No Effects of Intra-Articular Injections of Sodium Hyaluronate, Corticosteroids, Platelet-Rich Plasma on Temporomandibular Joint Osteoarthritis: a Systematic Review and Network Meta-analysis of Randomized Controlled Trials


 Background

Intra-articular injections of corticosteroids (CCS), hyaluronic acid (HA), and platelet-rich plasma (PRP) have often been used for temporomandibular joint osteoarthritis (TMJ OA). However, there is no guideline for the choice of pharmacological injections. The aim of this network meta-analysis (NMA) is to compare the efficacy of different intra-articular injectable treatments on TMJ OA.
Methods

Studies were identified from PubMed, Embase and Cochrane Central with date up to December 2020. Randomized Controlled Trials (RCTs) included were the studies of patients with TMJ OA who had intra-articular treatment with CCS, HA, PRP, placebo and follow-up assessing TMJ function in target outcome variables. The primary outcome was temporomandibular joint pain (VAS). The secondary outcomes were maximal mouth opening (mm), and lateral movement to the affected side (mm).
Results

Nine RCTs involving 316 patients were included. For primary pain outcome, no significance was detected when CCS, HA and PRP were compared with placebo by both short- (3-6 months) and long-term (>12 months) follow-up. In addition, these injectables did not significantly outperform placebo by evaluating secondary functional outcomes (maximal mouth opening and lateral movement) with the same follow-up. Subgroup analyses showed that the effect of CCS on subgroups with more than 70% women was statistically less effective compared with placebo.
Conclusion

Evidence suggested that intra-articular pharmacological injections of CCS, HA, and PRP had no effect on improving TMJ pain and functional outcomes compared with placebo injection.
Trial registration

This study is registered with PROSPERO, number CRD42021270914.


Conclusion
Evidence suggested that intra-articular pharmacological injections of CCS, HA, and PRP had no effect on improving TMJ pain and functional outcomes compared with placebo injection.

Trial registration
This study is registered with PROSPERO, number CRD42021270914.

Background
The temporomandibular joint (TMJ), or jaw joint, is a synovial articulation that performs complicated everyday functions such as speech and mastication. (1) Temporomandibular disorders (TMDs), representing a range of conditions with multifactorial complex pathophysiology, are one of the most common causes of pain in the mouth and face and have the potential to produce chronic pain conditions.(2, 3) Temporomandibular joint dysfunction not only causes pain and disability in 20 to 25% of adults worldwide, but also huge economic costs estimated at $4 billion per year.(4) Temporomandibular joint osteoarthritis (TMJ OA), characterized by progressive cartilage degradation, remodeling of the condylar bone, and synovitis in addition to chronic pain, (5) is one of the TMDs.
Several treatment modalities of TMJ OA includes nonsurgical therapies such as physical therapies, occlusal splints, NSAIDs, and minimally invasive therapy of intra-articular injections with hyaluronic acid (HA), corticosteroids (CCS), or platelet-rich plasma (PRP). (1,3,6) Injectable pharmacological treatments, which are more invasive than other topical or oral therapies, are often reserved for patients who have not responded favorably to most or all other nonsurgical treatment modalities and have mild to moderate OA. (7) Previous systematic studies have suggested intra-articular injection as a rst-line therapy for TMJ OA, (8) particularly in pain control. (9) However, there is no guideline for the choice of injection pharmaceutics.
While a previous NMA on knee OA suggested the superiority of intra-articular treatments over placebo, (10) a more recent NMA showed that intra-articular saline injections performed as well as any other injectable option in improving pain and functional outcomes in hip OA. (7) However, the e cacy and effectiveness of CCS, HA and PRP compared with placebo injections, as well as the rank probability as the best treatment among these three pharmacological treatments remains unkown in TMJ OA, in spite of the previous meta-analyses on the relevant topics. (11)(12)(13) Besides, osteoarthritis is universally acknowledged as an age-related and female-dominant condition, (14,15) and our previous work showed that the e cacy of PRP injection treatment for OA was related to sample size and gender composition. (16) Consequently, it implicated the location-dependent and subgroup-dependent effects of intra-articular injectables in TMJ OA. With new RCTs comparing various IA injections on TMJ OA conducted since, it is of great importance to provide useful messages about pharmacological injection therapies and potential patient subgroup information for TMJ OA in the clinical context. Thus, it warrants the need for a systematic review and network meta-analysis to determine: whether pharmacological injections of CCS, HA and PRP could outperform injection of placebo, and if so, the possible treatment effect ranking of the three pharmacological injections of TMJ OA. Also, we believe it is necessary to conduct subgroup analyses on age and gender composition for further investigations.

Search strategy and protocols
We searched PubMed, Embase and Cochrane Central Register of Controlled Trials from inception to December 2020 (Additional le 1). Only Randomized Controlled Trials (RCTs) for humans were included, with no language restriction. Additionally, reference lists of reviews on the relevant topic were manually scanned. This study follows the guidance of Cochrane Library and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for conducting and reporting systematic reviews and network meta-analysis. (17) Eligibility criteria The PICO framework was employed as the critera for selection of RCTs: Population (P): Adult patients who have been diagnosed as TMJ OA on the criteria of Research Diagnostic Criteria/ temporomandibular disorders (RDC/TMDs), Diagnostic Criteria/ temporomandibular disorders (DC/TMDs) (Axis I Group IIIb), or imaging ndings that demonstrated mild to severe degenerative changes.
Intervention (I) and Comparison (C): Included RCTs should involve two or more intra-articular injections of the following injectables -corticosteroids (CCS), hyaluronic acid (HA), platelet-rich plasma (PRP), and placebo injections (Ringer's Lactated Solution). Small differences of dosage and injection intervals between RCTs were ignored. Included studies were not necessarily required to include placebo injection. Instead, studies that compare two or more injectables were also included.
Outcome (O): Included RCTs had to report the primary outcome pain (VAS scale), one or both of the secondary outcomes (maximal mouth opening and lateral movement to the affected side) with a follow-up of at least 3 months.

Study selection
Two blinded reviewers independently reviewed all articles at the 1) title/abstract and 2) full text stages. Any discrepancy in the title/abstract stage resulted in the article being forwarded directly to the full-text stage. At the full-text stage, any discrepancy was resolved by consensus.

Data extraction
Data from each study was extracted and reviewed for potential inconsistency by two extractors. Basic data included study characteristics, country(s) of recruited patients, age and gender, dropout(s), dose(s) and frequency of the intra-articular injection(s), whether prior arthrocentesis was performed, and other details of intervention(s). Quantitative data included pain as the primary outcome variable, and secondary outcome variables including maximal mouth opening and lateral movement indicating functional scores. Quantitative data were recorded accurately along with follow-up time of measurements.

Quality assessment
Two reviewers independently assessed the risk of bias under aspects of allocation concealment, random sequence generation, blinding of investigators and patients, blinding of outcome variables examiners, integrity of data, and selective reporting using the Cochrane risk-of-bias tool (18). Each aspect was rated as low, high or unclear risk of bias. Any disagreement was resolved by consensus.

Measures of treatment effect and summary measures
The Visual Analogue Scale (VAS, 0-10) for pain and the maximal mouth opening (millimeters, mm) were two commonly assessed outcomes in both research studies and clinics. Lateral movement (millimeters, mm) to the affected side was also included as secondary outcome with MMO. VAS that was originally recorded on a 0-100 scale was converted to the standard 0-10 scale, with 0 indicating no pain and 10 indicating the worst imaginable pain.
The changes in the pain scores, maximal mouth opening and lateral movement were calculated, and a negative value represented pain relief, while a positive value stands for improvement in fuctional outcomes. When studies did not report mean change, these values were calculated as the arithmetic difference between baseline and follow-up. The calculation for SD was available for each study. Measurements of outcome variables, which were weighted by sample size, were performed according to the Cochrane Handbook methods; weighted mean scores were calculated at baseline, short-term, and long-term follow-up timepoints. (18) Mean Differences between pharmacological injections and placebo injections were summarized using forest plots, along with 95% Con dence Intervals (95%CI) for pairwise meta-analysis, 95% Credible Intervals (95% CrI) for NMA, and SUCRA values.

Network meta-analysis
The network plot, which includes the number of each intervention, how often they were evaluated and the frequency of comparisons between treatment options, was rst generated to visually illustrate the basic information of the NMA. The nodes represented treatments whose size was weighted according to the number of studies that examined the treatment, while the lines represented studies that compared the two connected nodes, with the thickness of the lines indicating the number of studies.
Next, we made the assumption of transitivity that there were no systematic differences between the available comparisons other than the treatments being compared, and thus performed an NMA. (19) We used Bayesian random-effects models for mixed multiple treatment comparisons. Results were presented, along with 95% central credible intervals (CrI), as mean difference effect sizes for pain, maximal mouth opening, and lateral movement to the affected side.
Forest plots, leaguetables, and Ranking diagrams (Rankograms) were created. The surface under the cumulative ranking curve (SUCRA) scores, which indicates the probability that a speci c treatment ranks rst for a given outcome, were reported. A SUCRA value of 100% means that the treatment is certainly the most effective in the whole network, while a value of 0% means that it is certainly the least effective. A higher SUCRA percentage indicates that the given treatment was more likely to rank rst among all treatment options for a given outcome. A higher SUCRA indicates a better rank of the given intervention in the network. (19,20) Plus, subgroup analyses on age and gender, and sensitivity analyses on arthrocentesis were performed.
Inconsistency was assessed using the node-splitting method, while heterogeneity was assessed using ANOHE method. The statistical package used for the pairwise meta-analysis was 'meta', and those for network meta-analysis were 'rjags' and 'gemtc', all built under R V.4.0.2.

Results
The search strategy we employed retrieved 234 related studies (99 duplicates included). After reviewing the titles/abstracts, 99 were excluded. The remaining 36 articles were retrieved for full-text review, of which 9 met the inclusion criteria for meta-analysis ( Figure 1). (21)(22)(23)(24)(25)(26)(27)(28) The characteristics of the selected studies were given in Table 1. All studies included the primary outcome of pain, uniformly scored on the scale of VAS (scales of 0-100 were converted to 0-10).
For the secondary outcomes, 6 of the 8 studies included maximal mouth opening (mm), while 5 included lateral movement to the affected side (mm). All included trials were published between 2007 and 2019, involving a total of 316 participants. Of the 9 studies included, only 2 were conducted without arthrocentesis. (21,28) The sample size for each study was calculated as the number of patients, rather than the number of affected joints, for consistency.
Four studies examined short-term (3-6 months) outcome, while the other 5 only reported long-term (>12 months) outcome. Two studies had three comparative groups, while the other seven studies had two (placebo was counted as an independent group, as were HA, CCS, and PRP). The age of patients recruited in each study ranged from 32.5 to 51.7 years (one study did not report the mean age); the proportion of women in each study ranged from 30-90%.

Results of pairwise meta-analysis
All of the direct comparison evidenced at both short-and long-term was imported. The mean differences in all the comparisons with their corresponding 95% CIs were calculated using a random-effects model (Figure 2 For the secondary outcome, no signi cance was detected in any of the pairwise comparisons (Additional le 2).

Results of network meta-analysis
The network of all treatment comparisons analyzed by pain, maximal mouth opening, and lateral movement to the affected side during the longest follow-up timepoint of each study, was shown in Figure 3. As indicated by the thickness of the branches between nodes, comparisons between HA and placebo were numerically the highest, whereas those between CCS and PRP were the lowest. Overall, the individual treatment groups included intra-articular placebo injection with or without arthrocentesis (n=70), HA (n=136), (n=69), PRP (n=56).
The quality assessment of the included trials with risk of bias both within and between studies were done (Additional le 3). The heterogeneity assessment and inconsistency assessment of both long and short term timepoints were also completed (Additional le 4).
Primary outcome -pain: None of the TMJ intra-articular injections showed signi cant improvement in pain by calculation using VAS scores via comparing with placebo injections at either 3-6 months (4 trials, 193 patients), or >12 months (5 trials, 133 patients) ( Figure 4A). League tables showed no signi cant result for pairwise comparisons by means of network meta-analyses ( Figure 4B, C).

Secondary outcome -maximal mouth opening (MMO), lateral movement to the affected side
None of the intra-articular TMJ injections showed signi cant improvement in either MMO or lateral movement compared with placebo injections at either short or long term follow-up timepoints (Additional le 5).
Ranking probabilities NMA allowed ranking of treatments. The probability of each treatment taking a particular rank was calculated by means of rank probability plots (Rankograms) (Additional le 6). In comparison, SUCRA values, given in Figure 3, accounted for uncertainty in relative effect estimates and relative ranking, whereas the original ranking method overlooked this. However, it should be emphasized that ranking of the SUCRA value only provided a possible prediction, which means that the context of treatment effects should be considered when deciding on a speci c treatment.
For primary outcome pain, PRP had the largest probability to be the best injectable option in our network at both 3-6 months (SUCRA = 72%) and >12 months (SUCRA = 74%). Nevertheless, non-signi cance of the result produced in our NMA could not substantiate the SUCRA value as statistically valid ranking and clinical reference.

Subgroup (gender, age) and sensitivity analysis
Subgroup analyses which investigated the impact of age and gender of patients on primary outcome pain were conducted ( Table 2). A percentage of female over 70% was de ned as high proportion of female, while low proportion of female was de ned as the opposite. Though statistically insigni cant, an exactly reversed SUCRA value was shown, in which CCS works the best and placebo the poorest in studies with less than 70% of women, conversely in studies with more than 70% of women. The effect of corticosteroids on studies with more than 70% of women was the only pharmacological treatment statistically worse Sensitivity analysis was performed by limiting the treatments to those that were combined with arthrocentesis, namely from the aforementioned 7 out of 9 studies, and no signi cant difference was found. Disccusion Intra-articular injections of CCS, HA and PRP showed no signi cant short-or long-term improvement in pain and function outcomes compared with placebo injection in TMJ OA patients. Interestingly, we found the the effect of CCS on subgroup with more than 70% women was statistically less effective compared with placebo.
As opposed to the ndings that intra-articular pharmacological injections in knee OA are of signi cant treatment effects,(10) the NMA we conducted found the effects the three different types of injectables to be non-signi cant in all outcome variables, which shares similarities with the study of injections in hip OA. (7) Nonetheless, there are 3 possibilities that may shed light on reasons behind the non-signi cant effect of CCS, HA and PRP compared with placebo: Firstly, emerging evidence indicates that the apparent e cacy of pharmacological intra-articular injection therapy is largely due to other factors, particularly either the placebo effect or the combined effect of arthrocentesis.(29) On one hand, the intrinsic effect of Ringer's lactate solution functions as a washout of in ammatory cells, which are associated with pain signaling and osseous change, and release of the stuck disc phenomenon, (6,14) thereby alleviating the symptoms of TMJ OA. The function of Ringer's lactate solution potentially accounts for the positive treatment effects of the placebo in our NMA and the consequent lack of competence between pharmacological and placebo injections. On the other hand, the sensitivity analysis of arthrocentesis did not indicate any combined treatment effect of injectables with arthrocentesis. Nevertheless, it could be owing to the fact that 7 studies were done with arthrocentesis while 2 studies without, so that the effect of arthrocentesis is not strong enough to break the non-signi cance of the result.
Secondly, compared with the NMA of knee joint and hip joint OA, our study revealed a non-signi cant result similar with that of the hip joint OA, while Bannuru et al. found that all the injectables outperformed placebo in knee join OA, with HA being the most prabable to be effective in alleviating pain and improving joint function.(10) It seems that the location-dependent effect of intra-articuar injection is due to the anatomical difference: knee joint has much more synovial tissues than hip joint and TMJ joint. Synovium, a complex structure in knee joint but not in either hip joint of TMJ, could provide a potential explanation for the results. Speci cally, the anti-in ammatory injectable corticosteroids could suppress synovitis in OA joint. Plus, the synovium contains a number of targets for pharmacological treatments, including steroid hormone receptor positive cells and senescent cells. (30,31) Besides, HA is an intrinsic component of synovial uid, and PRP could be bioactive factors to promote the function of synovium tissues for pain relief and cartilage repair. (32) Thirdly, the subgroup analysis of our NMA indicated that the effect is subgroup-dependent, particularly de ned by age and gender. The diversity of patients probably leads to the non-signi cance of all of the three injectables investigated in our study. This reminds us of the value of Individual Patient Data (IPD) Meta-analysis that examine patients individually, as noted in an NMA that some individuals with knee OA respond more robustly to the injections of HA than others,(33) which could also be performed in TMJ OA in the future. Thus, it is not accurate to de ne pharmacological injection treatment as contrary to the goal of cost-effective health care.
Subgroup analyses on age and gender included in our NMA indicated that female and elderly patients respond poorly to CCS intra-articular injection treatment. Injectables particularly CCS should be given with caution in the clinics. The proposed mechanism explaining the different effect of CCS -not only injections targeting muscle and joint, but also oral, inhaled, topical use targeting other diseases -on subgroups of patients is absent in current literature, which predicted future trends to encourage more clinical trials with large sample size to substantiate the effect of CCS on different age groups, and gender difference. Additionally, as suggested in 2015 by Lochbühler et al, intra-articular injections could reduce mandibular growth in adolescent patients. (34,35) Taken all together, despite that there is more likelihood for pharmacological injections to bene t younger age groups for pain relief, the potential adverse effects could not be overlooked. Clinicians should be aware of the precisive choice of injectables based on the subgroup-dependent bene ts and side effects.
Our study is equipped with strengths and advantages among systematic review of related topics. Liu et al. did a meta-analysis by comparing multiple injection options for TMJ OA, but their evaluation was at short-term, mostly 1-month follow-up results.(12) They found that PDGF was more effective in relieving pain and improving MMO, while HA was more effective in enhancing MMO. PRP ranked as the best treatment in alleviating pain, but our study revealed that this recommendation lacked signi cance. We also demonstrated that the mean difference of HA in improving MMO was positive in the short term but turned negative in the long term, also insigni cantly. With the new RCTs published since, (27,28) results of the effectiveness of pharmacological injection were updated, which extended the work of Liu et al. Additionally, the ndings of subgroup analysis of age and gender in our NMA further implicated the targeted e cacy of injectable drugs, which previous meta-analyses omitted. There have been systematic reviews with qualitative descriptions on comparing the intraarticular injections for TMJ OA with a follow-up of more than 6 months,(13) but the qualitative opinions had the disadvantage of lacking statistical validation.
Pairwise comparisons in our study demonstrated that PRP outperformed HA when outcome pain was considered, conforming to the the ndings of Liapaki et al. (13) However, the network meta-analysis, with which we used the extracted data to the full, further con rmed that no pharmacological injection was found to be signi cantly superior to placebo with regards to pain.
Our NMA has several limitations. Firstly, the number and size of RCTs included were limited, which may introduce bias due to small study effects. Secondly, heterogeneity between studies could not be ignored in our NMA. Thirdly, language bias could be a concern, as we only included English language trials in our NMA. Fourthly, strati cation of drug (e.g. low and high molecular weight (LMW / HMW) HA) (36) was insu cient in the RCTs included, which may bring confounding factors to our NMA.

Conclusion
We found that the intra-articular injections for TMJ OA showed no signi cant improvement in all but exhibited subgroup dependent effects. Further large RCTs are needed to con rm our results and to identify the subgroups of patients who might bene t from the speci c injectables.

Consent for publication
Not applicable.

Availability of data and materials
All data generated and analyzed during this study are included in this published article.

Competing interests
The authors declare that they have no competing interests.

Funding
This work was supported by the National Natural Science Foundation of China (81630065). The foundation had no role in the planning, conduct or reporting of this work.
Authors' contributions YX and KZ contributed to conception and design. YX, KZ and GY contributed to acquisition, analysis, and interpretation, while MY and HO contributed to analysis and interpretation. YX, KZ, GY and XY drafted the manuscript, while MY and HO critically revised the manuscript. All authors read and approved the nal manuscript.