < 0.01; *** p < 0.001. Between-groups comparison †p < 0.05; ‡p < 0.01; ‡‡p < 0.001.

Proportion Meeting the MCID for Changes in ISI Scores at Each Time Point

The proportion of participants achieving improvements in ISI scores exceeding the MCID was significantly higher in the experimental group than in the control group from week 4 post-intervention through the end of the intervention period (p = 0.02) (Table 4).

Table 4 MCID Achievement Rates Based on Insomnia Severity Index Score Changes

Changes in PSQI Scores Over Time

For PSQI scores, the main effect of group was not statistically significant; however, the main effect of time and the group × time interaction were significant (p < 0.001). In the between-group comparisons at each time point, the experimental group showed a trend toward lower PSQI scores than the control group at the end of the intervention (p = 0.05) and had significantly lower scores at the follow-up assessment (p = 0.01) (Figure 4 and Supplement 4).

Figure 4 Changes in Pittsburgh Sleep Quality Index scores. Notes: Data are presented as adjusted means ± standard error (SE) calculated using linear mixed-effects models. Within-group comparison with baseline * p < 0.05; ** p < 0.01; *** p < 0.001. Between-group comparison † p < 0.05.

Proportion Meeting the MCID for Changes in PSQI Scores at Each Time Point

The proportion of participants achieving improvements in PSQI scores exceeding the MCID was significantly higher in the experimental group than in the control group at week 4 (p = 0.04), but was not maintained at subsequent time points (Table 5).

Table 5 Comparison of MCID Achievement Rates Based on Pittsburgh Sleep Quality Index Score Changes

Sleep Diary

Among sleep diary-based parameters, TST and SOL showed a significant group × time interaction (F = 2.63, p = 0.03), (F = 2.80, p = 0.02) (Supplement 5–7). SE showed a significant main effect of time (F = 4.77, p < 0.001), whereas the main effect of group and the group × time interaction were not statistically significant (Supplement 5 and 8). No significant between-group differences were found in the number of awakenings at any time point (Supplement 5 and 9). For morning refreshment, analysis showed a significant group × time interaction (χ² = 23.84, p < 0.001), with the experimental group showing significantly more refreshed scores than the control group from Week 2 through the follow-up assessment (p < 0.02) (Figure 5 and Supplement 10).

Figure 5 Changes in the proportion of morning refreshment after awakening derived from sleep diary. Notes: Data are presented as adjusted means ± standard error calculated using cumulative link mixed-effects models. Within-group comparison with baseline * p < 0.05; ** p < 0.01; *** p < 0.001. Between-groups comparison †p < 0.05; ‡p < 0.01.

Actigraphy

Among actigraphy-based sleep parameters, TST, WASO, average awakening length, and SE did not show a significant group × time interaction (F = 1.29, p = 0.28), (F = 1.20, p = 0.31), (F = 0.70, p = 0.59), (F = 1.21, p = 0.31). No significant between-group differences were observed for TST, WASO, average awakening length, the number of awakenings, or SE at any assessment time point. Furthermore, neither the experimental group nor the control group demonstrated statistically significant within-group changes in any actigraphy-based sleep parameters across all assessment time points. (Supplement 11–16).

Sleep-Related Assessments After Short-Term EA Treatment

Changes in ISI Scores

One week after initiating EA treatment, ISI scores in the experimental group decreased significantly from baseline but did not reach the MCID of 6 points (p = 0.02) (Table 6).

Table 6 Change in Insomnia Severity Index Score from Baseline to Week 1 in the Experimental Group

Changes in PSQI Scores

One week after initiating EA treatment, PSQI scores did not show significant improvement from baseline. (Table 7).

Table 7 Changes in Pittsburgh Sleep Quality Index Scores from Baseline to Week 1 in the Experimental Group

VAS for Accompanying Physical Symptoms

No statistically significant changes were observed in the distribution of symptom severity for any item when comparing pre- and post-intervention assessments. Among participants reporting moderate-to-severe headache, chest discomfort, or loss of appetite, the mean reductions in VAS scores were 41.00, 52.00, and 30.00 points, respectively (Table 8).

Table 8 Change in VAS Scores for Accompanying Subjective Physical Symptoms of Insomnia Between Baseline and 1 Week After Intervention in the Experimental Group

Economic Evaluation

The experimental group showed significant improvements in EQ-VAS (p = 0.002) and SF-36 (p = 0.007) scores. The change in activity impairment score was also significantly greater in the experimental group than in the control group (p = 0.004) (Supplement 17). From the healthcare system perspective, the average 6-week medical costs were approximately KRW 192,000 (USD 128) higher in the experimental group (KRW 304,000; USD 203) compared to the control group (KRW 112,000; USD 75). The difference was mainly attributable to the cost of 12 EA treatments, whereas conventional medicine service costs were comparable between groups (approximately KRW 77,000; USD 51). Over the 6-week time horizon, mean QALYs were 0.1010 in the EA group and 0.0969 in the control group, yielding an incremental QALY of 0.0040. The ICER for EA compared with usual care in patients with insomnia was KRW 48,000,000 (USD 32,000) per QALY over 6 weeks from the healthcare system perspective in Korea. This ICER is higher than the commonly cited willingness-to-pay threshold typically applied in Korean health economic evaluations.

Feasibility Evaluation

The recruitment rate in this study was 71.43%. Twenty-five participants were allocated to each of the experimental and control groups. By the follow-up assessment, completion rates were of 96.00% and 72.00%, respectively in experimental and control group. The adherence rate in the experimental group was 96.00%.

Safety and Adverse Events

During the study period, a total of 18 adverse events were reported, including 16 AEs and 2 SAEs, with no significant difference between groups regarding overall incidence (Table 9). Nine AEs occurred in the experimental group: one oral cyst, one ankle sprain, four upper respiratory symptoms (cough, rhinorrhea, sore throat, nasal congestion), one knee pain, one cheilitis, and one low back pain. In the control group, seven AEs were reported, comprising knee pain, elbow pain, postherpetic neuralgia, periodontal disease, cold, and sinusitis. All AEs were judged to be unrelated to the intervention.

Table 9 Comparison of Adverse Event Incidence Between the Experimental and Control Groups

Two SAEs occurred only in the control group, both resulting in withdrawal: hospitalization due to a fall and hospitalization due to the recurrence of a pre-existing condition. Both SAEs were judged unrelated to the study intervention.

Discussion

EA treatment was associated with significant improvements in subjective sleep parameters in patients with TRI.

Mean ISI scores became statistically significant compared with controls from 2 weeks (p < 0.02) and persisted through the 4-week follow-up. From 4 weeks after treatment initiation, mean ISI reduction exceeded the MCID criterion of six points, corresponding to “moderate” improvement.³⁴ After 4 weeks, significantly more participants in the experimental group showed ISI reductions exceeding MCID than controls, consistent with studies reporting significant ISI improvements following 3–4 weeks of short-term EA in insomnia patients,^25,28 suggesting that EA is associated with meaningful improvements in subjective perceptions of insomnia severity, including in patients unresponsive to long-term pharmacotherapy.

PSQI scores showed significant sleep quality improvement in the experimental group at follow-up compared with controls (mean reduction, 2.8 points), approaching the MCID criterion of 3 points.³⁵ Findings align with previous multicenter studies reporting significant PSQI improvements following 3–4 weeks short-term EA.²⁵

Notable findings included significant ISI and somatic symptom improvements after two sessions, suggesting an early association between EA treatment and reductions in subjective symptom burden. VAS score improvements for moderate-to-severe somatic symptoms—headache, chest discomfort, and appetite loss—greatly exceeded the Substantial Clinical Benefit (SCB) criterion of 22.7 points,³⁶ suggesting EA broadly improves sleep-related somatic and insomnia symptoms. This early improvement could enhance treatment compliance and inform strategies for managing TRI in clinical settings.

Sustained morning refreshment improvement in sleep-diary analysis is noteworthy. Morning refreshment reflects fatigue resolution and negatively correlates with sleep-quality and depression.³⁷ Post-awakening fatigue causes insomnia patients to interpret stimuli negatively and feel persistently sleep-deprived, reinforcing poor sleep quality perceptions.³⁸ Therefore, morning refreshment improvement is expected to improve patients’ negative cognitive patterns and emotional states beyond increased sleep time, potentially benefiting long-term insomnia outcomes.

Sleep onset latency continuously decreased from 2 weeks post-EA, improving by 12–14 min at completion and follow-up. Similar improvements in subjective sleep onset latency have been reported in previous EA studies.²⁵

Total sleep time increased significantly at completion and follow-up, indicating patients perceived improved sleep onset latency as increased sleep time.

No significant changes were observed in actigraphy-based sleep parameters. Given the absence of a sham control group and participant blinding, the possibility that the observed subjective improvements might reflect non-specific or placebo-related effects, or changes in sleep perception rather than direct physiological changes in sleep architecture, cannot be ruled out. However, objective measurements may not fully capture patients’ actual sleep distress or degree of improvement.³⁹ Moreover, the discrepancy between subjective and objective sleep outcomes may also be affected by limitations of assessment tools.⁴⁰ Quiet wakefulness in insomnia involves minimal movement, making differentiation from actual sleep difficult when using actigraphy.⁴¹ In addition, the Cole-Kripke algorithm, developed for adults aged 18–65,³² may have limited specificity in older populations, such as the present cohort (mean age 64).⁴²

Since the diagnosis of insomnia relies on subjective sleep complaints,^43,44 and subjective improvement positively affects quality of life and mental health,⁴⁵ improvements in subjective parameter observed in this study may represent clinically meaningful results.

Previous CBT-I studies reported sleep perception improvements as major treatment effects.^46,47 Results showing higher subjective sleep complaints are associated with deteriorated physical and mental health outcomes^48,49 support this finding’s clinical importance.

Using the 2020 per capita GDP (KRW 35,130,000; USD 23,420) as the willingness-to-pay threshold, the estimated ICER of EA (KRW 48,000,000/QALY; USD 32,000/QALY) exceeded this benchmark from a healthcare system perspective within the 6-week analytic horizon.

However, the relatively short time horizon may limit the ability to fully capture the longer-term economic and health consequences of EA. In particular, improvements in quality of life, work productivity, and activity impairment observed in the EA group suggest that potential downstream benefits, including productivity gains and reduced indirect costs, may emerge over a longer follow-up period. Therefore, future economic evaluations adopting a longer time horizon and a broader societal perspective are warranted to better assess the full economic implications of EA. The present findings should be interpreted within the context of the study design and analytic assumptions.

With growing concerns about dependency and tolerance related to long-term sleep medication use, and increasing interest in integrative treatment approaches for chronic insomnia, this study is significant as the first to provide preliminary evidence regarding the short-term clinical effects and safety of EA in patients with TRI, reflecting real-world clinical settings in which a substantial proportion of patients chronically use hypnotic medications.

This study had several limitations. First, as a small-scale pilot study with a limited sample size, it faced constraints in establishing clinical evidence for the efficacy of EA. It was designed to explore the short-term clinical effectiveness and feasibility of EA rather than to provide definitive estimates of efficacy. Second, it should be noted that although the term “treatment-resistant insomnia” was used, the study population primarily consisted of patients refractory to pharmacotherapy rather than to CBT-I, reflecting real-world clinical practice in South Korea where pharmacological treatment remains the predominant approach. Third, the absence of a sham control group and participant blinding makes it difficult to exclude the possibility that improvements in subjective sleep outcomes were influenced by non-specific or placebo-related effects. Nevertheless, improvements in subjective sleep perception may still be clinically meaningful in insomnia, as it plays a central role in both severity and treatment outcomes. Fourth, while the dropout rate was higher in the control group than in the experimental group, baseline insomnia severity did not differ between completers and dropouts, suggesting a low likelihood of systematic attrition bias. Fifth, although significant between-group differences were observed in several sleep diary-based parameters, no statistically significant differences were detected in actigraphy-based sleep measures. This subjective–objective discordance may reflect changes in sleep perception or non-specific treatment effects rather than direct physiological changes in sleep, as well as limitations of assessment tools, including recall bias in sleep diaries and reduced sensitivity of actigraphy in older populations. These findings should therefore be interpreted with caution. Sixth, the relatively short follow-up period limits conclusions regarding the durability of EA effects, and longer-term studies are needed to determine whether observed improvements are sustained. Finally, subgroup analyses based on sleep medication type were not performed and should be considered in future research studies.

Future research should include large-scale, well-designed randomized controlled trials incorporating sham controls and longer follow-up periods to better evaluate the durability and clinical relevance of EA effects in TRI. In addition, further clinical studies are warranted to examine whether EA may support safe reduction of long-term hypnotic use within integrative medicine frameworks.

Conclusion

This is the first multicenter randomized controlled trial conducted in Korea to evaluate the effectiveness and safety of EA treatment in patients with TRI.

EA was associated with improvements in subjective sleep-related outcomes and somatic symptoms, whereas no significant changes were observed in actigraphy-based objective sleep parameters. Given the absence of a sham acupuncture control, the observed benefits should be interpreted cautiously and may reflect non-specific effects rather than direct physiological changes in sleep architecture. Further randomized controlled trials incorporating sham acupuncture, larger sample sizes, and longer follow-up periods are warranted to clarify the mechanisms and clinical role of EA in managing TRI.