Background

Osteoarthritis (OA) is an inflammatory joint disease that mainly damages articular cartilage. It is characterized by the degradation of articular cartilage and involving the entire joint tissue, which eventually leads to joint degeneration, fibrosis, fracture, defect and damage to the entire articular surface [1,2,3]. Epidemiological studies show that there are currently 355 million people with arthritis worldwide, and arthritis has become the world’s number one disabling disease. In China, as of 2015, there were 120 million people with arthritis in mainland China, with an incidence rate of about 13%, of which the number of people suffering from OA is the largest [4]. The management of patients with knee OA is mainly based on individualized and gradient treatment for the condition and severity of arthritis [5,6,7]. Among them, the drugs for OA pain are mainly non-steroidal anti-inflammatory drugs (NSAID) and specific (COX-2) cyclooxygenase II inhibitors [8]. However, these drugs may be expensive or cause stomach bleeding, and cannot repair cartilage and treat subchondral damage [8,9,10]. Therefore, due to the high incidence of NSAID-related adverse events, there is an urgent need for a safer and effective alternative to OA.

Boswellic acid is the active ingredient in Boswellia serrata; it has shown significant pharmacological activity in the treatment of inflammatory diseases such as rheumatoid arthritis, chronic bronchitis, asthma and chronic inflammatory bowel diseases (ulcerative colitis and Crohn’s disease) [11, 12]. Current research showed that 3-O-Acetyl-11-keto-beta-boswellic acid (AKBA) is the one boswellic acid with strong pharmacological activity; for example, AKBA has a powerful inhibitory effect on 5-lipoxygenase (5-LOX) [13, 14]. Clinical studies have shown that Boswellia serrata extract not only has anti-inflammatory and anti-arthritis properties, but also improves pain and physical function [15,16,17,18]; in vitro experiments also show that Boswellia serrata extract can inhibit the expression of inflammatory factors such as adhesion molecules [19,20,21,22,23]. With regard to the safety of Boswellia serrata, studies showed that Boswellia serrata extract (such as 5-Loxin and Aflapin) does not have toxic side effects at higher doses [23,24,25]. These indicate that the active compound of Boswellia extract (AKBA) is safe based on current evidence [23, 25].

As potential anti-inflammatory drugs for OA, the efficacy of Boswellia and Boswellia Extract have been reported in a lot of clinical trials [14,15,16,17, 24, 26]. Since 2003 [15], many trials have explored the feasibility of Boswellia and its extracts for the treatment of OA. In the meantime, the latest randomized controlled trials (RCTs) have assessed the benefits and adverse events of Boswellia and its extract for OA. To the best of our knowledge, this is the first systematic review and meta-analysis estimating the effectiveness and safety of Boswellia and its extracts for OA. Therefore, we hope that the results of this systematic review and meta-analysis will provide better evidence for the clinical application of Boswellia and its extract in the treatment of OA.

Why it is important to do the review

Boswellia and its extract have theoretical benefits for cultured cells in vitro and experimental animals [19, 21,22,23, 25, 27]; although several meta-analyses have widely analyzed the effect of herb or dietary supplements on OA [28, 29], to our knowledge, there is no systematic review and meta-analysis focused on assessing the efficacy of Boswellia and its extract on OA. For example, Liu et al. [28] conducted a systematic review about dietary supplements for patients with osteoarthritis; Of the studies it included, only three RCTs were about Boswellia and its extract. Meanwhile, its conclusions about Boswellia and its extract on OA failed to provide valuable reference information for clinical decision making. Another systematic review [29] only analyzed the RCTs before 2013. In 5 years, more RCTs about Boswellia and its extract on OA have been completed, which means that relevant conclusions need to be updated or confirmed. Therefore, this systematic review and meta-analysis focus on and summarizes available evidence from RCTs about the role of Boswellia and its extract in OA.

Materials and methods

Protocol

Systematic reviews and meta-analysis are carried out strictly in accordance with the protocol (CRD42018086785) and the PRISMA-guidelines (see Supplementary Materials) [30].

Search strategy and selection criteria

Web of Science, the Chinese Science and Technology Periodical Database (VIP), Wan Fang Database (Chinese Ministry of Science & Technology), EMBASE, Medline Complete, Chinese Biomedical Database (CBM), ClinicalTrials, the China National Knowledge Infrastructure Databases (CNKI), Pubmed, Cochrane Library (Until Issue 2, 2018) were searched with keywords “boswellic acid”, “Boswellia”, “Shallaki”, “Salai”, “aflapin”, “5-loxin” and “osteoarthritis”. The search period is from the establishment of the journal to January 2018.

Selection criteria

Participants

Human with specified diagnosis criteria of OA.

Intervention methods

Boswellia or its extract.

Comparison methods

Other treatments for OA, such as placebo or conventional medicine;

Outcomes

(1) Primary outcomes: visual analogue score (VAS), WOMAC pain, WOMAC stiffness, WOMAC function and lequesne index. (2) Secondary outcomes: WOMAC pain, WOMAC stiffness, WOMAC function at different points in time.

Study design

Randomized controlled trials (RCTs).

Information retrieval and data extraction method

Three researchers (Ganpeng Yu, Liuting Zeng and Kailin Yang) independently screened the literature based on protocol (CRD42018086785). The discrepancies between the three researchers would be resolved through discussion of all researchers. Literature retrieval is carried out according to the search strategy, the search strategy of Pubmed is shown in Table 1.

Table 1 Search Strategy for Pubmed
Full size table

After literature retrieval, three researchers (Ganpeng Yu, Liuting Zeng and Kailin Yang) extracted data from the included literature according to a pre-designed table (including author, year of publication, number of cases, inclusion and exclusion criteria of patients, duration of intervention, etc.). The discrepancies between the three researchers would be resolved through discussion of all researchers. The missing data in the literature would be obtained by contacting the author. If the author cannot be contacted, it would be estimated by the method provided by Cochrane Handbook 5.1.0 [31].

Literature quality evaluation methods

The Cochrane collaboration’s tool for assessing risk of bias provided by Cochrane Handbook 5.1.0 was used to evaluate the quality of the RCTs [32]. The evaluation content includes random sequence generation, allocation result concealment, blind method, incomplete result data, selective report and other biases. Four researchers (Ganpeng Yu, Wang Xiang, Tianqing Zhang, Liuting Zeng and Kailin Yang) independently assessed the quality of the literature. The discrepancies between the three researchers would be resolved through discussion of all researchers.

Statistical analysis

The RevMan5.3 software provided by the Cochrane collaboration network was used for system evaluation and meta-analysis. Enumeration data is represented by risk ratio (RR) and 95% confidence interval (CI), and measurement data is represented by mean deviation (MD) and CI. The heterogeneity between the results of the studies was tested by X2. If the heterogeneity of the study is low (P > 0.10, I2 < 50%), the fixed effect model is used for analysis; otherwise (P < 0.10, I2 > 50%), the source of heterogeneity is analyzed first and then subgroup analysis or random effect model is adopted.

Results

Results of the search

The total records identified through database searching and other sources were 513. Four hundred ninety-seven (497) were excluded based on the title and abstract and 16 for more detailed evaluation. After that, 7 studies were included [24, 26, 33,34,35,36,37], while 9 studies were excluded [15,16,17,18, 38,39,40,41,42] according to the inclusion criteria (Fig. 1). The 9 studies were Kimmatkar 2003 [15], Chopra 2004 [38], Shah 2010 [17], Chopra 2013 [39], Perera 2014 [16], Belcaro 2015 [40], Bolognesi 2016 [41], Gupta 2011 [18], Badria 2002 [42].

Fig. 1
figure 1

Flow diagram of searching and article selection

Full size image

Description of included trials

A total of 7 RCTs with 545 participants were included. Sengupta’s RCTs consist of 2 trial groups and 1 control group [24, 26]. According to the method of Cochrane Handbook 5.1.0, the control group was divided into two groups, matching the two trial groups (Sengupta 2008a and Sengupta 2008b; Sengupta 2010a and Sengupta 2010b) [31]. The characteristics of the RCTs were shown in Tables 2 and 3.

Table 2 The characteristics of the included studies
Full size table
Table 3 The characteristics of the included studies
Full size table

Risk of Bias in included studies

The summary and graph of risk of bias ware shown in Fig. 2.

Fig. 2
figure 2

The risk of bias

Full size image

Sequence generation

Among the 7 included RCTs, three studies [35,36,37] adopted unclear randomization procedures, we therefore rated it as having an unclear risk of bias. The other RCTs described their randomization procedures: Sengupta 2010 [24], Sengupta 2008 [26] and Vishal 2011 [33] utilized the computer-generated randomization scheme, while Haroyan 2018 [34] used the treatment randomization code to draw randomization. Thus, these RCTs were thought to have low risks of bias.

Allocation concealment

Sengupta 2010 [24], Sengupta 2008 [26], Vishal 2011 [33] and Haroyan 2018 [34] described that the appearance, smell and color of drugs preparations were similar and organoleptically indistinguishable, which is an acceptable method of allocation concealment. Hence, they were rated as having low risks of bias. Karimifar 2017 [35], Notarnicola 2016 [36] and Notarnicola 2011 [37] did not describe an acceptable method of allocation concealment; therefore, they were rated as having an unclear risk of bias.

Blinding

For blinding of participants and personnel, although all RCTs claim to use blinding, only Haroyan 2018 [34], Notarnicola 2016 [36] and Notarnicola 2011 [37] described the implementation process of blinding. Thus, we gave a low risk of bias for them. The other RCTs were rated as having an unclear risk of bias for they did not describe the blind implementation process.

For blinding of outcome assessment, the statisticians in Sengupta 2010 [24], Sengupta 2008 [26] and Vishal 2011 [33] wasn’t blinded, hence, they were rated as having a high risk of bias. Haroyan 2018 [34], Notarnicola 2016 [36] and Notarnicola 2011 [37] described the implementation process of blinding, hence, we gave a low risk of bias for them. Karimifar 2017 [35] were rated as having an unclear risk of bias for they did not describe the blind implementation process.

Incomplete outcome data

The missing outcome data of all RCTs balanced in numbers across intervention groups with similar reasons for missing data across groups. We gave them low risks of bias.

Selective reporting

One RCT (Vishal 2011 [33]) failed to provide all outcomes mentioned in its protocols, thus we thought its risk of bias was high. The other RCTs provided their protocols, and all of the study’s pre-specified outcomes that are of interest in the review had been reported in the pre-specified way; their risks of bias were low.

Other potential bias

In the RCTs of Sengupta 2010 [24], Sengupta 2008 [26] and Vishal 2011 [33], their protocols noted three primary outcome parameters, which means that when p = 0.05/3 = 0.017, the difference is significant. Strictly speaking, the statistics of this study were not carried out correctly. Hence, their risks of bias were high. Similarly, in Haroyan 2018 [34]‘s RCT, the difference is significant when p = 0.05 / 2 = 0.025. Its risk of bias was also high. Other sources of bias were not observed in other RCTs; therefore, the risks of other bias of them were low.

Primary outcomes

Visual analogue score

Six RCTs [24, 26, 33, 35,36,37] reported the changes of the visual analogue score (VAS) at the end of treatment. Due to the high heterogeneity (Tau2 = 10.85, I2 = 94%, P<0.00001), we used random effect model. In this index, it can be found that in improving VAS, Boswellia is better (WMD -8.33; 95% CI -11.19, − 5.46; P<0.00001). (Fig. 3).

Fig. 3
figure 3

The changes of VAS

Full size image

WOMAC

Four RCTs [24, 26-, 34–35] reported the changes of WOMAC pain. Due to the high heterogeneity (Tau2 = 94.69, I2 = 99%, P<0.00001), we used random effect model. According to the result, compared with the control group, Boswellia is better in improving WOMAC pain (WMD -14.22; 95% CI -22.34, − 6.09; P = 0. 0006). (Fig. 4).

Fig. 4
figure 4

The changes of WOMAC pain

Full size image

Four RCTs [24, 26, 33, 34] reported the changes of WOMAC stiffness. Due to the high heterogeneity (Tau2 = 44.40, I2 = 97%, P<0.00001), we used random effect model. According to the result, compared with the control group, Boswellia is better in improving WOMAC stiffness (WMD -10.04; 95% CI -15.86, − 4.22; P = 0. 0007). (Fig. 5).

Fig. 5
figure 5

The changes of WOMAC stiffness

Full size image

Four RCTs [24, 26, 33, 34] reported the changes of WOMAC function. Due to the high heterogeneity (Tau2 = 23.03, I2 = 93%, P<0.00001), we used random effect model. According to the result, compared with the control group, Boswellia is better in improving WOMAC function (WMD -10.75; 95% CI -15.06, − 6.43; P<0.00001). (Fig. 6).

Fig. 6
figure 6

The changes of WOMAC function

Full size image

Lequesne index

Six RCTs [24, 26, 33, 35,36,37] reported the changes of lequesne index. Due to the high heterogeneity (Tau2 = 0.55, I2 = 47%, P = 0.07), we used random effect model. According to the result, compared with the control group, Boswellia is better in improving Lequesne Index (WMD -2.27; 95% CI -3.08, − 1.45; P<0.00001). (Fig. 7).

Fig. 7
figure 7

The changes of Lequesne Index

Full size image

Secondary outcomes

Pain

Several RCTs reported Pain index (VAS and/or WOMAC pain) at week 4, 8, 12. The details of them were shown in Table 4 and Figure S1 ~ S6 (see Supplementary Materials).

Table 4 the pain index at week 4, 8, 12
Full size table

Stiffness

Several RCTs reported Stiffness index (WOMAC stiffness) at week 4, 8, 12. The details of them were shown in Table 5 and Figure S7 ~ S9 (see Supplementary Materials).

Table 5 the stiffness index at week 4, 8, 12
Full size table

Function

Several RCTs reported Function index (WOMAC function) at week 4, 8, 12. The details of them were shown in Table 6 and Figure S10 ~ S12 (see Supplementary Materials).

Table 6 the function index at week 4, 8, 12
Full size table

Adverse events

Five studies [24, 33,34,35, 37] reported AEs. Three of them were excluded because they reported no events in both arms. According to the results, there is also not strong evidence that which one is safer because there was no statistical difference (RR 0.63; 95% CI 0.22, 1.83; P = 0.39) (Fig. 8).

Fig. 8
figure 8

Adverse events

Full size image

Discussions

This systematic review and meta-analysis including 7 RCTs analyzes the effectiveness and safety of Boswellia and its extract for OA. Compared with the control group, Boswellia and its extract may relieve the pain (VAS and WOMAC pain) and stiffness (WOMAC stiffness), and improve the joint’s function (WOMAC function and lequesne index). In particular, since the doses of Boswellia and its extract reported in RCTs used for secondary outcomes analysis are 100–250 mg, based on the current evidence, pain, stiffness and joints’ function start to improve after 4 weeks of continuous Boswellia and its extract (at least 100–250 mg) intervention. While this finding seems promising, it should be interpreted with caution mainly due to the unclear risk of bias for selection bias (random sequence generation, allocation concealment) and attrition bias (incomplete outcome data), the high risk of bias for reporting bias (selective reporting), and the small number of participants.

Five studies [24, 33,34,35, 37] reported AEs. Three of them were excluded because they reported no events in both arms. According to the results, there is also not strong evidence that which one is safer because there was no statistical difference. However, safety studies conducted according to OECD guidelines and extensive acute and dose-dependent subchronic safety experiments on rats demonstrate that Boswellia and its extract does not exhibit toxic manifestations [23, 25], which means Boswellia and its extract might be a safety treatment option. But the absence of information on AEs does not mean that the intervention is safe [43]. Thus, although based on current evidences, we consider that Boswellia and its extract is a relatively safe treatment, we cannot assure it. Future clinical trials are required to report AEs with more explanations [44]. Thus, although based on current evidences, we consider that Boswellia and its extract is a relatively safe treatment, we cannot assure it. Future clinical trials are required to report AEs with more explanations [45].

Although there were several meta-analyses that have been widely analyzed the effect of herb or dietary supplements on OA [28, 29], as far as we know, this systematic review and meta-analysis is the first one that focused on evaluating the efficacy and safety of Boswellia and its extract for patients with OA. Compared with similar previous works [28, 29], we included more RCTs (including [33,34,35,36]) about Boswellia and its extract; thus, our conclusions were more accurate and reliable. In particular, compared with the work of Cameron et al. [29], we not only verified some outcomes (such as WOMAC) but also provide the recommended duration and dosage of Boswellia and its extract and so on. However, there are still some deficiencies in this study. For example, the risk of bias of some RCTs are high; the number of RCTs and cases included in this study is insufficient; the heterogeneity of some outcomes is high; these all affect the accuracy of the conclusions. The heterogeneity may be related to the discrepancies in the pharmacological effects of different Boswellia preparations. This may result from the different standardization of Boswellia and its extract manufacturing process, dosage, duration of treatment, units of laboratory tests and races of the selected patients or other places. Meanwhile, the recommended duration and dosage of treatment should also be interpreted with caution because the pain, stiffness and function index between week 1 and week 4 and the dosage of Boswellia and its extract outside 100–250 mg were not reported. Additionaly, since the quality of the RCTs is generally medium to low, all results should be interpreted more cautiously. Last but not least, further rigorously designed studies with high quality are needed to confirm the effectiveness and safety of Boswellia and its extract preparations for patients with OA.

Conclusion

This systematic review and meta-analysis showed that Boswellia and its extract may be a novel drug for patient with OA. Based on current evidence, the recommended duration and dosage of treatment with Boswellia and its extract is at least 100-250 mg 4 weeks. However, current RCTs have limitations, including the missing information of pain, stiffness and function index between week 1 and week 4 and small sample sizes. Meanwhile, since the quality of the RCTs is generally medium to low, we should formulate the conclusion more cautiously. More double-blind, large sample size RCTs of Boswellia and its extract for OA that reported pain, stiffness and function index before 4 weeks and the dosage of Boswellia and its extract outside 100–250 mg are needed in the future to confirm or modify the result of this work.