Pharmacopuncture in Korea: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Background. Pharmacopuncture is a new form of acupuncture combining acupuncture with herbal medicine, and it has been used under various conditions in Korea. The aim of this study is to establish clinical evidence for the safety and efficacy of pharmacopuncture in Korea. Methods. We searched 9 databases and two relevant journals up to December 2014 using keywords, such as pharmacopuncture. All randomized, controlled trials evaluating pharmacopuncture under any conditions in Korea were considered. Results. Twenty-nine studies involving 1,211 participants were included. A meta-analysis of two studies on obesity showed that 5 to 8 weeks of pharmacopuncture reduced weight, waist circumference, and body mass index (BMI) more than normal saline injections. In the 5 studies of musculoskeletal conditions, 7 to 30 days of pharmacopuncture had additional effects on the reduction of pain intensity, and this benefit was maintained by limiting analyses to studies with a low risk of bias for randomization and/or allocation concealment. Conclusions. This systematic review suggests the potential of pharmacopuncture for obesity and musculoskeletal diseases. However, it is difficult to recommend pharmacopuncture as an evidence-based treatment because of methodological flaws and small sample sizes of the included studies. Further well-designed trials are needed to draw a definitive conclusion.


Introduction
Pharmacopuncture (herbal acupuncture) is a new form of acupuncture treatment combining acupuncture and the injection of herbal medicine to the acupuncture points (acupoints). In Korea, pharmacopuncture was first officially introduced to the traditional Korean medicine (TKM) community in 1967 by Sang-Cheon Nam. While the conventional acupuncture treatment incorporates the physical stimulation of associated meridians and acupoints, pharmacopuncture adds chemical ingredients from therapeutic herbs with pharmacological effects [1]. In the treatment of pharmacopuncture, the typical constitution and conditions of the individual patient are diagnosed, and specific amounts of herbal extracts are injected into meridians and acupoints, providing the effect of both acupuncture and herbal medicine [1].
The effects of pharmacopuncture can be immediately observed after treatment because herbal extracts are directly absorbed without the need to pass through the gastrointestinal tract. Additionally, both patients with difficulty swallowing and those who refuse to take herbal medicine may profit by receiving pharmacopuncture treatment [1]. The major benefits of pharmacopuncture in comparison to conventional acupuncture are more rapid effects, additional synergistic effects of acupuncture and herbal extracts, and the ease for controlling dosage [2].
Since the 1970s, studies on pharmacopuncture conducted mostly in animals have reported that Astragali Radix [3,4],

Inclusion Criteria.
All RCTs evaluating pharmacopuncture treatment on various conditions were considered. Studies enrolling participants who reported any disorder or disease were eligible for inclusion. Hence, we classified each disorder or disease according to the International Statistical Classification of Diseases and Related Health Problems, 10th revision (ICD-10) [8] for the analyses.
Studies which assessed the combined effects of pharmacopuncture plus other interventions (e.g., pharmacopuncture plus acupuncture) were also considered when the identical intervention was administered to both the pharmacopuncture group and the control group.
For control groups, we considered placebo or sham, other interventions, and no interventions. Placebo or sham interventions were injections of normal saline or distilled water into the pharmacopuncture points or nonacupuncture points. Other interventions included acupuncture, herbal/western medicine, cupping, tuina, diet therapy, and physical therapy, including hot pack, transcutaneous electrical nerve stimulation (TENS), interferential current therapy (ICT), ultrasound, massage, and exercise.
There was no restriction on the type of outcome measures, but they had to be relevant to the conditions. All the trials were conducted in Korea.

Exclusion
Criteria. Nonrandomized trials, animal or cell studies, literature research, and quasi-RCTs (methods of allocating participants to a treatment group which are not truly random, e.g., hospital record number or alternation, and date of birth) were excluded. Trials including healthy participants were excluded.
We did not include trials testing bee-venom pharmacopuncture or injection of conventional medicine because they did not investigate the chemical effects of herbal medicine. Trials comparing different types of pharmacopuncture were excluded because the efficacy of control intervention could not be determined.

Data Extraction.
We reviewed all searched articles to evaluate their eligibility for inclusion. In case of uncertainties, authors were contacted for further information.
After the selection of studies, we extracted the following data from the selected articles: author, year of publication, study design, participants (age, gender), diseases or disorders, pharmacopuncture intervention, control intervention, outcome measures, main results, and adverse events ( Table 1). The outcome measures at the end of the treatments were used in data pooling. As for the pharmacopuncture interventions, we summarised each item in terms of the types and methods of pharmacopuncture, regimen, pharmacopuncture points, extraction methods, types of syringe, and amount, depth, and angle of the injection following STRICTA recommendations and the data was modified into the suitable form for trials of pharmacopuncture (Table 2).

Assessment of Risk of Bias (ROB).
We evaluated the ROB for the included studies according to the Cochrane Collaboration's ROB assessment tool [39].
We rated ROB for each item using "Yes (Y, low ROB)," "Unclear (U, uncertain or unknown ROB)," or "No (N, high ROB)." For patient blinding in studies with a placebo control, we assessed the study as having a low ROB when it clearly stated that patients were blinded. For the outcome assessor blinding, we judged that if it was clearly reported that the outcome assessor was blinded or the outcome measure was evaluated by blinded participants only, it was rated as having a low ROB. If the outcome measure was assessed by unblinded participants only, we rated it as having a high ROB. If the outcome measures were mixed with subjective and objective assessments and we could not obviously judge whether the outcome assessor was blinded or not, it was rated as having an unclear ROB. As for the incomplete outcome data reporting it was rated as having a low ROB if the number and reason for attrition were clearly reported in each group and were similar between groups and the percentage of withdrawals        Pharmacopuncture method was classified into three categories based on the levels of individualization: "fixed" means all patients receive the same treatment at all sessions, "partially individualized" means using a fixed set of points to be combined with a set of points to be used flexibly, and "individualized" means each patient receives a unique and evolving diagnosis and treatment [38].

Records identified through
Records excluded Full-text articles excluded,

Records after duplicates
Records screened based on the title and abstract   and drop-outs did not exceed 20% for short-term and 30% for long-term follow-up [39]. If there was disagreement, it was resolved by discussion with HL and JIK.

Data
Analyses. Meta-analysis was performed using the Review Manager software (version 5.2 for Windows; the Nordic Cochrane Centre, Copenhagen, Denmark). We used the mean difference (MD) and 95% confidence intervals (CI) to estimate the effect of an intervention for continuous outcomes using a random-effects model.
If it was impossible to perform statistical pooling, studies were assigned to 1 of 4 categories to classify the result for interpretation. The comparison between two groups was based on the results of original study: (1) positive when the pharmacopuncture group was significantly better than the control group, P; (2) negative when the control group was significantly better than the pharmacopuncture group, N; (3) neutral when there was no significant difference between the groups, NS; and (4) not assessable when the results were complicated or the presented data were insufficient, NA.
To address the heterogeneity among the included studies, the 2 test was used. An 2 value of 50% or more was considered to be an indicator of a substantial level of heterogeneity [40]. Sensitivity analyses were planned by including studies with low ROBs only or by including pain-related studies with sample sizes ≥40 per arm. We analyzed the trials with low ROBs for randomization and/or allocation concealment only and examined whether the estimate of the intervention effect was affected [41,42]. For the pain-related trials, studies with ≥40 participants per arm were separately analyzed to see whether any differences in the estimate emerged [43].

Results
Our search terms yielded 5,124 records: 49 from Ovid Medline or PubMed and 5,075 from domestic databases or relevant journals. After duplicated studies were removed, 3,030 records were screened. Based on the title and abstract, 2,929 records were excluded; 687 articles were inappropriate for the topic of this review; 2,105 were not clinical studies or were nonrandomized trials; and 137 trials did not satisfy the pharmacopuncture or control group criteria. Out of the remaining 101 studies, a total of 29 RCTs (Korean: = 27; English: = 2) were included in our review. Figure 1 shows a flow diagram of the literature searching as recommended in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [44]. Details of the included studies are summarised in Table 1. (29): data of 1,321 participants were included in the review. The number of participants in each group ranged from 10 to 37 in the pharmacopuncture group and from 9 to 46 in the control group. The median sample sizes per arm were 17 in the pharmacopuncture group and 18 in the control group.

Pharmacopuncture
Intervention. Details of pharmacopuncture interventions based on the revised STRICTA and modified to suitable patterns for pharmacopuncture are summarised in Table 2 [38].

Treatment Sessions.
The number of pharmacopuncture sessions ranged from 3 to 16 over 6 days to 8 weeks.  [26,28,30,34] used 12 meridian points, extra points, Ashi points, and tender points together. Two studies [24,33] used Ashi points only, and one study [35] used 12 meridian points plus Dr. Dong's acupuncture point. Three studies [9, 10, 16] mentioned approximate areas but not the accurate points, such as the abdomen [10], the area between the flexor carpi radialis tendon and the median nerve [16], or the left and right sides of four points inferior to and four points superior to the navel points on the stomach and the spleen and gallbladder meridians [9].

Extraction Methods.
As for the extraction methods of pharmacopuncture, 6 studies [11,13,15,21,27,30] used distillation of the herbal medicine; three studies [12,16,20] used alcohol immersion extraction methods; one study [29] used an extraction method that involved pressing from the herbs; one study [9] used distillation for wild ginseng and alcohol immersion for BUM; and the other 18 studies did not report details about the extraction method. Out of the 11 trials that mentioned extraction methods, 8 studies [9,11,13,15,16,21,27,29] followed the guidelines of the Korean Pharmacopuncture Institute. Two studies that used Carthami-Semen [12,20] did not follow the guidelines, in which the pressing extraction method is used to extract Carthami-Semen, but used an alcohol immersion extraction method instead. The other trial [30] used distillation to extract UDP, but there was no guideline for the extraction of UDP.

Amount of Injection.
Each amount of injection ranged from 0.05 mL to 1 mL, and the total amount of injection ranged from 0.2 mL to 5 mL. Only one study [30] did not report the amount of injection.

Depth of Injection.
Seven studies [16,18,20,27,29,30,36] reported the depth of injection. The depth of injection ranged from 5 to 30 mm according to the pharmacopuncture points. angles, and one study [16] used oblique angle (45 degrees) when injecting into the wrist area.

Outcome Measures.
Outcome measures reported in the included studies were very diverse because of the various types of focused diseases. Intensity of discomfort (e.g., measured with the visual analogue scale, the numeric rating scale) was investigated in 20 trials [13-16, 20-33, 35, 37]. All studies focusing on diseases of the musculoskeletal system and connective tissue adhered to it except one [34]. A qualityof-life-related scale was applied in six studies [12,15,19,20,29,30]. All trials on Bell's palsy utilized the Yanagihara score [13,14,17]. All studies on cervicalgia used a neck disability index [22-24, 28, 29]. Out of the five studies that treated low back pain, three studies were applied using the Oswestry disability index [25][26][27] (Table 1).

ROB Assessment.
The majority of the included trials were assessed as having a high ROB. Details of the ROB assessments are presented in Table 3.
The participant, practitioner, and outcome assessor each were blinded in only two trials [12,30]. Double-blinding of the participant and practitioner was conducted in two studies [18,35]. Participant blinding was performed in four trials [10,11,17,37]. The participant and outcome assessor were blinded in three trials [15,20,36] as outcome measures were all subjective and assessed by blinded participants in two trials [15,36], and the other one mentioned that an independent assessor evaluated constipation symptoms [20].
In terms of addressing incomplete outcome data, 13 studies [13-16, 19, 22-24, 27, 29, 31, 34, 36] were assessed as having a low ROB, as they had no missing outcome data. Nine trials [9,12,18,20,21,30,32,33,35] had missing outcome data, but the drop-out rate did not exceed 20% for shortterm and 30% for long-term follow-up, and the number and reasons for drop-out in each group were similar. Six trials [10,11,25,26,28,37] also had missing outcome data, but the drop-out rate exceeded 20% for short-term and 30% for longterm follow-up. The other study [17] had missing outcome data, but we could not calculate the drop-out rate, as the number of participants randomized in each group was not reported.
As for the selective outcome reporting, we could not locate and compare the protocols of any of the included studies. Therefore, we judged the ROB based on the described methods in each study. One study [16] had a high ROB because the authors (Lim et al.) were supposed to report visual analog scale (VAS) in the methods part, but there was no VAS data in results section.

Effects of Pharmacopuncture.
The key outcomes from the included studies are provided in Table 1.
A total of 10 studies were available for statistical pooling (Figures 2 and 3). As for the other 19 trials in which statistical pooling was impossible because of the substantial heterogeneity of the diseases, types of pharmacopuncture, control groups, or outcome measures, we classified the results into four categories: positive (P), negative (N), neutral (NS), and not assessable (NA).

Effects of Pharmacopuncture in Obesity.
Among the three studies on obesity, two studies [10,11] showed that Capsicum frutescens L. or Ephedra sinica Stapf-Aconitum carmichaeli Debx. pharmacopuncture significantly reduced weight, waist circumference, and BMI compared with the normal saline injection group by 1.36 kg, 4.59 cm, and 0.52 kg/m 2 , respectively, immediately after treatment (Figure 2 Another study [9] also reported that body weight, waist circumference, and BMI were more reduced than in the normal saline group, but we did not obtain sufficient data for statistical pooling. Thus, the result was not assessable.
As statistical pooling was impossible in the other seven trials, detailed results were described as follows. Two trials  [39]; "Y" indicates "yes (low risk of bias)"; "U," "unclear"; "N," "no (high risk of bias)." on HNP of the L-spine [31,33] showed unassessable effects of Soyeom and Ouhyul pharmacopuncture. Of the two studies on osteoarthritis of the knee, one study [30] contrasted root bark of UDP pharmacopuncture with normal saline injection, and the effect was not assessable. Another study [34] that compared Hominis Placenta Pharmacopuncture with acupuncture showed no significant difference between the groups. In shoulder pain caused by stroke [21,35], the effect of pharmacopuncture was not assessable. For acute ankle sprain [32], acupuncture improved the symptoms better than Hwangryunhaedok-tang pharmacopuncture after 9-12 days of treatments; in other words, they reported a negative effect of pharmacopuncture.

Effects of Pharmacopuncture on Diseases of the Nervous
System. The results of pharmacopuncture were composited for each disease. For headache, both Carthami-Semen and Hwangryunhaedok-tang pharmacopuncture improved symptoms compared with normal saline injection [12,15]. For Bell's palsy, the effect of Hominis Placenta Pharmacopuncture was significantly better than normal saline injection [17]. However, for the postauricular pain that accompanies Bell's palsy, the effect of Soyeom pharmacopuncture was not assessable. It may be due to the fact that the pain intensity or duration of pain decreased significantly, while the Yanagihara score did not show a significant difference between the groups [13,14]. One study on scolopendrid  pharmacopuncture treatment did not show additional effects on pain intensity in carpal tunnel syndrome [16].

Effects of Pharmacopuncture on Diseases of the Circulatory
System. The effect of Hominis Placenta Pharmacopuncture compared with distilled water injection was not assessable in leg spasticity of stroke patients due to the complexity of the results. For modified Ashworth scale (MAS), H-reflex/Mresponse ratio (H/M ratio), and Berge balance scale (BBS), there were no significant differences between groups, while the time up and go (TUG) in pharmacopuncture group was significantly lower than in distilled water injection group [18].

Effects of Pharmacopuncture on Diseases of the Digestive
System. One study of Hominis Placenta Pharmacopuncture had a similar effect to acupuncture in dyspepsia [19]. The effect of Carthami-Semen pharmacopuncture on chronic constipation was not assessable [20].
3.6.6. Effects of Pharmacopuncture on Diseases of the Genitourinary System. One study on Hominis Placenta Pharmacopuncture for dysmenorrhea showed a similar effect to normal saline injection [36].
3.6.7. Effects of Pharmacopuncture on Pregnancy, Childbirth, and the Puerperium. Hominis Placenta Pharmacopuncture had complicated results compared with normal saline injection [37]. There were no significant differences between groups in VAS for heating feeling, sweet during movement and sleeping, and complete blood cell (CBC) count. For thirst, Hominis Placenta Pharmacopuncture group showed significant higher VAS than normal saline injection group before the treatment ( = 0.023). After treatment, two   groups reported similar thirst symptom ( = 0.510) without correcting the baseline value. Therefore, we could not assess the results.

Adverse Events (AEs).
Only five studies reported AEs. In the study by Seo [21], they compared Ouhyul pharmacopuncture with normal saline and reported general pain in the Ouhyul group and transient local site pain or fatigue in the normal saline group. Kim [10] compared Capsicum frutescens L. with normal saline and reported moderate AEs related to anesthesia cream or pharmacopuncture (4 in the Capsicum frutescens L. group, 2 in the normal saline group). In each of the two studies by Park et al. [12,20], they compared Carthami-Semen with normal saline and reported mild AEs, such as pain during the injection, ecchymosis, and redness in the Carthami-Semen group and moderate pain during the injection in the normal saline group. One study by Kim et al. [30] compared UDP with normal saline and reported mild AEs, such as nausea and itching in the UDP group and slight dizziness in the normal saline group. These AEs disappeared in a short time without specific treatment, and no serious AEs were reported. Another two studies by Jun et al. [27] and Kim et al. [37] reported that AEs did not occur, and the other 22 trials [9, 11, 13-19, 22-26, 28, 29, 31-36] did not mention AEs.

Sensitivity Analyses.
We performed sensitivity analyses by excluding studies with predefined less desirable characteristics, and the results from the musculoskeletal studies were robust.
3.8.1. ROB. When the analyses were limited to two studies with a low ROB for random sequence generation and/or allocation concealment [27,29], pharmacopuncture had additional benefits in terms of pain relief in musculoskeletal diseases immediately after treatments (MD 1.76, 95% CI: 0.80-2.71, 2 = 0%). One study with adequate random sequence generation and/or allocation concealment did not provide enough information; thus, the effect of pharmacopuncture compared with normal saline could not be assessed in shoulder pain caused by stroke [21].

Sample
Size. There was no study with ≥40 participants per arm.

Discussion
Our review on pharmacopuncture aimed to establish the evidence of pharmacopuncture treatment of any disease. The analyses of two trials on obesity [10,11] demonstrated a significant benefit from 5 to 8 weeks of pharmacopuncture treatments compared with normal saline injection. The analyses of five trials on musculoskeletal diseases [22,[26][27][28][29] represented a significant effect from 7 to 30 days of combined treatment of pharmacopuncture with other interventions (e.g., acupuncture, herb medicine, tuina, and physical therapy) compared with other interventions only. However, these analyses were based on small studies and other interventions used in these trials were varied; thus cautious interpretation is needed. In the musculoskeletal diseases, pharmacopuncture's benefits were maintained by limiting the analyses to studies with a low ROB for randomization and/or allocation concealment [27,29], which means that they presented robust evidence for the treatment of musculoskeletal diseases. However, the number of participants in these studies was too small (less than 10 per arm), so careful interpretation is required. Pharmacopuncture does not appear to be associated with serious AEs, but the evidence is limited.
Most of the included studies had methodological weaknesses. Thirteen out of 25 studies [9, 11-13, 18, 20-22, 24, 27, 29, 30, 36] had low ROBs for adequate randomization and/or allocation concealment. Among them, only three studies [12,21,27] had both appropriate randomization and allocation concealment. It is well known that inadequate allocation concealment/random sequence generation leads to the overestimation of treatment effects [41,42], and unconcealed allocation is the most important source of bias in RCTs [45]. When we limited our analyses to the studies rated as having low ROBs for randomization/allocation concealment, pharmacopuncture's benefit was maintained.
There were some limitations in this review. Our review only included trials conducted in Korea and published in Korean or English. Therefore, we could not necessarily remove a potential language bias. Egger et al. [46] reported that studies published in non-English languages or studies published in journals that are not indexed in Medline are likely to increase the effect estimates, and this may have relevance to this review. In addition, pharmacopuncture, an acupuncture-related intervention, may be highly culture specific. According to the 2007 National Health Interview Survey (NHIS) data, only 6.5% of the Americans reported ever receiving acupuncture [47]. Upchurch et al. also reported that there was significant difference in the use of acupuncture by ethnicity and race. Asian women reported the highest usage [48]. Thus, further research is necessary to determine whether the interventions are applicable and acceptable in other countries.
The included trials were mostly of poor quality; thus, the reported data are likely to be overestimated. In addition, most of the included studies were small. Median sample sizes per arm were 17 in the pharmacopuncture group and 18 in the control group. The effect size of small studies may have been inflated due to poor methodological design and conduct [49]. Moore et al. [43] reported in a simulation study that at least 40 participants per arm are required to obtain clinically relevant results in trials of pain; however, there were no studies with ≥40 participants per arm, and we could not analyze it separately.
The efficacy of the treatment used for the control group, such as acupuncture, herbal medicine, and tuina manual therapy, was not yet established; therefore, we could not attribute "positive" results solely to the effectiveness of pharmacopuncture. Additionally, clinically meaningful information on follow-up results was sparse in the majority of the included trials. Therefore, the available evidence prevented us from further examining how long its benefit was maintained.
As we included all conditions/diseases, the focus of our review may seem blurred. However, this review provides an overview of the entire primary pharmacopuncture researches conducted in Korea. The results help to set priorities and directions for future research on pharmacopuncture.
Although this review represented the applicability of pharmacopuncture, the standardisation of pharmacopuncture intervention was not performed. Thus, in the future, it is absolutely necessary to standardise it to apply pharmacopuncture in routine clinical practice. The degree of pharmacopuncture stimulation could be influenced by the following factors: (1) pharmacopuncture types; (2) concentration and extraction methods of pharmacopuncture; (3) amount, depth, and angle of injection; (4) syringe types, including thickness and length; (5) pharmacopuncture points; and (6) number of sessions based on the STRICTA guidelines [38]. Currently, the Korean Pharmacopuncture Institute suggests guidelines for pharmacopuncture treatment. The classification of pharmacopuncture, such as meridian field, eightprinciple, or monoherbal medicine-type, is determined by the diagnosis of patient's conditions. The total amount of injections depends on the severity of the disorder, the age of the patient, the injecting area, and the characteristics and concentration of the pharmacopuncture extract. Using various types of injectors or syringes depends on the type of pharmacopuncture, its dosage, the area of the body part, and the depth of the injection. Syringe needles are generally between 26 and 32 gauges. Different needles are utilized for different uses.
However, the standardisation of these factors has yet to be completed, and there is no firmly established research method for pharmacopuncture studies; therefore, pharmacopuncture interventions of the included trials were very heterogeneous. In addition, future studies should include not only a test of the efficacy and safety of pharmacopuncture but also an examination of the validity of the intervention based on the standardised guidelines.

Conclusions
The results of this review demonstrate the effectiveness of pharmacopuncture for the treatment of obesity and musculoskeletal diseases compared with normal saline injections and other interventions, respectively; however, given the methodological flaws and small sample sizes, the available evidence is insufficient to recommend pharmacopuncture as an evidence-based treatment option. In the future, the standardisation of pharmacopuncture intervention and the adequate reporting of pharmacopuncture intervention in accordance with STRICTA guidelines are needed.