Effectiveness and Safety of Manufactured Chinese Herbal Formula for Knee Osteoarthritis: Insights from a Systematic Review

Objective. To assess the current clinical evidence of manufactured Chinese herbal formulae (MCHF) for knee osteoarthritis (KOA). Methods. Seven databases were searched from inception to May 2015. Eligible randomized controlled trials investigating the effectiveness of MCHF for KOA were included. Data extraction, methodological assessment, and meta-analyses were conducted according to the Cochrane standards. Results. A total of 17 kinds of MCHF were identified from the twenty-six included trials. Meta-analyses showed that MCHF significantly relieved the global pain of knee joints, either used alone or combined with routine treatments. Additionally, MCHF plus routine treatments significantly decreased the scores of WOMAC and Lequesne index. However, there were no statistical differences between MCHF group and routine treatment group in walk-related pain and WOMAC scores. No significant differences were found in Lysholm scores. There were twenty-one trials that mentioned adverse events. A pooled analysis showed that adverse events occurred more frequently in control group compared with MCHF group. Conclusions. Our results indicated that MCHF showed some potential benefits for KOA. However, we still cannot draw firm conclusions due to the poor methodological quality of included trials. More high-quality RCTs would help to confirm the evidence.


Introduction
Knee osteoarthritis (KOA) is a serious rheumatic disease in which the cartilage breaks down and causes the narrowing of joint space. In severe conditions, when knee joints completely lose cartilage, the periarticular bone and soft tissue structures would start to change and this may cause joint pain, swelling, misshapenness, and disability [1,2]. In America, symptomatic KOA occurs in approximately 10% of adults aged 60 years or older [3]. According to recent statistics, there were 9.3 million adults who suffered from KOA in the United States [4]. As a consequence of the aging process, the number of people with KOA is expected to increase in the next decades [5,6]. Osteoarthritis is predicted to be the fourth leading cause of disability by 2020 [7].
The aims of management for KOA are to relieve pain or swelling, improve mobility of the joint, and minimize disability. The current treatment options recommended in several evidence-based clinical guidelines include weight loss, exercises, oral pharmacological medications, topical therapies, intra-articular therapies, nonpharmacological therapies, and surgical treatments [8][9][10]. Among these various 2 Evidence-Based Complementary and Alternative Medicine interventions, nonsteroidal anti-inflammatory drugs and intra-articular hyaluronic acid or corticosteroids were more commonly used in clinical practice [11,12]. However, such treatments may be ineffective or lead to serious adverse events in some patients [13]. Under this circumstance, clinicians and patients were becoming more willing to adopt complementary and alternative medicine as a treatment option for KOA [14][15][16].
In mainland China and increasingly worldwide, Chinese herbal medicine has been used in varied forms for the treatment of KOA [17][18][19]. Manufactured Chinese herbal formulae (MCHF), also known as Chinese patent medicine, combined several Chinese herbs together to improve the therapeutic effects and reduce the side effects [20]. Hundreds of MCHF were approved by China Food and Drug Administration (CFDA) and listed in the "National Essential Drug List" (2012 Edition), providing basic medical support for patients in China. Although the mechanism has not been fully illuminated, animal experimental studies indicated that Chinese herbs decreased the levels of nitric oxide in the serum, synovium, and joint cartilage in OA rabbits [21]. Another research showed that Du-huo-ji-sheng decoction (a traditional Chinese herbal medicine) significantly exerted therapeutic effects on OA rabbits, probably through inhibiting the expressions of vascular endothelial growth factor and hypoxia inducible factor-1 [22].
Scientificity is a big obstacle for the acceptance of Chinese medicine in western countries. Previous meta-analyses have been conducted to evaluate oral or topical applications of herbal medicine in treating osteoarthritis [23,24]. So far, there has been no systematic review of randomized controlled trials of MCHF in treating KOA. Therefore, we performed a systematic review in an attempt to better define the efficacy and safety of MCHF in the treatment of KOA.

Eligibility Criteria.
We included all published and unpublished randomized controlled trials (RCTs) testing the effectiveness of MCHF in the treatment of KOA. Studies should involve human subjects and original data should be presented. The patients should be diagnosed by standard criteria of KOA, for instance, American College of Rheumatology (ACR) criteria in 1995 [25], Chinese Medical Association criteria in 2007 [26], or European League against Rheumatism (EULAR) criteria in 2010 [27].
The interventions in experimental group were MCHF, either used alone or combined with routine treatments, regardless of dose and duration. Manufactured Chinese herbal formulae included in the trials should be approved by China Food and Drug Administration (CFDA). The control group received no treatment, placebo, or routine treatments. Routine treatments were in accordance with the recommendations in clinical practice guideline, such as the Osteoarthritis Research Society International (OARSI) guidelines [28]. Cotherapies including herbal medicine, acupuncture, moxibustion, manual therapy, tai chi, and yoga were excluded.
One or more of primary or secondary outcomes should be reported in the included RCTs. The primary outcomes were the pain and function of knee joint measured by recognized scales, including Visual Analog Score (VAS), Western Ontario and MacMaster universities arthritis index (WOMAC), Lequesne functional index (Lequesne), Lysholm knee score scale (Lysholm), or analogous pain scales. In this review, we focused on the total score of these scales. The secondary outcomes were adverse events and quality of life (36-item short form health survey, SF-36).

Database and Search
Strategies. This review was guided by the PRISMA statement [55]. Studies were identified by electronic database search from inception until May 31, 2015, in PubMed, Cochrane Central Register of Controlled Trials (CENTRAL), Web of Science (SCI), https://clinicaltrials.gov/, China National Knowledge Infrastructure (CNKI), Wanfang Database (Wanfang), and Chinese Biomedical Literature Database (CBM). There were no restrictions on language or publication type. Titles and abstracts were scanned by two authors (Linghui Li and Hao Gong) independently to make an initial assessment based on the inclusion and exclusion criteria. Full-texts were assessed by two independent reviewers (Linghui Li and Hao Gong) when the information of titles and abstracts was not enough to judge the relevance. The reference list of retrieved papers was also searched. Disagreements were resolved by a third reviewer (Liguo Zhu). The strategy for searching PubMed was listed as follows: ("osteoarthritis, knee" [MeSH Terms] OR ("osteoarthritis" [All Fields] AND "knee" [All Fields]) OR "knee osteoarthritis" [All Fields] OR ("knee" [All Fields] AND "osteoarthritis" [All Fields])) AND "Chinese" [All Fields].

Data Extraction and Quality Assessment.
The following data were extracted: the first author names, year of publication, sample size, the characteristics of the patients (age and sex), diagnosis criteria, details of interventions in experimental and control groups, adverse events, and outcome measures for each study. Intention-to-treat data was extracted from the primary articles [56,57]. Where authors indicated a trial protocol registry number, the protocol of previous study was retrieved. Supplementary table provided the full names and components of MCHF. Data extraction was conducted by two authors (Shaofeng Yang and Shangquan Wang) independently according to predefined criteria.

Quality Assessment.
We adopted Cochrane Collaboration's tool for assessing risk of bias [58]. Six different domains, including a random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), and selective outcome reporting (reporting bias), and other potential sources of bias were used to evaluate methodological quality of included studies. Each study was judged to be low, high, or unclear risk of bias. (iii) Inappropriate outcome n = 32) ( the quality of the included studies. A third reviewer (Liguo Zhu) resolved discrepancies.

Data Synthesis.
We performed a meta-analysis using RevMan (version 5.3) [59]. For continuous data of our review, the difference in change from baseline to posttreatment between two groups was expressed as mean differences (MD) or standardized mean differences (SMD) with 95% confidence interval (CI). For dichotomous data, the effect of the intervention was expressed as relative risk (RR) with 95% CI. Random or fixed-effect model was used to calculate summary effects estimates [60] and heterogeneity was assessed using the I2 statistic which showed the proportion of variability between effect estimates [61]. One reason for high heterogeneity may be due to variability in the routine treatments. So we considered subgroup analysis to explore heterogeneity. Forest plots were created to depict estimates and confidence intervals of the interventions for each study. And plotting symbol size represented the weight of included studies in the meta-analysis [62]. Asymmetry and publication bias of the studies were evaluated by a funnel plot [63].

Description of Included Trials.
A total of 11277 records were identified through electronic databases: CNKI ( = 4356), CBM ( = 4565), Wanfang data ( = 1910), Medline ( = 255), CENTRAL ( = 113), SCI ( = 63), and https:// clinicaltrials.gov/ ( = 15). After removing duplicates, we screened 7092 records through titles and abstracts and selected 149 for full-text assessment. There were 6 nonrandomized controlled trials, 115 trials with inappropriate interventions or outcomes, and 2 trials with incomplete data. In the end, twenty-six trials (all conducted in China and published in Chinese) met the inclusion criteria ( Figure 1). Searches for trial registries yielded no ongoing trials. The characteristics of the 26 included studies are given in Table 1   Evidence-Based Complementary and Alternative Medicine not reported [35,48]. The patients ranged from 38 to 79 years old. Baseline demographic characteristics (age, sex, and disease course) were provided in all included studies except one [34] and all showed no significant differences between groups. All trials used MCHF alone or combined with routine treatments in the experimental group. Seventeen kinds of MCHF were involved in the current review (as shown in Table 2), six of which were recorded in China Pharmacopoeia (2015 Edition) with the indication of chronic joint pain rather than the standard name of KOA. All of the included MCHF were approved by China Food and Drug Administration (CFDA) and were commercially available. The interventions for control group were only routine treatments, including oral nonsteroidal anti-inflammatory drugs (NSAIDs: diclofenac sodium, aceclofenac, diacerein, ibuprofen, indometacin, ketoprofen, celecoxib, naproxen, and nabumetone), oral glucosamine, intra-articular injection (sodium hyaluronate and ozone), and nonpharmacotherapy (infrared therapy). The treatment duration varied from 15 days to 3 months.
Participants in four trials [31,39,40,46] were divided into three groups comparing MCHF used alone and combined with routine treatments to routine treatments. One trial [38] had four intervention arms (MCHF versus MCHF plus routine treatments versus diclofenac sodium versus ozone). One trial [47] had three intervention arms (MCHF versus glucosamine versus Shi's manipulation), so only the first two related arms comparing MCHF to routine treatments were included. The other twenty-one trials were all randomized controlled trials with two groups comparing MCHF intervention alone or MCHF plus routine treatments to routine treatments alone. Various outcomes were compared between different treatment groups (as shown in Table 1).

Methodological
Quality. The methodological quality in the included studies was assessed according to the criteria in the Cochrane Handbook for Systematic Review of interventions. The methodological quality of the majority studies was "poor." All of the included studies stated randomization; however only 13 trials mentioned the specific methods for sequence generation, such as drawing of lots [31], computer software [39], stratified randomization [46], and random number table. Allocation concealment was only mentioned in one trial [29]. One trial was of double-blind design [30] and one trial was not blinded [29]. The rest of trials did not report any information on blinding at all. Only eight trials reported information on drop-out. Three trials conducted follow-ups of 8 weeks [31,39] and 1 year [51]. Selective reporting was unclear due to the unavailability of the research proposal. None of the studies described sample size calculation. Risk of bias summary of included studies was shown in Table 3.

Effect on Joint
Pain. There were sixteen trials reporting pain VAS as primary outcome measure. However the reporting of measures was different in these studies, which could limit the utility of studies for meta-analysis. So we conducted meta-analyses as two parts: VAS-1 (global pain) and VAS-2 (pain on walking). VAS pain scale was generally considered as 10 cm lines from 0 (nil symptom) to 10 (worst possible symptom), but in several trials it is reported on a centimeter scale in the range from 0 to 100. We converted these outcomes to the 0 to 10 cm scale in meta-analyses. One reason for high heterogeneity may be due to variability in the routine treatments, so subgroup analyses were conducted to explore the heterogeneity. = 0.08) (as shown in Figure 2).
Eight trials (involving 870 participants) compared the effects of MCHF plus routine treatments to routine treatments. A random effects model was used for meta-analysis due to the significant heterogeneity ( 2 = 17.26, = 0.02; 2 = 59%). The results showed that MCHF plus routine treatments significantly reduced the mean VAS-1 score compared with routine treatments  Figure 3).

Visual Analog Score-2 (Pain on Walking).
A total of six trials with 7 comparisons (involving 657 participants) reported pain VAS-2 as outcome measure. Among the six trials, one trial [40] compared the effects of both MCHF used alone and combined with routine treatments to routine treatments (sodium hyaluronate injection). The results showed no significant differences between MCHF group and sodium hyaluronate injection group. However, MCHF plus sodium hyaluronate injection treatments significantly reduced the average pain score compared with both MCHF and sodium hyaluronate injection used alone. The other five trials [29,33,42,48,53] all compared the effects of MCHF used alone to routine treatments. A random effects model was used for meta-analysis    Heterogeneity: 2 = 0.37; 2 = 8.01, df = 3 (P = 0.05); I 2 = 63% Test for overall effect: Z = 2.85 (P = 0.004) Heterogeneity: 2 = 0.00; 2 = 0.00, df = = 0.98); I 2 = 0% Test for overall effect: Z = 6.57 (P  Figure 4).

Effect on Joint Function.
Physical function of knee joints was measured by WOMAC, Lequesne, Lysholm scores, and Hospital for Special Surgery (HSS) scales. Subgroup analyses were conducted to explore heterogeneity (based on various routine treatments).

WOMAC.
WOMAC was used as outcome measure in six trials [29,32,38,40,41,47]. However, there were two scoring methods used for the scale. These two scoring methods were both acceptable for clinical researches, so mean difference (MD) was replaced by standardized mean difference (SMD) for meta-analyses. Five trials [29,38,40,41,47] Figure 5).  Three trials [32,38,40] (involving 294 participants) contributed to the meta-analysis of comparison of MCHF plus routine treatments versus routine treatments (as shown in Figure 6). The results showed favorable effect of MCHF plus routine treatments compared to routine treatments (total: SMD:  [31,34,45] used Lequesne algofunctional index as outcome measure. One trial [31] compared the effects of MCHF both used alone and combined with routine treatments to routine treatments (oral NSAIDs therapy). The results showed no significant differences between MCHF group and oral NSAIDs therapy group.

Lequesne. Three trials
Three trials [31,34,45] (involving 218 participants) contributed to the meta-analysis of comparison of MCHF plus routine treatments versus routine treatments. A random effects model was used for meta-analysis ( 2 = 7.31, = 0.03; 2 = 73%). The results showed small beneficial effects of MCHF plus routine treatments compared to routine  [39,42,49,51,52] with six comparisons involving 604 participants used Lysholm as outcome measure. One trial [39] compared the effects of MCHF both used alone and combined with routine treatments to routine treatments. As shown in the result of meta-analyses, no significant differences were found in the comparisons of MCHF individually versus routine treatments (MD: 5. = 0.21) (as shown in Figure 9). Subgroup analysis revealed significant improvements on Lysholm scores in MCHF plus sodium hyaluronate injection group, compared with sodium hyaluronate injection used alone (MD: 9.53 [6.42, 12.64]; < 0.00001), but no significantly differences were found in other comparisons.

HSS.
One trial [50] involving 100 participants compared 6 weeks of intervention with MCHF (Bi qi capsule) against glucosamine (250 mg, tid). As reported, participants in both groups showed significant improvements in HSS score. However, between-group differences were not statistically significant suggesting that Bi qi capsule has comparable efficacy to glucosamine in the improvement of physical function of knee joints.

Effect on Quality of Life.
Two trials [29,34] involving 153 participants measured quality of life by SF-36. They assessed the general health concepts by the following eight subscales: physical functioning, role limitations due to physical health problems, bodily pain, general health perceptions, vitality, social functioning, role limitations due to emotional problems, and mental health. Data from one trial [29] showed no statistically significant differences in each subscale between MCHF group and routine treatment group ( > 0.05). However, the result for the comparison of MCHF plus routine treatments versus routine treatments revealed that the SF-36 scores were significantly increased in experimental group ( < 0.05).   3.6. Adverse Events. Twenty-one trials (involving 2836 patients) mentioned the occurrence or absence of adverse events and the other four trials [36,43,50,53] provided no data about adverse events. Six trials [34,37,42,45,51,52] reported that there were no adverse events in both experimental group and control group. The reported adverse events included gastrointestinal symptoms, abnormal liver function, dry mouth, local swelling, dizziness, and eruption. The majority of adverse events were not severe and they spontaneously recovered after drug discontinuance or symptomatic treatment. No serious adverse event was reported.

Publication Bias.
A funnel plot revealed an asymmetrical distribution of studies comparing MCHF (used alone or combined with routine treatments) to routine treatments for global pain of KOA ( Figure 11). These nineteen trials were all published in Chinese and most of them reported positive results, which indicated that there is publication bias.  MCHF (Yaotongning capsule and Tougu Xiaotong capsule) promoted the proliferation and glycosaminoglycan synthesis in IL-1 induced chondrocytes and may have the potential activity on treating chondrocytes degeneration caused by osteoarthritis [70,71]. A large number of clinical trials reported that MCHF could alleviate the symptoms of KOA [72]. It is important to perform objective reviews to assess the clinical evidence of MCHF for KOA. Several reviews provided evidence for the effectiveness of oral and topical use of herbal medicines for osteoarthritis. For example, one study investigated the effect of traditional Chinese herbal patches in the treatment of osteoarthritis [24]. There were 86 kinds of TCHPs identified in the review and the result showed certain evidence of TCHPs in improving the global effectiveness rate for OA. One study [23] assessed the effect of oral medicinal plant products used in various countries for osteoarthritis. There were 31 single medicinal plant products and several polyherbal preparations identified in 49 studies. However, meta-analyses were only possible for two kinds of herbal products due to the different study protocols and interventions. All other herbal medicines in this review were investigated in single trials, limiting conclusions. To date, the role of MCHF for KOA is still unknown. Therefore, the current review aims to investigate both of the therapeutic benefits and the safety concerns of MCHF.

Discussions
The current systematic review included 26 randomized controlled trials involving 3472 participants with KOA. All trials used MCHF alone or combined with routine treatments in the experimental group. The interventions for control group were routine treatments, such as oral drugs (NSAIDs and glucosamine), intra-articular injection (sodium hyaluronate and ozone), and nonpharmacotherapy (infrared therapy). A total of 17 kinds of MCHF were identified and six of them were recorded in China Pharmacopoeia (2015 Edition) with the indication of chronic joint pain rather than the standard name of KOA. It is reported in previous studies that the clinical efficacy of MCHF (Kang gu zeng sheng tablets) in KOA might be related to regulation of basic Fibroblast Growth Factor protein and mRNA expression in chondrocytes [73]. One study [74] claimed that MCHF (Bi qi capsule) could protect articular cartilage of patients with KOA by reducing the activity of MMP-3 in serum and synovial fluid, improving the activity of TIMP-1 and inhibiting the degradation of cartilage matrix.
Meta-analyses revealed that MCHF significantly relieved global pain of knee joint when compared with routine treatments (medication alone: = 0.0003). Since four of twenty-six trials provided no data about adverse events, we still cannot draw definitive conclusions about the safety of MCHF.

Limitations.
This study is the first systematic review of MCHF in treating KOA. We performed the review referring to the PRISMA through all stages to ensure the research quality. The results of meta-analysis showed that MCHF seemed to be an effective and safe intervention for global pain relief and function improvement of knee joint. However, there were several limitations existing in this review. Though we performed a thorough search strategy in seven electronic databases, we still cannot be absolutely certain that all relevant trials were found. All of the included trials were conducted in China and published in Chinese. Only KOA was referred to in our studies, so conclusions regarding osteoarthritis of other joints could not be drawn from our systematic review. Even though all the included trials mentioned randomization, half of them did not describe the specific methods of random sequence generation. Allocation concealment and blinding were only mentioned in one trial. The treatment duration in this review was short, varying from 15 days to 3 months. The majority of included trials did not report information on drop-out and follow-ups. Selective reporting was unclear in all trials due to the unavailability of the research proposal. None of the included trials reported a pretrial estimation of sample size. The biases of selection, detection, implementation, and publication should be taken into consideration when interpreting the conclusion in this review.

Conclusion
The current review indicates that MCHF have considerable therapeutic effect in the treatment of KOA with no serious side effects. Meta-analyses showed that MCHF seemed to be an effective method for global pain relief, either alone or in combination. In addition, MCHF combined with routine treatments appeared to be more effective in improving the physical function of knee joints compared with routine treatments. There was a slightly raised risk of adverse events in control group compared with MCHF group. However, we still cannot draw definitive conclusions due to the poor methodological quality of included studies. More highquality RCTs would help clarify the definite efficacy and safety of MCHF in the treatment of KOA.