How Effective Are Non-Operative Intra-Articular Treatments for Bone Marrow Lesions in Knee Osteoarthritis in Adults? A Systematic Review of Controlled Clinical Trials

Knee osteoarthritis (KOA) is a progressive joint disease and a leading source of chronic pain and disability. OA-bone marrow lesions (BMLs) are a recognised aetiopathological feature of KOA. Several intra-articular injectable therapies are recommended and used for management of KOA. This systematic review assessed the efficacy and safety of intra-articular therapies for improving OA-BMLs and reducing pain in adults with KOA. The study was conducted following registered review protocol (PROSPERO CRD42020189461) and six bibliographic databases, and two clinical trial registries were searched. We included eight randomised clinical trials involving 1294 participants, reported in 12 publications from 2016 to 2021. Two studies of sprifermin, one of autologous protein solution (APS) and one of high-dose TissueGene-C, reported a positive effect on OA-BMLs under 1-year follow-up. Two studies with corticosteroids reported mixed findings with no beneficial effect beyond 14 weeks of follow-up. One study assessing platelet-rich plasma found no significant improvement in OA-BMLs at 12 months follow-up. Knee pain was improved in two studies evaluating TissueGene-C and one study assessing APS; the remaining studies found no improvement in knee pain. Overall, we found mixed evidence on the efficacy of intra-articular therapy for improving OA-BMLs in KOA. Additional studies with long-term follow-up are needed to confirm the effect of various intra-articular therapies on OA-BMLs in KOA.


Introduction
Osteoarthritis (OA) is a chronic, progressive, and painful condition that affects many joints, including the knee. Knee OA (KOA) is the most common form of OA and is characterised by an inflammatory-degenerative process of all joint structures, which involves reduced physical activity and social and occupational functioning, mainly due to pain, the primary symptom [1]. It is the leading cause of disability in OA patients worldwide and the predominant condition leading to total knee replacement (TKR) surgery from a global perspective [2].
Despite the substantial global burden of KOA, no disease-modifying osteoarthritis drugs (DMOADS) are available to treat OA [3]. The reasons are multifactorial and include multiple heterogeneous causes of OA, which are difficult to target given the multiple pathways of causality [3]. Furthermore, dry biomarkers, such as traditional radiography,

Criteria for Study Inclusion
Studies were included if they compared intra-articular injectable non-operative interventions with placebo or other active treatments for KOA in an adult population with at least one OA-BML on MRI. Eligible study designs included randomised, quasi-randomised, or non-randomised controlled clinical trials. Sub-group analysis and post hoc analysis were also eligible. No restrictions were placed on the type of publication (full-text papers or conference abstracts), provided they had reported measuring the required outcomes.

Criteria for Study Exclusion
We excluded studies with no comparison group and those published in a language other than English.

Population, Intervention, Comparator and Outcomes
A detailed account of the Population, Intervention, Comparator, and Outcomes (PICO) is provided in the published protocol [20]. Briefly, the population constituted adult (≥18 years) humans with KOA; the intervention group for this review was intra-articular injectable non-operative therapy, and the comparator group was a placebo or any active intra-articular pharmacological intervention.
The primary outcomes of interest included structural changes (maximal area/volume) of OA-BMLs, determined at baseline and treatment intervals measured by quantitative or semi-quantitative measurements from MRI. The secondary outcomes of interest included changes in knee pain intensity determined at baseline and at treatment intervals measured by validated patient-reported outcome measures (PROMS) such as the Knee Injury and Osteoarthritis Outcome Score (KOOS); Western Ontario and McMaster University Arthritis Index (WOMAC); and changes in health-related quality-of-life (HRQoL) were assessed using the Assessment of Quality of Life-8 Dimension score (AQoL-8D) and the 36-Item Short Form Health Survey score (SF-36).

Search Strategy and Study Selection
Systematic and comprehensive searches were conducted from database inception to 16 May 2022 in the bibliographic databases Ovid MEDLINE ® , Embase, CENTRAL, CINAHL, SPORTDiscus and pEDro. Additionally, ClinicalTrials.gov and the Australian New Zealand Clinical Trials Registry (ANZCTR) were also searched. The search strategy was first optimised for MEDLINE and then adapted for other databases. The search strategy for MEDLINE is shown in Supplement Table S1. The complete search strategy, including MeSH terms, was developed and validated by the first author (A.K.) with assistance from a Medical Librarian. Bibliographic database searches were supplemented by hand-searching the reference lists of included articles and by contacting study authors.
Search results were imported into Covidence (Veritas Health Innovation, Melbourne, Australia), and two reviewers (A.K. and A.W.S.) screened the titles and abstracts of all articles against the eligibility criteria. Full-text copies of studies identified by the title/abstract screen as having met the inclusion criteria were obtained. Any disagreements or conflicting decisions were resolved through discussion and consensus with the other authors. Reasons for excluding studies were documented.

Data Extraction and Risk of Bias Assessment
Relevant data were extracted independently for each included study by two authors using a prespecified MS Excel-based data extraction template. The data extracted included publication details, study design, follow-up duration, population, intervention details, key outcomes, conclusion, etc.
Two reviewers (S.H. and A.S.) independently assessed the risk of bias in the included studies according to the Cochrane Handbook 5.0.1 RCT risk of bias assessment (RoB-I) tool [22]. As per the Cochrane RoB-I tool, the bias was assessed for items such as sequence generation, allocation concealment, blinding of participants, study personnel, outcome assessors, incomplete outcome data, selective outcome reporting, and other potential sources of bias.

Data Synthesis and Statistical Analysis
Due to the heterogeneity among included studies in terms of intervention, doses, duration of follow-up, and outcome reporting (OA-BMLs reported as a continuous outcome as well as categorical outcome), a meta-analysis was not deemed appropriate. Hence, we analysed the data qualitatively and presented it in the form of a narrative synthesis.

Results
Following database searches, hand-searching references, and trial registry screening, 1245 records were identified ( Figure 1). After removing duplicates, 740 records had titles and abstracts screened, and 97 full-text articles were screened for eligibility. After the full-text screening, eight studies reported in 12 publications met the eligibility criteria and were included in the review (Supplement Figure S1: PRISMA 2020 flow diagram).
Relevant data were extracted independently for each included study by two authors using a prespecified MS Excel-based data extraction template. The data extracted included publication details, study design, follow-up duration, population, intervention details, key outcomes, conclusion, etc.
Two reviewers (S.H. and A.S.) independently assessed the risk of bias in the included studies according to the Cochrane Handbook 5.0.1 RCT risk of bias assessment (RoB-I) tool [22]. As per the Cochrane RoB-I tool, the bias was assessed for items such as sequence generation, allocation concealment, blinding of participants, study personnel, outcome assessors, incomplete outcome data, selective outcome reporting, and other potential sources of bias.

Data Synthesis and Statistical Analysis
Due to the heterogeneity among included studies in terms of intervention, doses, duration of follow-up, and outcome reporting (OA-BMLs reported as a continuous outcome as well as categorical outcome), a meta-analysis was not deemed appropriate. Hence, we analysed the data qualitatively and presented it in the form of a narrative synthesis.

Results
Following database searches, hand-searching references, and trial registry screening, 1245 records were identified ( Figure 1). After removing duplicates, 740 records had titles and abstracts screened, and 97 full-text articles were screened for eligibility. After the fulltext screening, eight studies reported in 12 publications met the eligibility criteria and were included in the review (Supplement Figure S1: PRISMA 2020 flow diagram).

Risk of Bias
The overall risk of bias in included studies was low, with five trials assessed as having a low risk of bias according to the Cochrane RoB-1 tool. Three studies were considered at risk of bias due to one or more RoB-1 tool domains (High risk: random sequence generation, blinding; Unclear risk: random sequence generation, selective reporting allocation concealment) being at high or unclear risk of bias ( Figure 2).

Risk of Bias
The overall risk of bias in included studies was low, with five trials assessed as having a low risk of bias according to the Cochrane RoB-1 tool. Three studies were considered at risk of bias due to one or more RoB-1 tool domains (High risk: random sequence generation, blinding; Unclear risk: random sequence generation, selective reporting allocation concealment) being at high or unclear risk of bias ( Figure 2).

Primary Outcome-Effect of Intra-Articular Treatments on MRI-Assessed OA-BMLs
In all included studies that used MRI to assess the outcomes of OA-BMLs, four of those studies assessed change in OA-BML size using a modified Whole-Organ Magnetic Resonance Imaging Score (WORMS) system [7,17,25,29], while one study used MRI Osteoarthritis Knee Score (MOAKS) to assess BMLs [14]. The remaining studies did not specify using any specific type of MRI scoring tool (Tables 1 and 2). None of the studies assessed change in OA-BMLs as a primary outcome.

Primary Outcome-Effect of Intra-Articular Treatments on MRI-Assessed OA-BMLs
In all included studies that used MRI to assess the outcomes of OA-BMLs, four of those studies assessed change in OA-BML size using a modified Whole-Organ Magnetic Resonance Imaging Score (WORMS) system [7,17,25,29], while one study used MRI Osteoarthritis Knee Score (MOAKS) to assess BMLs [14]. The remaining studies did not specify using any specific type of MRI scoring tool (Tables 1 and 2). None of the studies assessed change in OA-BMLs as a primary outcome.

Sprifermin
Among the two studies that used intra-articular injectable sprifermin [7,25], Roemer et al. (2020) reported a positive effect of sprifermin in OA-BML on the patellofemoral joint (PFJ); however, there was no significant difference observed in OA-BML changes when accounted for the entire knee at 24 months of follow-up (p > 0.05) [7]. In an earlier study by Roemer et al. (2016), the OA-BMLs analysed for the whole knee showed significant (p = 0.042) improvement from 6 to 12 months but not from baseline to 6 months or 12 months (p = 0.237) [25].

Platelet-Rich Plasma and Autologous Protein Solution
Bennell et al. compared platelet-rich plasma against placebo in patients with symptomatic medial KOA (KL grade 2-3). The study found no significant (p = 0.31) benefit of platelet-rich plasma at 12 months follow-up for the reduction in OA-BML progression (24.3% vs. 18.9%) [15].
Kon et al. compared APS with placebo and reported a beneficial effect of APS on OA-BMLs in patients with KOA. At 12 months follow-up, the study found a significant (p = 0.041) reduction in progression in OA-BMLs grade favouring APS [14].
A significant improvement in WOMAC pain subscale scores (p = 0.02) was observed in the study evaluating APS, compared to placebo, over 12 months of follow-up [14]. Similarly, the study comparing TissueGene-C with placebo reported a significant improvement in pain in the TissueGene-C group compared to placebo at 52 weeks follow-up [27,29]. Another study evaluating high-dose TissueGene-C with low-dose TissueGene-C found a significant improvement in pain within both groups (p < 0.001); the between-group difference, however, was not significant (p > 0.05) [17,28].
Two studies evaluating corticosteroids reported no significant benefit compared to placebo in pain reduction at 14 and 26 weeks of follow-up [13], or at 24 months [26]. Similarly, studies evaluating PRP [15] and sprifermin [7,23,25] reported no benefit, compared to placebo, in pain reduction over a follow-up period of 12 and 24 months. However, Lohmander et al. reported statistically significantly lower improvement in pain compared to placebo at 12 months follow-up (p = 0.0013) (Table 3) [24,25].

Safety Outcomes
The overall safety profile of intra-articular therapies was acceptable, with no noticeable concerns (Table 5). No serious adverse events (SAEs) were reported with PRP, although the PRP group experienced more commonly encountered adverse events (AEs), such as knee joint pain, swelling, and stiffness after injections compared to the placebo [15]. Kon et al. demonstrated a favourable safety profile of APS at 12 months follow-up with no significant difference in the frequency and severity of AEs between groups with SAEs unrelated to the treatment [14].
Two studies evaluating sprifermin reported an acceptable safety profile with no treatment-related SAEs or AEs reported in both trials [7,23,25]. The typical local treatmentemergent AEs were arthralgia, joint swelling, and injection-site pain [7,23,25]. Similarly, TissueGene-C trialled in two studies showed no noticeable safety concerns. One study reported no significant difference in AEs [17,28], and the second reported joint inflammation, arthralgia, and effusion to be commonly experienced AEs in the TissueGene-C arm [27,29]. Both studies reported no SAEs related to treatment [17,[27][28][29]. Likewise, corticosteroids assessed in two studies were reported to have no noticeable safety concerns [13,26]. While one study reported no SAEs [13], another found no significant difference in SAEs in the two arms (p = 0.06) [26]. APS, n (%) 14 (45.2%) 2 (bladder cancer and kidney stone) Total number of AEs: 48 APS displayed a positive safety profile; no significant differences in frequency and severity of AEs between groups.

Cho 2016 * NCT02341378
High-dose TissueGene-C, n (%) Major AEs 10 (71) None High-dose treatment group had a higher incidence of AEs, which may be attributed to a dose-dependent increase in TGF-b.

Discussion
To the best of our knowledge, this is the first study to systematically review the efficacy and safety of intra-articular therapies for the treatment of KOA with a primary focus on structural changes assessed using OA-BMLs and symptomatic improvement assessed using knee pain.
This systematic review found mixed evidence from the included primary studies. High-quality evidence from the RCTs demonstrated improvement in the whole knee MRI-assessed OA-BMLs with high dose sprifermin at 6 to 12 months [25] and in PFJ OA-BMLs up to 12 months [7]. A significant reduction in OA-BML grade was seen with APS at 12 months of follow-up [14], and OA-BMLs improved in the high-dose cohort with TissueGene-C over the same follow-up period [17]. Similarly, a statistically significant reduction in OA-BML volume in the short term (14 weeks) was observed with intra-articular corticosteroids [13]. On the other hand, the beneficial effect of sprifermin on OA-BMLs was not significant when assessed from baseline to 12 months [25] and when accounted for the entire knee region at a longer follow-up (24 months) [7]. Another study assessing TissueGene-C found no differences in the progression of OA-BMLs when compared with placebo at 12 months [29]. The study that reported a positive effect of corticosteroids at 14 weeks found the difference in OA-BMLs levelled out at 26 weeks follow-up [13], whereas another study found no benefit of triamcinolone compared to placebo over 24 months [26]. The only study assessing PRP found no significant difference in OA-BML progression at 12 months of follow-up [15]. The improvement in knee pain outcome was reported in two studies evaluating TissueGene-C [17,29] and one study assessing APS. The remaining studies found no improvement in knee pain. The HRQoL outcomes assessed using the SF-36 and AQoL-8D in four studies found no significant improvement in scores with any of the intra-articular therapies compared with placebo [14,15,26,29].
Previous studies have found a discrepancy in the association between structural changes and pain in patients with OA [9]. However, OA-BMLs correlate with pain and changes in pain [30] in patients with KOA [9,11]. Furthermore, OA-BMLs are thought to drive OA-associated pain and may help predict treatment outcomes and prognosis [9,31]. Hence, researchers have argued that OA-BMLs could be the appropriate target for novel interventions that might reduce symptoms and improve the structural progression of KOA [9,31].
Two studies that evaluated sprifermin reported positive findings for improvements in OA-BMLs when considered for PFJ [7] or at a shorter follow-up [25]. Morphologically, positive findings in the PFJ can be attributed to its indirect effect on cartilage thickness [23] and the different loading patterns compared to the more load-bearing tibiofemoral joint (TFJ). Furthermore, less worsening of cartilage surface morphology in the PFJ may lead to an improvement in OA-BMLs at the PFJ [7]. However, an improvement in the whole knee region from 6 to 12 months and not from baseline to 12 months is not easily explained, and the exact causes are still to be understood [25]. The study evaluating APS showed the beneficial effect of APS, improving OA-BMLs with a significant improvement in knee pain, compared with placebo at 12 months [14]. However, the authors noted no improvement in cartilage and suggested future studies to confirm whether OA-BML improvements could be attributed to APS or if the observed improvements were the result of other unexplained factors [14].
TissueGene-C was evaluated in two studies that reported mixed findings. Cho et al.'s. study was constrained by a smaller sample size (n = 27), shorter follow-up duration, and no placebo control [17]. Guermazi et al. found no difference in OA-BMLs with TissueGene-C when compared to placebo, however an improvement in KOA structural features and other MRI markers such as Hoffa-synovitis and effusion-synovitis was observed [29]. Furthermore, the study found no improvement in meniscal damage or hypertrophic osteophyte formation [29]. Future studies of TissueGene-C in KOA should use a sufficient sample size and use a placebo-controlled study design.
The two studies evaluating corticosteroids uniformly demonstrated no benefit in OA-BML reduction in KOA [13,26] and found no association between OA-BMLs and knee pain [13]. Although Nielsen et al. found a significant positive effect at 14 weeks of follow-up, the effect levelled out at 26 weeks. The study concluded that there was no relation between corticosteroids and OA-BML volume [13]. This finding was consistent with the McAlindon et al. study that reported significantly greater cartilage volume loss and no improvements in OA-BMLs with triamcinolone compared with saline placebo. To be noted, earlier KOA trials have reported a strong placebo response to intra-articular injection, and a higher placebo effect has been a known phenomenon in OA studies [13,32]. Bennell et al. found no benefit of PRP compared to placebo for improvements in OA-BMLs, cartilage volume loss, or pain reduction and did not support the use of PRP for treating KOA [15]. The findings were inconsistent with earlier studies reporting the beneficial effects of PRP in pain reduction [33]. This inconsistency could be attributed to discrepancies in preparing PRP, injection regimens, outcome measures, and patient characteristics. Furthermore, the lack of blinding in the earlier trials may have influenced the positive outcome with PRP [15,33].
Additionally, although intra-articular therapies involving surgery were out of the scope of this paper, we note that studies using stem cell therapy, such as bone marrow aspirate concentrate (BMAC) or mesenchymal stem cells (MSCs), showed some promise by demonstrating regression of subchondral OA-BMLs, improvement in pain and subsequent reduction in progression to TKR [12,16]. However, further studies are warranted in this area as well. Additionally, given that poor knee alignment and resultant dynamic load are known to impact significantly on the natural history of KOA, this may contribute to the heterogenous OA-BML outcomes seen in our study [9,14]. Furthermore, variability among studies was observed in terms of MRI scoring tools used to assess OA-BMLs. While the majority of studies used tools such as WORMS and MOAKS, other studies, such as Nielsen et al., used computer-assisted segmentation (CAS) for OA-BML scoring, reflecting the heterogeneity of BML measurement [13]. Future studies should use validated tools (WORMS and MOAKS) to standardise the reporting of structural changes in KOA (OA-BMLs) to allow effective comparison across studies [34,35].
The safety of intra-articular therapies across the included studies was favourable, with no notable AEs or SAEs reported. The implication of these findings is of particular interest as it confirms the absence of any harmful effect of intra-articular therapies delivered in a clinical setting. Given that the intra-articular therapies in this review are safe in a controlled environment, we suggest further studies investigate their use in the clinic over a more extended period.
Notable strengths of this systematic review include a registered protocol-based method, exhaustive database and hand-searching, and a transparent risk of bias assessment. However, this study has certain limitations. Of note, we included only the RCTs that assessed intra-articular therapy delivered in a clinic setting through a non-operative procedure for treating KOA and reported OA-BMLs as an outcome. Considerable variability existed in reporting the OA-BML outcomes across the studies, limiting the possibility of a metaanalysis. Additional work could be done to assess the effect of intra-articular therapies on structural changes associated with pain in KOA. Future work should aim to incorporate outcomes such as osteophytes, effusion, synovitis, cartilage thickness, cartilage defect, and meniscal damage. Nonetheless, our study sets the priority for future systematic reviews to focus on structural outcomes and their association with symptom improvement in patients with KOA in this rapidly evolving research area.

Conclusions
This systematic review found mixed evidence on the efficacy of intra-articular therapies for improving OA-BMLs in patients with KOA. While high-dose sprifermin, TissueGene-C, and APS showed some promise for improving OA-BMLs, corticosteroids and PRP showed no improvement in OA-BMLs. In addition, no intra-articular therapy, except TissueGene-C and APS, showed any improvements in knee pain. Overall, although some