Cost-Effectiveness of Pharmacopuncture Versus Conventional Acupuncture for Insomnia: An RCT-Based Economic Evaluation with a Markov Model

Abstract

Introduction

Pharmacopuncture, which combines traditional acupuncture with herbal medicine, may offer synergic therapeutic effects for insomnia management; however, evidence on its economic value remains limited. This study evaluates the cost-effectiveness of pharmacopuncture for insomnia disorder compared with other acupuncture modalities in South Korea.

Methods

Two Markov models were developed from a healthcare system perspective to assess the short- and long-term cost-effectiveness of pharmacopuncture. The models included four levels of insomnia severity based on the Insomnia Severity Index (ISI). Transition probabilities, medical costs, and utility weights were derived from a pragmatic randomized controlled trial (KCT0006803). A 4-week cycle length was applied to stimulate 100 participants with moderate-to-severe insomnia. The short- and long-term models employed 8- and 24-week time horizons, respectively. Effective outcomes included the proportion of remission (ISI < 8) and quality-adjusted life-years (QALYs).

Results

Pharmacopuncture incurred an incremental cost of KRW 368,000 (USD 270) per additional remission achieved over 8 weeks. Long-term analysis showed an incremental cost-effectiveness ratio of KRW 16,940,000 (USD 12,600) per QALY. Sensitivity analyses revealed that cost-effectiveness was sensitive to pharmacopuncture pricing.

Conclusion

Pharmacopuncture can be considered cost-effective if decision-makers accept an additional cost of KRW 400,000 (approximately USD 280) per responder over 2 months. At a willingness-to-pay threshold of KRW 27,660,000 (approximately USD 21,740), pharmacopuncture is a cost-effective option for managing insomnia compared with other acupuncture modalities.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40120-026-00936-w.

Plain Language Summary

Pharmacopuncture combines acupuncture with herbal medicine and is commonly used to treat insomnia in Korea, but is not currently covered by the National Health Insurance. We conducted a clinical trial to compare pharmacopuncture with acupuncture and electroacupuncture, which are covered by insurance. We also evaluated whether pharmacopuncture provides good value for money by analyzing its costs and health effects. Pharmacopuncture led to additional costs of about USD 270 per extra patient who achieved remission over 8 weeks and USD 12,600 per quality-adjusted life year (QALY) gained over 6 months. Based on a commonly used threshold in Korea (USD 21,740 per QALY), pharmacopuncture is likely to be a cost-effective treatment compared with other acupuncture therapies. These findings suggest that pharmacopuncture could be considered for future insurance coverage, while further research may help confirm its value in broader populations.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40120-026-00936-w.

Key Summary Points

Why carry out this study?

Pharmacopuncture, which integrates acupuncture with herbal medicine, has been proposed as a potential treatment option for insomnia; however, its economic value has not been systematically evaluated

This study evaluated the cost-effectiveness of pharmacopuncture compared with traditional acupuncture and electroacupuncture using data from a pragmatic randomized trial (KCT0006803)

What was learned from the study?

Pharmacopuncture incurred an additional medical cost of approximately USD 270 per patient achieving remission over 8 weeks

From a healthcare system perspective, the long-term incremental cost-effectiveness ratio (ICER) was estimated at USD 12,600 per quality-adjusted life year (QALY), below the willingness-to-pay threshold of USD 21,740

Pharmacopuncture may represent a cost-effective option for insomnia compared with traditional acupuncture and electroacupuncture

Introduction

Insomnia is a significant public health issue, with symptoms affecting 30–50% of adults worldwide [1]. Insomnia disorder is diagnosed when an individual experiences difficulty initiating or maintaining sleep or early morning awakening, at least 3 times per week for 3 months, accompanied by distress or clinically significant impairment in daytime functioning [1, 2]. It negatively impacts patients’ quality of life (QoL) and work productivity and exacerbates comorbid conditions [3]. Moreover, it imposes a substantial economic burden on society [3]. Traditional treatments for insomnia include cognitive behavioral therapy (CBT-I) and short-term pharmacotherapy (zolpidem, benzodiazepines, or antidepressants) [4, 5]. CBT-I is recommended as first-line but underutilized because of limited trained providers and low patient awareness [6]; however, pharmacotherapy has safety concerns, including cognitive decline, fall risk, and dependence [7].

Complementary and alternative medicine (CAM) offers options with fewer serious side effects for patients dissatisfied with sleep medications or seeking dose reduction [8, 9]. Visits to providers of Korean medicine (KM), a widely used form of CAM, for insomnia management increased 52% between 2010 and 2016 in South Korea [10]. According to the 2021 KM clinical practice guidelines, acupuncture is one of the recommended treatment options for insomnia [11]. Several acupuncture modalities are available, including traditional acupuncture, electroacupuncture (EA), and pharmacopuncture (PC) [11]. In line with the KM clinical practice guidelines, a 2016 survey found that most KM doctors employed acupuncture for insomnia treatment, with traditional acupuncture being the most commonly used (98%), followed by PC (30%) and EA (15%) [8].

Traditional acupuncture has long been used to support individuals experiencing insomnia in East Asia and is one of the most widely used CAM therapies for insomnia in the US, Europe, and Australia [12, 13]. It involves inserting thin needles at specific acupoints to restore physiological balance and improve sleep, potentially improving sleep by enhancing central inhibition through increased serotonin and gamma-aminobutyric acid levels [14]. EA, a modified form of traditional acupuncture, delivers electrical stimulation through inserted needles at acupoints [12]. This technique may amplify therapeutic effects by enhancing acupoint stimulation through the body’s natural electrical conductivity [12]. A recent meta-analysis demonstrated that both traditional acupuncture and EA significantly improve subjective sleep quality with no significant differences between the two modalities [15].

PC combines traditional acupuncture with herbal medicine by injecting herbal extracts into acupoints [16]. This approach offers synergistic effects by integrating the physical stimulation of acupuncture with the pharmacological actions of herbal extracts [16]. Herbal extracts used in PC are tailored to each patient's condition based on pattern identification, enabling personalized treatment [17]. A meta-analysis demonstrated PC improved self-reported sleep quality more than traditional acupuncture [18]. PC benefits patients with dysphagia or those avoiding oral medicine [19]. Because the extracts are directly absorbed and bypass the gastrointestinal tract, PC may provide a more rapid onset and improved dose control [19]. For these reasons, PC is widely adopted in KM practice, with Placenta Hominis extract and Hwangryunhaedok-tang being frequently utilized formulations [8, 20].

Despite the potential of PC, evidence of its clinical and economic value for insomnia remains insufficient [18]. Unlike traditional acupuncture and EA, PC is not currently covered by the Korean National Health Insurance (NHI). Well-designed studies comparing PC with established insomnia treatments are needed to support rational allocation of limited healthcare resources. Such evaluations should reflect real-world practice since PC treatment regimens are highly individualized. Variations exist in the types of herbal solutions, dosages, and the duration or frequency of administration, depending on the patient’s clinical condition and the practitioner’s judgment [20]. Because PC is not covered by the NHI, its price varies considerably across KM clinics [16]. Therefore, it is necessary to determine the appropriate price range that demonstrates cost-effectiveness.

The objective of this study was to assess the cost-effectiveness of PC compared with other acupuncture modalities for the treatment of insomnia disorders. To reflect real-world clinical practice, we utilized data from a previously conducted pragmatic randomized controlled trial (pRCT; KCT0006803). Using these data, we developed a Markov simulation model to evaluate the short- and long-term cost-effectiveness outcomes of PC.

Methods

Clinical Trial Design and Participants

This economic evaluation was conducted alongside a pragmatic, randomized, active-controlled trial (KCT0006803; registration date: 2021-12-02), described in detail in a previous study [21]. Eligible participants met DSM-5 criteria for insomnia, had an Insomnia Severity Index (ISI) ≥ 15, and were aged 19–80 years. They were recruited from three KM hospitals and randomly assigned (2:1 ratio) to either the PC or conventional acupuncture (CA) group. Everyone received personalized treatment based on their clinical status. The PC group received one of the following treatments: Placenta Hominis, Hwangryunhaedok-tang, Venenum Bufonis, Dendropanax morbifera, or Se-sim, while the CA group received traditional acupuncture or EA treatment. The treatment period was 4 weeks (10 sessions) followed by an additional 4-week observation phase. All participants could continue prior sleep medications, but they were withdrawn if hypnotic dosage/type changed.

This study was reviewed and approved by the institutional review board at each study site (Pusan National University Korean Medicine Hospital: PNUKHIRB-2021004, Semyung University Korean Medicine Hospital in Chungju: SMCJH 2110-12, and Dong-eui University Korean Medicine Hospital: DH-2021-13). Informed consent was obtained from all participants.

Effectiveness Measurement

Clinical effectiveness and QoL outcomes were collected at baseline, post-treatment, and follow-up. The ISI was used to evaluate clinical effectiveness. The ISI is a reliable and valid instrument widely used to assess insomnia severity [22, 23]. The ISI scores range from 0 to 28, with higher scores indicating more severe insomnia symptoms. Based on the ISI scores, insomnia severity was classified as no (ISI: 0–7), mild (ISI: 8–14), moderate (ISI: 15–21), or severe insomnia (ISI: 22–28).

Health-related QoL (HRQoL) was assessed using a Health-Related Quality of Life Instrument with Eight Items (HINT-8), a tool specifically developed to measure HRQoL in the Korean population. The HINT-8 consists of eight domains—climbing stairs, pain, vitality, working, depression, memory, sleep, and happiness—each rated on four levels. Responses were converted into a utility index using the Korean tariff [24], and these values were used to calculate quality-adjusted life years (QALYs).

Cost Measures

All relevant medical costs related to insomnia management were included in the analysis. Medical costs comprised expenses incurred for KM therapies during the 4-week clinical trial and costs related to participants’ use of Western medicine (WM) services for insomnia throughout the 8-week study period. All costs were estimated in KRW for the reference year 2024. Discounting was not applied because of the short study duration. Detailed cost information is provided in Supplementary Material (Table S1).

RCT-associated costs included diagnostic fees and acupuncture treatment costs. As PC is not covered by Korea’s NHI, its costs were calculated by combining procedure and material fees obtained from the Pusan National University Korean Medicine Hospital (Supplementary Material Table S1). Material costs were based on the type of PC each participant administered. Costs for NHI-reimbursed acupuncture therapies (traditional acupuncture and EA) included both reimbursement fees listed in the NHI fee schedule and patient co-payments (Supplementary Material Table S1).

WM-related costs were calculated for participants who used WM services, including hospital visits and prescribed hypnotic medications. WM visit costs were estimated using data from the Disease Statistics by Subcategory (F51, G47) provided by the HIRA website (Supplementary Material Table S1). Drug costs were calculated based on the type and frequency of prescribed hypnotics, with unit prices referenced from published weighted average prices for NHI-covered drugs [25].

Analysis of Cost-effectiveness

This economic evaluation was conducted and reported in accordance with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) 2022 checklist [26]. Two Markov models were developed to compare the cost-effectiveness of PC relative with other acupuncture modalities from a healthcare perspective. Both models incorporated transition probabilities, medical costs, and state-specific utility weights derived from the pRCT (KCT0006803). Simulation results are presented as incremental cost-effectiveness ratios (ICERs), comparing differences in mean costs and effectiveness between the PC and CA groups.

The two models differed in effectiveness outcomes and time horizons. Model 1 used an 8-week horizon, matching the pRCT, and the remission rate (ISI < 8) as the effectiveness measure. Model 2 applied a 24-week time horizon and QALYs, consistent with previous economic evaluations of insomnia treatments, in which follow-up periods typically did not exceed 6 months [3]. This duration also reflects real-world clinical practice patterns, in which PC is commonly administered for up to 6 months, with clinical guidelines recommending reassessment at 6 months [1, 20]. The ICER per QALY was compared with a KRW 27,660,000 threshold, the median ICER of pharmaceuticals in the national benefit package (2019–2023) based on reimbursement decisions with pharmacoeconomic evidence [27].

Our state-transition models included four health states (Fig. 1): no (ISI: 0–7), mild (ISI: 8–14), moderate (ISI: 15–21), and severe (ISI: 22–28) insomnia, based on the validated ISI cutoffs [22]. A 4-week cycle length was applied to simulate 100 patients diagnosed with insomnia disorder according to the DSM-5 criteria. The model’s initial distribution reflected proportions observed in our clinical trial, with 85% of participants entering the model in moderate and 15% in severe states. An ISI < 8 indicated remission. In the base case analysis, remission was assumed to persist without relapse during the 6-month Markov model. Evidence from sleep trajectory studies suggests that individuals with restored sleep are more likely to remain symptom-free than to relapse over time [28]. To explore structural uncertainty regarding the maintenance of remission, an additional scenario allowing for relapse was evaluated.

Fig. 1.

Diagram of the Markov model. ISI, Insomnia Severity Index. The model included four health states defined by ISI categories: no (0–7), mild (8–14), moderate (15–21), and severe (22–28) insomnia. Transition probabilities were derived from observed transitions in the clinical trial and allowed to vary between treatment and follow-up phase. In the base case analysis, remission was assumed to persist without relapse over the 6-month time horizon

Transition probabilities, utility weights, and costs were derived from our pRCT. Transition probabilities were calculated directly from observed transitions between insomnia severity states in the trial data, using the proportion of patients moving between states at consecutive time points. Transition probabilities were time-dependent to account for differences between the treatment and follow-up periods. As the model cycle length (4 weeks) corresponded to the observation interval, no transformation was required. Utility weights for each state were estimated using HINT-8 and ISI scores from our pRCT and applied to the Markov model. To estimate input utility values by insomnia severity (no, mild, moderate insomnia), a linear mixed-effects model was used to examine the association between ISI categories and HINT-8 scores. To account for within-subject variance, an unstructured covariance structure was selected based on its lowest Akaike information criterion value compared with an autoregressive (1) structure. Each health state was assigned a fixed utility weight that remained constant across all treatment phases (pre-treatment, post-treatment, and follow-up). KM treatment was administered only in the initial 4 weeks. Treatment-period costs (first 4 weeks) included KM and WM interventions for both groups, while follow-up costs were based on WM use for managing ongoing symptoms.

Sensitivity Analysis

Sensitivity analyses were conducted to assess the robustness of base case results to variations in procedure fees for PC, treatment duration, transition probabilities, and utility weights. In Scenario 1, PC fees were varied (KRW 15,000, 25,000, and 30,000, vs. base case KRW 20,000), yielding total per-session costs of KRW 31,000–46,000 for Placenta Hominis (the highest-cost formulation) and KRW 19,000–34,000 for Venenum Bufonis (the lowest-cost formulation). Scenario 2 extended treatment duration to 8, 12, and 24 weeks, after which remission patients discontinued KM treatment.

Scenario 3 assumed no difference in the maintenance of treatment effects between the two groups during the post-intervention period. As the follow-up estimates in the base case model were derived from a limited 4-week observation period, they may be insufficient to represent the long-term maintenance effects. Accordingly, both groups were assigned identical transition probabilities during the post-treatment period rather than group-specific values. Scenario 4 relaxed the base case assumption of stability over 6 months in symptom-free patients. In this scenario, the remission maintenance rate observed between weeks 4 and 8 was extrapolated over the remainder of the follow-up period. Patients who did not maintain remission were assumed to have relapsed to a mild insomnia state. This assumption was informed by evidence suggesting that relapse from a symptom-free state more commonly leads to milder insomnia states than to severe insomnia [28].

Finally, Scenario 5 incorporated utility values stratified by insomnia severity as reported by Chalet et al. (2023), who provided EuroQoL 5-Dimension (EQ-5D)-based utility weights according to ISI severity using a large population-based sample (n = 8920) [29]. Details of the input parameters used in the sensitivity analyses are presented in Supplementary Material (Table S2).

Statistical Analysis

This economic evaluation used per-protocol data from patients completing all ten treatment sessions. Baseline characteristics were summarized as means and standard deviations or frequencies, with independent t-tests and chi-squared tests used to assess group homogeneity at baseline. Within-group changes in ISI and HINT-8 from baseline to endpoint were evaluated using paired t-tests. Between-group differences in outcome changes over the 4- and 8-week periods were analyzed using independent t-tests. Significance was set to 0.05 (two-tailed). All statistical analyses were performed using RStudio 4.5.0 and Microsoft Excel.

Results

Baseline Characteristics During pRCT

The CONSORT flowchart is shown in Fig. 2. A total of 138 participants were randomized to the PC group (n = 91) or the CA group (n = 47) in a 2:1 ratio. Of these, 126 (n = 83 in PC, n = 43 in CA) who completed all ten sessions were included in our study. At baseline, among the 98 participants with ISI scores > 15, 86.7% (n = 85) had moderate and 13.3% (n = 13) had severe insomnia. Table 1 presents the baseline characteristics of the participants. The average age of both groups was in the early 50s, and approximately 80% of participants were female. Most participants were in the moderate or severe insomnia category (mean ISI scores: 18 in PC and 16 in CA). No significant differences were observed between the PC and CA groups in demographics, clinical outcomes, or QoL at baseline.

Fig. 2.

CONSORT flow chart of the pragmatic RCT. PC, pharmacopuncture; CA, conventional acupuncture (traditional acupuncture, electroacupuncture); ISI, Insomnia Severity Index; HINT-8, Health-related Quality of Life Instrument with Eight Items

Table 1.

Baseline characteristics for study participants (N = 126)

	PC (N = 83)	CA (N = 43)	p Value
Age; years, mean (SD)	51.8 (14.9)	54.3 (14.4)	0.3667
Gender, n (%)			0.5310
Female	69 (83.1%)	33 (76.7%)
Male	14 (16.9%)	10 (23.3%)
Smoker, n (%)	6 (7.2%)	2 (4.7%)	0.8592
Alcohol use, n (%)	34 (41.0%)	16 (37.2%)	0.8287
Regular exercise, n (%)	69 (83.1%)	32 (74.4%)	0.3538
ISI, mean (SD)	17.6 (4.0)	16.3 (3.7)	0.0730
HINT-8, mean (SD)	0.786 (0.091)	0.788 (0.072)	0.8695

PC, pharmacopuncture; CA, conventional acupuncture (traditional acupuncture, electroacupuncture); ISI, Insomnia Severity Index; HINT-8, Health-related Quality of Life Instrument with Eight Items

Treatment Outcomes and Costs During pRCT

Table 2 presents the changes in treatment outcomes. ISI scores decreased significantly from baseline to post-treatment and follow-up in both groups. During the 4-week treatment period, the decrease in ISI scores was significantly greater in the PC group than in the CA group (10.46 vs. 8.19; p = 0.005). From baseline to week 8, reductions were 10.63 in the PC group and 8.30 points in the CA group, respectively, with a significant between-group difference (p = 0.007). In terms of treatment remission (ISI < 8), 64% of patients in the PC group and 58% in the CA group achieved a symptom-free state at the 4-week treatment endpoint. Upon completion of the follow-up phase, remission was sustained in > 50% of patients in the PC group compared with < 50% in the CA group. HINT-8 scores in the PC group improved from baseline (0.786) to week 4 (0.823), with gains maintained at week 8. The CA group showed an initial improvement at week 4 (0.821), followed by a modest decline at week 8 (0.813). Although the PC group exhibited numerically greater HINT-8 improvement over the 4 weeks (0.037 vs. 0.033) and the 8 weeks (0.037 vs. 0.024), these differences were not statistically significant (p = 0.676, p = 0.319).

Table 2.

Treatment outcomes during the clinical trial period

Treatment outcomes: raw score
	Baseline	After Tx	Changea (base ~ Tx)	p Value	After F/u	Changeb (base ~ F/u)		p Value
ISI
PC	17.59 ± 3.95	7.13 ± 4.45	10.46 ± 5.05	0.005	6.96 ± 4.45		10.63 ± 4.85	0.007
CA	16.30 ± 3.69	8.12 ± 4.03	8.19 ± 3.79		8.00 ± 4.03		8.30 ± 4.32
HINT-8
PC	0.786 ± 0.091	0.823 ± 0.088	0.037 ± 0.071	0.676	0.823 ± 0.010		0.037 ± 0.074	0.319
CA	0.788 ± 0.072	0.821 ± 0.075	0.033 ± 0.042		0.813 ± 0.088		0.024 ± 0.063

Treatment outcomes: remission rate
	Baseline ~ after Tx				Baseline ~ after F/u
	No./total				No./total
PC	53/82d	(64.4%)			50/82d	(61.0%)
CA	25/43	(58.1%)			19/43	(44.2%)

Tx, treatment; F/u, follow-up; PC, pharmacopuncture; CA, conventional acupuncture (traditional acupuncture, electroacupuncture); ISI, Insomnia Severity Index; HINT-8, Health-related Quality of Life Instrument with Eight Items

^aChange from baseline to post treatment

^bChange from baseline to end of follow-up

^cRemission defined as Insomnia Severity Index score < 8

^dParticipants with baseline ISI < 8 excluded from remission analysis

From the healthcare perspective, the mean medical costs for KM and WM services over the 8-week study period were higher in the PC group (KRW 487,000) by approximately KRW 78,000 compared to the CA group (KRW 409,000). The average KM costs for ten sessions over 4 weeks were KRW 475,000 in the PC group and KRW 400,000 in the CA group. About 10% of patients in both groups used WM services, incurring a mean costs of approximately KRW 6000 over the 4 weeks, with no notable difference between the two groups.

Estimated Parameters from pRCT Data

Table 3 summarizes model inputs from the pRCT, including transition probabilities for state-to-state movement, utility weights, and medical costs. The 4-week PC treatment (10 sessions) resulted in a remission rate (ISI < 8) of 65% among participants with moderate insomnia and 50% among those with severe insomnia. After the 4-week CA therapy (10 sessions), the probability of remission was 52% in patients with moderate insomnia, while no remission was observed in those with severe insomnia.

Table 3.

Summary of input parameters for the base case model

		Base case
		PC	CA
Transition probability (1 cycle = 4 weeks)
Assumption	No → no insomnia	1.00	1.00
Treatment phase	Severe → no insomnia (remissiona)	0.50	0.00
	Severe → mild	0.30	0.67
	Severe → moderate	0.20	0.33
	Moderate → no insomnia (remissiona)	0.65	0.52
	Moderate → mild	0.27	0.32
	Moderate → moderate	0.08	0.16
Follow-up phase	Moderate → mild	0.43	0.80
	Moderate → moderate	0.57	0.20
	Mild → no insomnia (remission)	0.30	0.15
	Mild → mild	0.57	0.77
	Mild → moderate	0.13	0.08
Medical costs for 4 weeks (KRW)
Treatment phase (KM and WM)		481,000	403,000
Follow-up phase (WM)		6100	6200
Utility
No insomnia		0.844	0.844
Mild		0.791	0.791
Moderate		0.784	0.784

PC, pharmacopuncture; CA, conventional acupuncture (traditional acupuncture and electroacupuncture); KM, Korean medicine; WM, Western medicine

^aRemission defined as Insomnia Severity Index score < 8

^bTransition probabilities were derived from observed transitions between insomnia severity states across consecutive time points in the clinical trial. No relapse was assumed in the base case analysis

^cMedical costs were estimated using trial-based data (PC pricing, WM utilization) and national fee schedules (acupuncture, electroacupuncture, hypnotics, and WM visit costs)

^dUtility values were derived from HINT-8 scores using ISI-based severity categories estimated from the clinical trial using a linear mixed-effects model

According to the pRCT data, no patient in either group remained in a state of severe insomnia after 4 weeks of treatment. During the follow-up period, the proportion of patients who sustained remission was 81% in the PC group and 64% in the CA group (Supplementary Material Table S2). The probability of achieving remission from mild insomnia was 30% in the PC group and 15% in the CA group.

Based on the pRCT, the estimated utility values were 0.844 for symptom-free, 0.791 for mild, and 0.784 for moderate insomnia. These utility weights were applied uniformly across both groups, irrespective of treatment allocation. During the treatment phase, the total medical costs (KM therapy and WM services) per patient were KRW 481,000 in the PC group and KRW 403,000 in the CA group. During the follow-up phase, the WM-related costs per cycle were KRW 6100 in the PC group and KRW 6200 in the CA group.

Cost-effectiveness Outcomes

At baseline, 100 patients were assumed to have either moderate (n = 85) or severe (n = 15) insomnia. Based on transition probabilities derived from our pRCT data, 63% of patients in the PC group and 44% in the CA group achieved a symptom-free status (remission) after ten sessions over 4 weeks. After one subsequent cycle without KM therapy, 71 patients in the PC group and 50 patients in the CA group remained symptom-free.

Table 4 summarizes the simulated cost-effectiveness (Model 1) and cost-utility (Model 2) outcomes. In Model 1, the ICER was estimated at KRW 368,000 per additional participant achieving a positive treatment response over 8 weeks. In Model 2, the 4-week PC treatment yielded 0.0046 incremental QALYs over 6 months at an additional medical cost of KRW 77,000 compared with CA, resulting in an ICER of KRW 16,940,000 per QALY gained. The ICER for the base case falls below the cost-effectiveness threshold of KRW 27,660,000.

Table 4.

Base case cost-effectiveness results

	Expected effectiveness	Expected costs	Incremental effectiveness	Incremental costs	ICER
Model 1: Treatment response rate over an 8-week period (4-week treatment plus 4-week follow-up)
PC	0.71	KRW487,000	0.21	KRW78,000	KRW368,000
CA	0.50	KRW409,000	–	–	–
Model 2: QALYs for 24 weeks (4-week treatment plus 20-week follow-up)
PC	0.3838	KRW511,000	0.0046	KRW77,000	KRW16,940,000
CA	0.3793	KRW434,000	–	–	–

ICER, incremental cost-effectiveness ratio; PC, pharmacopuncture; CA, conventional acupuncture (traditional acupuncture, electroacupuncture); QALYs, quality-adjusted life-year

Sensitivity Analysis Results

Figure 3 and Supplementary Material (Table S3) show the one-way deterministic sensitivity analysis. For Scenario 1, when the reimbursement fee for the PC procedure was set at 15,000, the ICER was estimated to be KRW 6,000,000 per QALY gained, which was below the cost-effectiveness threshold. However, ICERs exceeded the threshold when practice fees increased to KRW 25,000 (KRW 27,881,000 per QALY) and KRW 30,000 (KRW 38,822,000 per QALY). In Scenario 2, extending treatment to 8, 12, and 24 weeks reduced ICERs and remained below the cost-effectiveness threshold (Supplementary Material Table S3). With a longer intervention period, the differences in both QALYs and medical costs between the groups diminished.

Fig. 3.

Tornado diagram of one-way sensitivity analysis of ICER estimates across scenario assumptions. ICER, incremental cost-effectiveness ratio; WTP, willingness to pay; QALY; quality-adjusted life years; PC, pharmacopuncture; CA, conventional acupuncture (traditional acupuncture, electroacupuncture); SA, scenario analysis; F/u, follow-up. One-way sensitivity analysis showing the impact of key parameters on the ICER for Model 2. Bars represent the range of ICER values obtained by varying each parameter. The dashed line indicates the willingness-to-pay threshold (27,660,000 KRW/QALY), and the dotted line represents the base case ICER (16,940,000 KRW/QALY)

Scenario 3, assuming identical transition probabilities for both groups throughout the post-intervention period, diminished incremental QALY gains compared with the base case model; however, the estimated ICERs remained below the threshold (Supplementary Material Table S3). Scenario 4, which incorporated the remission maintenance probability observed in the pRCT, produced increased incremental QALY gains (0.062 over 24 weeks) compared with the base case model, resulting in a more favorable ICER (Supplementary Material Table S3). In Scenario 5, utility weights were replaced with values reported by Chalet et al. (2023). This substitution led to a slight reduction in incremental effectiveness (incremental QALYs gained over 24 weeks: 0.0033) compared with the pRCT-based base case (0.0046). Consequently, the ICER increased by 37% relative to the base case model (KRW 23,195,000 vs. KRW 16,940,000), approaching the cost-effectiveness threshold (Supplementary Table S3).

Discussion

This study evaluated the cost-effectiveness of PC versus other acupuncture modalities for insomnia using a state-transition model based on a pRCT (KCT0006803). Both intervention groups showed clinically meaningful improvements in insomnia symptoms after ten treatment sessions over 4 weeks, with mean reductions in ISI scores (PC: − 10.46, CA: − 8.19) exceeding the 6-point threshold commonly used to define meaningful change (MCID) [30]. By the end of the clinical trial (8 weeks), participants receiving PC demonstrated significantly greater improvement than those receiving traditional acupuncture/EA (ISI: − 10.63 vs. − 8.30, p < 0.05). The remission rates at 4 weeks were also higher in the PC group among patients with severe insomnia (50% vs. 0%) and moderate insomnia (65% vs. 52%). Similar patterns were observed in the 6-month remission rates (61.0% vs. 44.2%). The 6-month remission rate for PC (61%) was comparable to or higher than previously reported rates for CBT-I (55%–62%) [31]. The remission rates of traditional acupuncture and EA (44.2%) were similar to those reported for hypnotic pharmacotherapy (48%) [31]. However, these comparisons should be interpreted with caution, as differences in study design, patient populations, and intervention delivery may influence the observed outcomes. Furthermore, PC demonstrated more sustained treatment effects than CA during the non-intervention period. The difference in remission rates between PC and CA was more pronounced at follow-up (61.0% vs. 44.2%) than immediately post-treatment (64.4% vs. 58.1%), indicating better maintenance of effect with PC.

In the short-term analysis (Model 1), PC demonstrated superior effectiveness but incurred higher medical costs than CA during the 8-week study period. The ICER indicated that PC required approximately KRW 370,000 per additional treatment responder over 8 weeks, relative to CA. Given the absence of an established cost-effectiveness threshold for insomnia symptom improvement, the cost-effectiveness of PC remains indeterminate. Nevertheless, this finding provides valuable information on the cost required to achieve additional treatment benefits and may support informed decision-making on individual or system-level willingness to pay. For the long-term analysis (base case Model 2), we employed a 24-week (6-month) time horizon with QALYs as the effectiveness measure. The resulting ICER of KRW 17,000,000 per QALY falls below the median ICER of KRW 28,000,000 for reimbursed pharmaceuticals in Korea (2019–2023), suggesting that PC may be considered cost-effective from a Korean healthcare system perspective. This indicates that the additional health gains associated with PC may be achieved at an acceptable cost within the current reimbursement landscape.

This study assessed the robustness and generalizability of key model parameters, including the costs of non-reimbursed interventions (PC), follow-up transition probabilities, and utility weights. Sensitivity analysis revealed that the cost-effectiveness of PC is highly sensitive to the pricing structure. As PC is not covered by NHI, its pricing varies substantially across KM providers. In the base case model, we used a practice fee of KRW 20,000 to reflect the costs of the institutions participating in our pRCT. PC remained cost-effective when reduced by 25% (KRW 15,000), suggesting competitive pricing could enhance its economic attractiveness. Conversely, when practice fees for PC increase by 25% and 50% (KRW 25,000 and KRW 30,000, respectively), the intervention exceeded the acceptable cost-effectiveness thresholds. This price sensitivity suggests that government reimbursement for PC could improve both cost-effectiveness and patient accessibility. Unlike the current non-reimbursement status, in which pricing varies across providers, the inclusion of PC in the NHI system would establish uniform reimbursement rates, potentially ensuring cost-effective treatment delivery and reducing financial barriers for patients seeking insomnia care.

Extending treatment duration increased the number of symptom-free patients in both groups, resulting in reduced medical costs per responder and enhanced overall effectiveness (Scenario 2). The PC group achieved a more rapid remission than the CA group, further reinforcing the cost-effectiveness of PC for insomnia treatment. However, these findings should be interpreted with caution, given the potential influence of treatment duration on transition probabilities and the assumption of no post-remission relapse. Scenario 3 applied identical transition probabilities for both groups during the follow-up as a conservative measure to account for potential bias introduced by extrapolating short-term (4 weeks) outcomes. Despite this conservative assumption, the ICERs remained below the cost-effectiveness threshold. In Scenario 4, permitting transitions from no insomnia to mild insomnia yielded results that were more favorable to PC cost-effectiveness than the base case. The utility weight for mild insomnia observed in our pRCT (0.791) was consistent with that reported in a previous cohort study (0.790) [29], whereas differences between the no and moderate insomnia states may reflect demographic variations across countries, age groups, and sexes. Nevertheless, Scenario 5, which incorporated utility inputs from a larger sample dataset, continued to support the cost-effectiveness of PC.

To our knowledge, this is the first trial-based Markov model economic evaluation of PC for reducing insomnia symptoms. In the absence of directly comparable studies, we contextualized our findings against prior cost-effectiveness studies of CBT-I, the recommended first-line treatment for insomnia. Focusing on ISI-based responder outcomes, our study showed a smaller incremental effect (ΔE = 21%) than the Baka et al. (28%) and Buntrock et al. (30%) studies, which may reflect the shorter study duration (8 weeks vs. 6 months) and the use of an active comparator [32, 33]. In contrast, prior studies employed comparators such as usual care or waitlist, which may have inflated effect estimates [32, 33]; therefore, our findings may represent more conservative estimates. Furthermore, differences in outcome definitions—responder (≥ 8-point reduction in ISI) vs. remission (ISI < 8)—may also influence the magnitude of observed effects and should be considered when interpreting the reported ICERs (€764 per responder [2018, The Netherlands] and €650 per remission [2013, Germany]).

A similar pattern was observed in the study by Kim et al., which conducted a model-based economic evaluation using clinical trial data in a Korean setting [34]. Kim et al. reported an ICER (KRW 8,719,727 per QALY) lower than ours (KRW 16,940,000 per QALY), despite comparable time horizons (27 weeks) and cost-effectiveness at a willingness-to-pay threshold of approximately KRW 30 million per QALY. This difference was primarily driven by incremental effectiveness (ΔE = 0.0091 vs. 0.0064), while incremental costs were similar (KRW 79,691 vs. 77,000). Despite differences in analytical perspective (societal vs. healthcare), ICER differences were largely driven by effectiveness rather than cost. Notably, Kim et al. used usual care (sleep hygiene education) as the comparator [34, 35], whereas our study employed an active comparator, which may have attenuated the incremental effectiveness and resulted in more conservative estimates.

In line with our findings, Baka et al. reported clinically meaningful improvements in insomnia severity that were not reflected in QALY gains [32]. This may be attributable to the limited sensitivity of the EQ-5D-5L in capturing QoL changes associated with improvements in insomnia symptoms [32, 36]. To address this limitation, we employed HINT-8 rather than EQ-5D-5L, which demonstrated greater sensitivity in detecting health state variations and showed substantially lower ceiling effects in our study (24% vs. one patient reporting full health at baseline). By reducing ceiling effects, HINT-8 may better capture improvements in HRQoL and provide more accurate cost-effectiveness estimates. As a preference-based instrument developed for the Koreans [37], it may also better reflect culturally relevant aspects of health.

This study has several strengths. First, our analysis was grounded in data from a pRCT, which provides empirical evidence that closely reflects real-world clinical practice. Unlike explanatory trials, pragmatic designs accommodate the inherent complexities of routine care—patient heterogeneity, comorbidities, and resource constraints—making it the preferred approach for economic evaluations. Our study design captures the individualized nature of the KM practice, where practitioners select PC types, acupoints, and injection volumes based on patient-specific clinical presentations. Institutional variations in PC availability were also incorporated into the cost estimations, reflecting actual resource utilization patterns. Furthermore, the pRCT design enabled the comprehensive collection of cost data encompassing all healthcare services received by the participants throughout the study period, including those related to study interventions and concurrent conventional treatments. This comprehensive approach strengthens the external validity of our findings and supports their applicability in clinical practice and healthcare policy decision-making.

Second, we addressed the inherent limitations of short-term economic evaluations by extending our analysis beyond the 8-week trial period using Markov modeling. A fundamental challenge in cost-effectiveness analysis is that ICERs become unstable when the differences in effectiveness approach zero, potentially yielding misleading estimates that tend toward infinity. This phenomenon was evident in our 8-week analysis, in which, despite statistically significant improvements in ISI scores favoring PC over CA, the incremental QALY gains remained minimal (0.0005), resulting in an ICER approaching KRW 100 million, which far exceeded the acceptable thresholds. This limitation reflects the inherent properties of QALYs as composite measures that incorporate both the quality and duration of life, making them less sensitive to short-term clinical changes. The 8-week timeframe was insufficient to capture meaningful differences in HRQoL between the two treatment groups. By extending our analysis to a 6-month horizon using Markov simulation, we obtained more stable and clinically relevant estimates that better reflected the sustained benefits of PC treatment for insomnia management.

Third, our model incorporated multiple insomnia severity states (no, mild, moderate, severe insomnia) from validated ISI classifications, providing a more nuanced representation than dichotomous outcomes. A previous economic evaluation employed simplified models that distinguished only between insomnia presence and remission, owing to limited transition parameter data in the literature. Such a binary approach fails to capture the spectrum of treatment response and may overlook clinically meaningful improvements that fall short of complete remission. Our pRCT-informed model enabled a detailed assessment of severity-specific treatment effects, revealing that patients with moderate baseline insomnia achieved higher remission rates than those with severe insomnia following 4 weeks of PC therapy. By incorporating these nuanced treatment trajectories, our model provides a more clinically realistic assessment of PC effectiveness for healthcare decision-making.

This study has limitations. First, the modeling framework used may lack generalizability. Evidence on the natural history of insomnia and utility weights stratified by severity remains limited. To our knowledge, no standardized modeling framework for insomnia currently exists that incorporates key epidemiological parameters, such as incidence, persistence, and relapse rates. This gap may be attributed to the traditional view of insomnia as a secondary symptom of other conditions rather than a distinct disorder. Furthermore, the complex and heterogeneous nature of insomnia makes it difficult to develop a universally applicable model that accurately reflects its pathophysiology. Consequently, transitions between severity states and key input parameters—such as transition probabilities and utility weights—in our model were derived from a single clinical trial (KCT0006803). This reliance on a single data source may limit the generalizability of the findings. However, the pragmatic trial design enhances external validity by more closely reflecting real-world clinical practice. To improve generalizability in further research, it is important to incorporate validated transition pathways and parameter estimates derived from broader epidemiological datasets.

Second, the relatively modest sample size introduced inherent uncertainty into the model estimates, particularly affecting transition probabilities, costs, and utility estimates. Stratification by severity substantially reduced subgroup sizes, most notably in the CA group, where only three participants presented severe insomnia at baseline, potentially compromising the precision of transition probability estimates. Such methodological challenges are inherent to subgroup analyses in limited-sample RCTs and highlight the trade-off between granular clinical categorization and statistical power. Cost and utility precision, however, were less likely to be substantially affected. Sensitivity analyses using validated utility decrements from large population-based cohorts [29] yielded consistent results. Non-intervention medical costs showed no meaningful between-group differences, suggesting a limited impact of sample size. Future studies with larger and more diverse populations are warranted to confirm these findings and provide more precise estimates.

Third, although PC improved insomnia in our study, the contribution of specific herbal extracts was unclear. Given KM’s reliance on pattern identification, evidence on pattern-specific efficacy is needed. Insomnia patterns (deficiency vs. excess) are treated with different formulations, e.g., Placenta Hominis for deficiency [17] and Hwangryunhaedok-tang for heat/fire [38]. This study did not consider pattern differentiation, limiting interpretation. Supporting evidence is mostly preclinical or from small-scale studies [39, 40], warranting further research on individual extract effects.

Finally, this evaluation was conducted within the Korean healthcare context, which may limit generalizability. Differences in cost structure and willingness-to-pay thresholds across countries may influence cost-effectiveness conclusions. In addition, PC is less commonly used outside East Asia, which may further constrain the application of these findings to other healthcare systems. Accordingly, the results should be interpreted with caution when extrapolated beyond this context.

Conclusions

The study findings demonstrated that PC was superior to other acupuncture modalities, including traditional acupuncture and EA, in reducing insomnia severity. From an economic perspective, PC represents a cost-effective therapeutic option contingent upon the acceptability of the incremental cost of KRW 400,000 per responder over 8 weeks. Long-term analysis over 6 months supports the cost-effectiveness of PC relative to alternative acupuncture interventions within Korea’s healthcare context. These findings provide compelling evidence to inform policy decisions regarding the inclusion of PC in Korean NHI coverage for insomnia. Future validation studies are warranted to corroborate these results and facilitate enhanced access to PC therapy for insomnia management.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Conceptualization: Eun Cho; methodology: Eun Cho; investigation: Eun Cho, Jung-Hwa Lim, Jae-Hyok Lee, Bo-Kyung Kim; formal analysis: Ji-Eun Lee, Chan-Young Kwon; data curation: Ji-Eun Lee, Chan-Young Kwon; visualization: Ji-Eun Lee; writing—original draft: Eun Cho, Ji-Eun Lee; writing—review and editing: Eun Cho, Ji-Eun Lee, Jung-Hwa Lim, Jae-Hyok Lee, Chan-Young Kwon, Bo-Kyung Kim; supervision: Eun Cho, Bo-Kyung Kim; funding acquisition: Bo-Kyung Kim.

Funding

This research was supported by a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant no. RS-2020-KH114669). The journal’s publication fee was funded by the authors.

Data Availability

The data analyzed in this study are not publicly available, as they were collected and analyzed in a clinical trial.

Declarations

Conflict of Interest

Eun Cho, Ji-Eun Lee, Jung-Hwa Lim, Jae-Hyok Lee, Chan-Young Kwon, and Bo-Kyung Kim declare that they have no conflict of interest.

Ethical Approval

This study was reviewed and approved by the Institutional Review Boards at each study site (Pusan National University Korean Medicine Hospital (PNUKHIRB-2021004), Semyung University Korean Medicine Hospital in Chungju (SMCJH 2110-12), and Dong-eui University Korean Medicine Hospital (DH-2021-13). Informed consent was obtained from all participants.