An economic evaluation of pharmacopuncture versus usual care for chronic neck pain: a pragmatic randomized controlled trial

A cost-effectiveness analysis was conducted with a 12-week pragmatic RCT [12]. The results of this clinical trial with the protocol have been published separately [12, 13]. This RCT was briefly conducted alongside a two-armed, parallel, multicenter (four KM Hospitals) RCT that included 101 patients across four institutions in Korea. Enrolled patients were randomly allocated to the pharmacopuncture and PT groups in a 1:1 ratio. The participants received a total of eight treatment sessions for four weeks, followed by eight weeks of follow-up. During this period, pain indicators, quality of life (QOL) indicators, and medical expenditure were investigated.

Schedule of the participants is shown in Supplementary Table A1 (see Additional file 1). The protocol has been registered with Clinicaltrials.gov (Identifier: NCT04035018 (29/07/2019)) and the Clinical Research Information Service (Identifier: KCT0004243 (26/08/2019)). This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guidelines.

Eligible patients were aged 17–70, had been experiencing neck pain for more than six months, and had a VAS score of more than five for their neck pain. The exclusion criteria for participation were as follows: having cancer metastasis or fracture in the spine; having progressive or severe neurologic deficits or other disorders that may have affected the results; suffering from other types of cancer, rheumatoid arthritis, or stroke; taking medications, such as steroids; being pregnant; having had cervical surgery during the last three months; and facing difficulties participating in the trial as determined by the researcher. The specific inclusion and exclusion criteria can be found in the study conducted by Park et al. [12].

The general rule for pharmacopuncture treatment was undergoing two sessions per week for four weeks; however, one to three sessions per week were permitted depending on the patient’s condition and the physician’s determination. The type of pharmacopuncture solution and treatment points were determined by the physician. All intervention-related matters were recorded in the patient’s electronic medical records (EMR) and case report form (CRF).

The general rules for usual care were similar to those for pharmacopuncture treatment. A Korean Health Insurance Review and Assessment’s (HIRA) analysis claimed that data in a previous study showed that combinations of various types of PT (e.g., superficial heat therapy, deep heat therapy, traction, and electrotherapy) were prescribed [14]. PT was selected based on clinical determination by a physician from those used in previous studies, while the type, duration, and area where all interventions were to be applied were recorded in the EMR and CRF.

Participants who were determined to be eligible based on the inclusion and exclusion criteria and signed the Informed consent form for this clinical study were assigned to two groups at a ratio of 1:1 using a randomization table. The randomization table was created in advance by a statistician using R studio 1.1.463 (© 2009–2018 RStudio, Inc., Boston, Massachusetts, United States). The random sequence was generated by block randomization, and the size of one block was randomly set between 2, 4 and 6.

The randomization results were sealed in an opaque envelope and delivered to each institution for storage under a double lock. Randomization and assignment of registration numbers for eligible participants was done by opening sealed envelopes. A random number assigned to each participant was recorded on an electronic chart. Due to the nature of our intervention, only evaluators were blinded to group assignments. The evaluator performed outcome evaluation in a separate space before intervention.

Both EQ-5D and SF-12 scores were measured at baseline and at intervals of 5, 8, and 12 weeks after randomization. The Korean version of the EQ-5D with verified validity was used, and utility was calculated by applying the tariffs used by Kim et al. [15]. In the trial, a 12-item short-form health survey version 2 (SF-12v2) was used to assess QOL, and the values obtained by SF-12v2 were converted to SF-6D values using the equation described by Brazier et al. [16]. The QALY was calculated using the area under the curve (AUC) approach and trapezoidal rules, using regression analyses adjusted for baseline values [17].

Supplementary Table A2 (see Additional file 1) shows the sources and unit costs of the interventions applied during the trial. Because the cost of pharmacopuncture intervention may vary across hospitals, as it is a non-reimbursed item, the fees published on various hospitals’ websites were surveyed, and a cost of 20,000 won (around $17) per session was derived. Costs of examination and PT were obtained from the 2019 HIRA Health Insurance Medical Care Benefit Expenses data [18]. The costs were calculated using the type and frequency of treatments prescribed to the participants. Among the types of PT used, the non-reimbursed items’ cost not covered by health insurance was determined by investigating the costs generally prescribed by medical institutions.

In addition to the intervention used in the trial, healthcare services used personally by the patients for neck pain were investigated at baseline and at 5, 8, and 12 weeks using a questionnaire. The questionnaire was used to survey the non-reimbursement and copayments made by the patients as well as the frequency of use. Payer reimbursement was calculated by gender and age by analyzing the records for cervicalgia (Korean Standard Classification of Diseases code: M54.2) in the 2018 HIRA-National Patient Samples (HIRA-NPS) data [19].

The transportation costs were determined through a questionnaire survey. The time spent by patients to receive treatment was surveyed, and the time cost was calculated using the time spent by each individual based on the standard wages for their corresponding gender and age given in the 2019 Survey Report on Labor Conditions by Employment Type [20].

Productivity loss was investigated using the Work Productivity and Activity Impairment – Specific Health Problem (WPAI-SHP) questionnaire [21] at baseline and 2, 3, 4, 5, 8, and 12 weeks after randomization. The WPAI-SHP is used to measure absenteeism (missed work time), presenteeism (impairment while working), overall productivity loss (absenteeism + presenteeism), and activity impairment (impairment in regular activities) in the past week due to a specific health problem (i.e., chronic neck pain) [22, 23]. The present study attempted to account for not only productivity loss among waged workers but also the productivity loss and opportunity cost associated with health problems for self-employment and household work activities. Accordingly, the WPAI was calculated by applying the overall work productivity loss for waged workers and activity impairment for all other participants. The costs of economic loss due to low productivity were estimated by multiplying the WPAI value obtained by surveying the patient with the standard wage for the corresponding gender and age [20].

As the total study period was 12 weeks, a discount rate was not applied. Moreover, an inflation rate of 0.46% for change in the consumer price index for medical care was applied to the costs estimated using the 2018 HIRA-NPS data to derive the costs for 2019. An exchange rate of 1156.4 won/dollars (as of 2019) was applied to derive values in US dollars [24].

Two perspectives were considered, and the primary analysis was conducted from a societal perspective, reflecting productivity loss due to neck pain. The societal perspective includes direct medical costs, direct non-medical costs, and costs associated with productivity loss. The analysis from a healthcare perspective included only direct medical costs and direct non-medical costs. Direct medical costs include the cost of intervention applied during the trial and the costs of informal care incurred during the trial period, such as the cost of treatment for neck pain from other Western medicine or KM institutions, the cost of analgesics, including over-the-counter drugs, and medical devices or exercise therapy. Direct non-medical costs include time and transportation costs.

A primary analysis examining incremental cost utility from a societal perspective was performed to calculate the incremental cost-effectiveness ratio (ICER) by comparing the differential mean costs and QALY between the pharmacopuncture and usual care groups. ICERs were calculated by dividing the difference in total costs by the difference in effects.

Intention-to-treat (ITT) analysis was performed, and the missing values were imputed with multiple imputations (MI) using the Markov Chain Monte Carlo method and predictive mean matching. Twenty imputed datasets were generated; the covariates for imputation were treatment allocation, gender, age, and other correlated variables. Correlated variables (≥ 0.4) were included in the imputation model. The missing values for EQ-5D and SF-6D were 2–3%. The mice package in R version 4.0.1 was used for the imputation of missing values.

The significance level was set to 0.05 for all statistical analyses, and SAS 9.4 (© SAS Institute, Inc., Cary, NC, USA) and R studio 1.1.463 (© 2009–2018 RStudio, Inc., Boston, Massachusetts, United States) were used for the statistical analyses.

The uncertainty of ICER and costs was estimated using the bootstrap residual technique [17]. To obtain valid bootstrap inferences while dealing with multiple imputation, a total of 10,000 datasets were generated by applying 500 replications to 20 MI sets in accordance with Schomaker et al. [25], who recommended a bootstrap for each MI set. Residual bootstrapping was applied to the adjusted variables. The uncertainty of ICER was expressed as incremental cost-effect pairs (CE pairs) bootstrapped in cost-effectiveness planes (CE planes), and the percentage of pairs distributed in each quadrant was derived. Moreover, cost-effectiveness acceptability curves (CEACs), which indicate the probability of pharmacopuncture being cost-effective according to WTP, were derived. The threshold for willingness to pay was set at $26,374, based on the WTP amount survey result in Korea.

Three sensitivity analyses were conducted. First, a per-protocol analysis was performed only on patients who had received six or more treatments. Second, the method for calculating the productivity loss was altered. In the primary analysis, overall work impairment was calculated by including all patients and not just waged workers. Because it was assumed that only employed patients would have time and productivity loss in the sensitivity analysis, the loss of patients other than waged workers was processed as zero. Third, pharmacopuncture fees may vary from one hospital to another as they are not reimbursed. Accordingly, pharmacopuncture fees were assumed to be 1.5 and 2 times the pharmacopuncture cost in the primary analysis while performing the cost-effectiveness analysis. Finally, a cost-effectiveness analysis was performed on the Numerical rating scale(NRS) score for neck pain and Neck Disability Index (NDI). The VAS is a numeric pain scale that can be used for an objective assessment of subjective pain felt by patients. The patient selects the appropriate point along a 100 mm-long line with one end labeled 0, indicating no pain, and the other end labeled 10, indicating the worst pain possible [26]. The NDI is an assessment tool for disabilities caused by neck pain while performing daily activities. Ten items were rated on a 5-point scale for a total of 50 points, and the scores were converted to percentages, with higher scores indicating greater disability in daily living [27, 28]. The differences in NRS and NDI scores between the groups were assessed at baseline and 12 weeks after randomization, from which the differential NRS and differential NDI scores were calculated.

Between September 2019 and June 2020, 263 patients were screened, of which 101 were randomly allocated to the pharmacopuncture and usual care groups. One participant from each group withdrew their consent before receiving the treatment; moreover, one participant from the PT group was dismissed by the investigator due to an administrative error in the random allocation process. Ultimately, 98 patients (49 each in the pharmacopuncture and usual care groups each) who received at least one treatment session were included in the ITT analysis. The flow chart of the study is shown in Supplementary Figure A2 (see Additional file 1). The patients’ baseline characteristics are shown in Supplementary Table A3 (see Additional file 1). At baseline, although there were no differences in pain and function, a significant difference was found in EQ-5D. The EQ-5D-5 L score of the pharmacopuncture group was 0.69 ± 0.13, which was significantly lower than that of the control group. There were no significant differences in SF-12 when analyzed by MCS and PCS or in SF6D. Meanwhile, the pharmacopuncture and usual care list used in this study are shown in Supplementary Table A4 (see Additional file 1).

The pharmacopuncture group showed a significantly lower EQ-5D-5 L score at baseline; however, the score was similar to those of the control group after 12 weeks. There were no significant differences between the two groups’ SF6D scores. The QALY was calculated using regression analyses adjusted for baseline values (Table 1; Fig. 1).

The adjusted QALY of the pharmacopuncture group for 12 weeks was 0.168 (95% confidence interval [CI], 0.165–0.171), while that of the control group was 0.166 (95% CI, 0.163–0.169). Accordingly, the adjusted differential QALY calculated based on the EQ-5D was 0.002. The adjusted QALY calculated based on SF-6D was 0.159 (95% CI, 0.155–0.164) in the pharmacopuncture group and 0.156 (95% CI, 0.151–0.160) in the control group, with an adjusted differential QALY of 0.003.

From a healthcare perspective, the combination of medical costs and direct non-medical costs were similar between the pharmacopuncture and control groups during the treatment period, whereas the cost was lower by $26 (95% CI, -123 to 64) in the pharmacopuncture group during the entire period. Regarding direct non-medical costs, the time cost was significantly lower in the pharmacopuncture group ($201; 95% CI, 186–217) than in the control group ($236; 95% CI, 210–263). Specifically, the time cost was significantly lower in the pharmacopuncture group because they spent less time being treated than the control group (Supplementary Table A5 (see Additional file 1). Moreover, the cost of productivity loss was also lower in the pharmacopuncture group than in the control group by $1,130 (95% CI, -1738 to -536); thus, the cost from a societal perspective was lower in the pharmacopuncture group by $1,157 (95% CI, -1868 to -573; Table 2). Specific details are provided in Supplementary Table A5 (see Additional file 1).

From a societal perspective, pharmacopuncture was superior as it showed a slightly higher QALY and a lower cost of $1,157 compared to usual care. Based on $26,375 as the WTP among the Korean public, the probability of being cost-effective from a societal perspective was 100% based on both EQ-5D and SF-6D (Table 3; Fig. 2).

An analysis from a healthcare perspective also showed a cost difference of $26, confirming that pharmacopuncture was better than usual care. Based on $26,375 as the WTP, the probability of cost-effectiveness from a societal perspective was 83.7% and 89.6% based on EQ-5D and SF-6D, respectively.

The results of the sensitivity analysis are presented in Table 4. Concerning the cost-effectiveness plane in the analyses from each perspective, the probability of being distributed in the south quadrant, which is cost-saving and more effective, was 80.5% based on EQ-5D from a societal perspective.

In the main analysis, patients assigned to each group were followed-up up to 12 weeks, and missing values were imputed with multiple imputation. The costs from the healthcare system perspective include the costs of formal and informal healthcare involved in chronic neck pain treatment and the transportation and time costs. For the costs in the societal perspective, productivity costs from chronic neck pain are included.

Sensitivity analysis. 1 was a per-protocol analysis, which included patients who received at least six treatment sessions (47 in the pharmacopuncture group and 45 in the physical therapy group). Sensitivity analysis 2. Assuming that only employed patients suffered income loss due to time and productivity losses. Sensitivity analysis 3. We applied $26 by multiplying 1.5 with the pharmacopuncture cost in the base case analysis. Sensitivity analysis 4. We applied $35 by multiplying 2 with the pharmacopuncture cost in the base-case analysis. Sensitivity analysis 5. Cost effectiveness analysis based on NRS and NDI.

First, a cost-effectiveness analysis was performed on patients who had received at least six treatment sessions. The difference in costs was $1,149 from a societal perspective and $36 from a healthcare perspective. On the CE plane, CE pairs showed the highest distribution of 81.3% in the SE quadrant; thus, the probability of pharmacopuncture being more cost-effective from a societal perspective was 100% based on QALY.