Skip to main content

29-11-2018 | Osteoarthritis | Article

Intra-articular treatment options for knee osteoarthritis

Nature Reviews Rheumatology

Authors: Ian A. Jones, Ryan Togashi, Melissa L. Wilson, Nathanael Heckmann, C. Thomas Vangsness Jr

Publisher: Nature Publishing Group UK


Intra-articular drug delivery has a number of advantages over systemic administration; however, for the past 20 years, intra-articular treatment options for the management of knee osteoarthritis (OA) have been limited to analgesics, glucocorticoids, hyaluronic acid (HA) and a small number of unproven alternative therapies. Although HA and glucocorticoids can provide clinically meaningful benefits to an appreciable number of patients, emerging evidence indicates that the apparent effectiveness of these treatments is largely a result of other factors, including the placebo effect. Biologic drugs that target inflammatory processes are used to manage rheumatoid arthritis, but have not translated well into use in OA. A lack of high-level evidence and methodological limitations hinder our understanding of so-called ‘stem’ cell therapies and, although the off-label administration of intra-articular cell therapies (such as platelet-rich plasma and bone marrow aspirate concentrate) is common, high-quality clinical data are needed before these treatments can be recommended. A number of promising intra-articular treatments are currently in clinical development in the United States, including small-molecule and biologic therapies, devices and gene therapies. Although the prospect of new, non-surgical treatments for OA is exciting, the benefits of new treatments must be carefully weighed against their costs and potential risks.


Osteoarthritis (OA) is a chronic and debilitating joint disease that causes damage to the articular cartilage and underlying bone 1. Although commonly referred to as a ‘wear and tear’ disease, complex interactions between genetic, metabolic, biochemical and biomechanical factors are also thought to be important in disease progression 2,3. Indeed, osteoarthritic chondrocytes are not apoptotic, but degenerated and deranged, as evidenced by ultrastructural changes and an uncoordinated gene expression pattern 4. Moreover, the whole joint is involved in the progression of the disease 5 and the roles of the synovium, muscles and ligaments are likely to be underestimated 6. Intra-articular drug delivery, in which a concentrated therapeutic dose is distributed throughout the joint capsule 7, might therefore be the ideal mode of drug delivery for OA therapies (Fig. 1).
In this Review, we briefly introduce intra-articular therapies before critically appraising the evidence supporting the use of standard intra-articular treatment options. We also discuss clinical studies that have investigated single-molecule biologic therapies and provide a high-level overview of cell-based therapies. Finally, we deliver an update on and a critical assessment of some of the most anticipated and promising intra-articular OA therapies that are currently in clinical development for the United States market. The surgical treatment of OA and the basic biology of the joint are not discussed in detail, as these topics have been covered extensively elsewhere 811, as have other clinically investigated treatments 12 and administration routes 13.

Intra-articular drug delivery

Intra-articular drug delivery has a number of advantages over systemic delivery, including increased local bioavailability, reduced systemic exposure, fewer adverse events and reduced cost 7,8. However, the efficacy of intra-articular therapies remains controversial and clinical guidelines regarding their use are often inconsistent with one another 14,15. In addition, factors such as drug residence time 16, systemic effects 17 and administration technique 18 contribute to treatment variability.
Intra-articular therapies are rapidly cleared from the synovial fluid by lymphatic drainage at a rate that largely depends on the size of the molecule. For example, the half-life of albumin in the joint is roughly 1–13 hours, whereas hyaluronic acid (HA) takes approximately 26 hours to clear the joint 16. Additionally, the half-life of NSAIDs and soluble steroids in the joint is only 1–4 hours 19. Despite the short residence time of intra-articular therapies, studies frequently report effects that last for several months 20. The mechanism behind these long-term effects is treatment-specific and is not well understood.

The placebo effect

When considering the evidence for or against an intra-articular therapy, it is important to understand that intra-articular injections elicit a strong placebo effect 21. Self-reported parameters such as pain and stiffness are particularly responsive to intra-articular placebo 22. In fact, the effect size of intra-articular placebo injections might be greater than that of both topical and oral placebos 23. Although intra-articular therapies are widely used in the treatment of OA, conflicting evidence exists as to whether standard intra-articular treatment options (HA and glucocorticoids) are beneficial compared with joint aspiration alone for many patients 24.
The strong placebo effect of intra-articular injections might account for the difficulty in detecting differences between treatment groups in clinical trials, especially when the difference between the groups is small. Even in patients who receive mock injections (an injection without therapeutic agent or placebo), the reported benefit is not statistically different to that of those who receive saline, lactic acid, procaine or hydrocortisone 25. A 2017 meta-analysis showed the effect of intra-articular saline on pain scores (visual analogue scale (VAS) and Western Ontario and McMaster Universities osteoarthritis index (WOMAC)), revealing statistically significant differences at 6 months for both the VAS and the WOMAC score 26.

Interpreting clinical findings

The degree to which a treatment affects the outcome of interest is called the effect size. Effect size is the most important statistic reported — not the P value — because it communicates how effective the treatment is at alleviating OA symptoms 27. Any given treatment might or might not have a large effect on the clinical population, irrespective of the statistical significance of the outcomes of the treatment. When an effect size is very large, it suggests possible clinical significance even in the absence of statistical significance, as the latter is heavily dependent on sample size. Even studies that meet standard power parameters (80% statistical power) still might not have sufficient power to detect a clinically significant difference (a type II error), which, by design, occurs 20% of the time. Likewise, a small effect size might be statistically significant but could have little to no clinically meaningful significance.
A tool that can be used to understand the clinical relevance of reported differences is the minimal important difference (MID), also known as the minimal clinically important difference (MCID). The MID was originally defined as the smallest difference in a score that patients perceive as being beneficial 28, and is most frequently calculated using the mean change method, which is the average score of patients who report feeling ‘slightly better’ minus the average score of patients who report feeling ‘about the same’ 29,30. In other words, the MID attempts to capture both the magnitude of the improvement and the value patients place on that improvement 31.
A 2017 systematic review revealed plausible MID values for the WOMAC, the knee injury and osteoarthritis outcome score (KOOS) and the European quality of life five-dimension questionnaire (EQ-5D) 32. These estimates can be used to provide pretext when clinically appraising a treatment’s effectiveness. However, the MID cannot be used to assess individuals 30 and does not take into account the overall risk–benefit ratio or the cost of a treatment 33,34. In addition, the MID should not be treated as a universal fixed value 35,36, as it tends to vary between populations and contexts, and by the calculation method used 33,37,38.

Standard intra-articular treatments

Although not categorically defined as standard-of-care, glucocorticoids and HA are standard intra-articular treatment options for the management of OA-related knee pain in patients who fail to respond to non-pharmacological therapy, NSAIDs or analgesics. However, an emerging body of evidence exists that calls the efficacy of standard intra-articular treatments into question 39, and an increasing number of professional organizations are questioning their appropriateness (Table  1).
Table 1
Hyaluronic acid and glucocorticoids to treat osteoarthritis a
American Academy of Orthopaedic Surgeons (2013) 41
Osteoarthritis Research Society International (2014) 48
National Institute for Health and Care Excellence (2014) 52
American College of Rheumatology (2012) 53
Hyaluronic acid
Not recommended
Not recommended
aSelected clinical guidelines published in the past 10 years.
Rather than putting forward specific treatment recommendations on the basis of existing evidence, the following sections explore several of the most widely cited and clinically used guidelines on the treatment of OA. The evidence supporting each recommendation is discussed to highlight potential strengths and weaknesses. Importantly, what is appropriate for use in the general population of patients with OA might differ for specific subsets of patients, and patients with some phenotypes of OA might respond to standard treatment options differently than others 40. Additionally, efficacy should be carefully weighed against both the cost of the treatment and the potential risk of harm, especially when the efficacy of a given treatment is questionable.


In their 2013 guidelines, the American Academy of Orthopaedic Surgeons (AAOS) found a lack of compelling evidence to support the use of glucocorticoids for the treatment of OA, as well as an unclear balance between the benefits and potential harms of this treatment 41. This recommendation was formed on the basis of the quality and generalizability of the included studies, which was determined in advance using an automated coding scheme. The final recommendation of the AAOS was made on the basis of six key studies, four of which were placebo-controlled trials that evaluated pain for at least 4 weeks 4245. The results of these studies were mixed, and all of these studies included one or more design flaws. Owing to the weak efficacy data of these studies 4245, as well as the results of other studies that suggested that glucocorticoid injections might be inferior to HA injections 46 or tidal lavage 47, the AAOS determined that inconclusive evidence existed to support the use of glucocorticoids for knee OA 41.
Guidelines published in 2014 by the Osteoarthritis Research Society International (OARSI) 48 issued a more favourable recommendation than the AAOS, concluding that intra-articular glucocorticoid injections were ‘appropriate’ and that the quality of the evidence was ‘good’. These conclusions were supported by the results of two systematic reviews 49,50 that were published in 2009 and 2006. However, a close inspection of the data in conjunction with current evidence suggests that the OARSI guidelines 48 might have overestimated the therapeutic efficacy of glucocorticoid injections. The authors of the 2009 systematic review 49 concluded that glucocorticoids provided benefits over HA at 2 weeks, but not at 4, 8, 12 or 26 weeks. By contrast, the authors of the 2006 systematic review 50, which was published in the Cochrane Database, took a more favourable stance concerning the effectiveness of glucocorticoids. However, an updated Cochrane meta-analysis published in 2015 showed that, although glucocorticoids seemed to offer small-to-moderate benefits over placebo for ≤6 weeks (standardized mean difference (SMD) -0.41; 95% CI -0.61 to -0.21), it was unclear whether this difference was clinically important 51. The authors also pointed out inconsistent and highly variable treatment effects, imprecise pooled estimates that did not rule out potentially relevant clinical effects, a high or unclear risk of bias, considerable heterogeneity between trials and evidence of small-study effects 51.
The guidelines published by the National Institute for Health and Care Excellence (NICE) 52 in 2014 and by the ACR 53 in 2012 both recommended glucocorticoids for patients with knee OA; however, both groups provided weak support for their recommendations. NICE recommended that glucocorticoid injections be considered as an adjunct to core treatments for the relief of moderate-to-severe pain in patients with OA on the basis of the ability of glucocorticoids to provide short-term (1–4 weeks) pain relief 52, but their recommendation relied on the out-of-date 2006 Cochrane Database review 50 that was also referenced in the OARSI guidelines 48, and failed to consider any literature published between 2006 and 2014. The ACR guidelines ‘strongly’ recommended glucocorticoid injections for patients who do not respond to full-dose paracetamol (acetaminophen), but provided no supporting references to justify this recommendation 53.
Since the aforementioned recommendations were published, concern has increased around the detrimental catabolic effects that glucocorticoids have on cartilage 54. For example, the results of a 2017 study suggested that administering glucocorticoid injections before total knee arthroplasty might increase the risk of postoperative infection 55. Notably, the results of one study showed that patients who had been treated with intra-articular glucocorticoids (40 mg triamcinolone) had significantly greater cartilage volume loss than those who had been given intra-articular saline over a 2 year period (between group difference −0.11 mm; 95% CI −0.20 to −0.03 mm) 56. This result is concerning because a 1% increase in the rate of tibial cartilage loss between baseline and 2 years corresponds to an ~20% increase in the risk of undergoing knee replacement surgery at 4 years 57, and because high rates of cartilage loss have been directly associated with an increased risk of undergoing arthroplasty 58.

Hyaluronic acid

Clinical recommendations for the use of intra-articular HA for knee OA tend to be less favourable than those for glucocorticoids. The 2013 AAOS guidelines strongly recommend against the use of HA for symptomatic knee OA 41. This recommendation was formed on the basis of a meta-analysis of 14 studies that showed an effect that was <0.5 MID units. The reliance of the AAOS guidelines 41 on a MID of 0.5 on a five-point scale has prompted debate, with some researchers arguing that a MID should only be used as a supplementary instrument and not as a basis for clinical decision making 59. However, the results of a 2017 systematic review 32 that aimed to establish credible, anchor-based MID values for patients with OA support the approach taken by the AAOS. Nevertheless, a lack of treatment consensus in the orthopaedic community continues to persist, which has led to an apparent disconnect between the AAOS recommendations and what occurs in clinical practice 60.
The OARSI guidelines 48 were more favourable towards HA than the AAOS guidelines 41 and suggested that the efficacy of HA for knee OA was ‘uncertain’. The recommendation of uncertain by OARSI was formed on the basis of three meta-analyses 20,49,61. The first of these meta-analyses showed that the long-term effects of intra-articular HA were superior to intra-articular glucocorticoids 49; however, HA was not compared with saline as a placebo in this study, which raises doubt over the validity of the conclusions. The second meta-analysis from 2011 compared HA to placebo and showed that the effect size favoured HA by week 4 (0.31; 95% CI 0.17–0.45), peaked at week 8 (0.46; 95% CI 0.28–0.65) and then trended downwards, with residual effects still present at week 24 (0.21; 95% CI 0.10–0.31) 20. By contrast, the results of the third meta-analysis from 2012 revealed that the benefits of HA for patients with symptomatic OA were minimal-to-non-existent and discouraged the use of HA owing to an increased risk of harm 61. The apparent disconnect between the conclusions of these two studies 20,61 is probably caused by the use of newer data, the exclusion of studies with fewer than 100 patients per treatment group and the inclusion of an additional five unpublished studies that showed that HA was not superior to placebo in the 2012 meta-analysis 61.
Similar to the AAOS guidelines 41 and the OARSI guidelines 48, the NICE guidelines 52 and the ACR guidelines 53 were both less favourable towards HA injections than they were towards glucocorticoid injections. NICE recommended against the use of HA for the management of knee OA 52 — a recommendation that relied on a 2006 Cochrane Database review 62 and an additional 20 studies that were published between 2006 and 2014. The NICE guidelines concluded that the evidence to show that HA was clinically effective was uncertain and determined that HA was unlikely to be cost effective. The ACR did not issue broad recommendations either for or against the use of HA in patients with knee OA 53.
Systematic reviews published after the AAOS guidelines and OARSI guidelines (2013 and 2014, respectively) have drawn inconsistent conclusions regarding the clinical utility of HA injections. A 2015 meta-analysis of double-blinded, sham-controlled trials that had at least 60 patients found an average treatment effect that was only 29% of the MID compared with placebo and no clinically important improvement in pain or other outcomes, even after subdividing HA preparations by molecular mass ( M r) 63. By contrast, a 2018 systematic review in which non-operative treatments for knee OA were compared showed ‘strong evidence’ for clinically important treatment effects when using intra-articular HA formulations with an M r of between 1,500 kDa and 6,000 kDa 64. Continued publication of contradictory recommendations, in conjunction with a lack of treatment consensus among clinicians, indicates that a need still exists for well-designed, pragmatic trials to evaluate the real word effectiveness of intra-articular HA for OA.

Intra-articular delivery of biologics

The idea that the progression of OA could result from an imbalance of catabolic and anabolic factors, as well as the known effectiveness of biologics in the treatment of inflammatory forms of arthritis, has raised hope that biologic agents might be used to treat OA 5. Over the past 10 years, several notable studies investigating the use of biologic agents for OA have been published (Table  2), but to date, results from clinical trials have mostly been disappointing. In the following sections, we discuss some of the most notable clinical studies to have used intra-articular biologic agents.
Table 2
Clinical trials using intra-articular biologic therapies to treat osteoarthritis
Study (year)
Study size (groups)
Final follow-up
Chevalier et al. (2009)
Anakinra (rhIL-1RA)
170 (101 anakinra, 69 placebo)
12 weeks
Low-dose anikara was inferior to placebo on pain score; high-dose anikara was superior to placebo on pain score
Ohtori et al. (2015)
Etanercept (anti-TNF)
39 (19 etanercept, 20 HA)
4 weeks
Etanercept was inferior to HA for VAS, WOMAC pain score, WOMAC stiffness score, WOMAC physical function score and total WOMAC score
Hunter et al. (2010)
Eptotermin alfa (rhBMP7)
33 (25 eptotermin alfa, 8 placebo)
24 weeks
Eptotermin alfa was superior to placebo for achieving WOMAC pain 20%, 50% and 70% reduction scores
Lohmander et al. (2014)
Sprifermin (rhFGF18)
192 (126 sprifermin, 42 placebo)
52 weeks
Sprifermin was superior to placebo for WOMAC pain scores and showed a dose-dependent response
BMP7, bone morphogenic protein 7; FGF18, fibroblast growth factor 18; HA, hyaluronic acid; IL-1RA, IL-1 receptor antagonist; rh, recombinant human; VAS, visual analogue scale; WOMAC, Western Ontario and McMaster Universities osteoarthritis index.

Targeting IL-1

IL-1β is a key mediator of the inflammatory and catabolic processes that lead to cartilage degradation and the destruction of joint tissues 65,66. IL-1β has been used to induce the dedifferentiation of chondrocytes in vitro 67, and the results of in vivo experiments suggest that IL-1β might directly mediate the erosive processes that lead to OA 6870. A 2009 randomized, multicentre, double-blind, placebo-controlled trial showed the IL-1β antagonist anakinra to be well tolerated, but not associated with improvements in OA symptoms compared with placebo 71. A 2012 study, in which the effects of anakinra following anterior cruciate ligament injury were examined, found that patients treated with anakinra had reduced pain and improved knee function compared with the control group 72.

Targeting TNF

TNF is a pro-inflammatory cytokine that interacts with chondrocytes 73 and is associated with a loss of knee cartilage 74,75. Infliximab was one of the first anti-TNF therapies to be clinically investigated as a potential intra-articular treatment for OA. Despite evidence of initial tolerability, the development of infliximab for OA never progressed beyond early exploratory trials 76. Another TNF inhibitor, etanercept, was subsequently evaluated for pain relief in patients with moderate-to-severe knee OA 77. In this study, 39 patients were treated with a single intra-articular injection of either HA or etanercept and followed for 4 weeks 77. For patients treated with etanercept, the VAS at week 1 and week 2 was improved compared with the VAS of those treated with HA, but these differences were diminished by week 4. Intra-articular administration of the TNF inhibitor adalimumab has also been investigated in patients with knee OA in an open-label randomized controlled trial (RCT) 78. The authors of this study reported safety and improvements in VAS 78; however, major methodological flaws, the most notable being the failure to register the trial and to pre-specify the outcomes, limit the utility of their findings.

Growth factor therapy

Only two clinical trials have been published to date that use intra-articular growth factor therapy to treat knee OA. The first study, published in 2010, was a randomized, double-blind, placebo-controlled, single-dose escalation study in which the safety, tolerability and dose-limiting toxicity of recombinant human bone morphogenetic protein 7 (rhBMP7) was evaluated 79. All 33 enrolled participants completed the 24-week study. No dose-limiting toxicity was found and the WOMAC scores suggested an improvement in pain, particularly in the mid-dose cohort. Although the findings of this study 79 generally supported continued development of rhBMP7, no further intra-articular clinical investigations have been initiated.
The second study to use an intra-articular growth factor therapy in patients with knee OA was a randomized, double-blind, placebo-controlled, proof-of-concept trial that aimed to evaluate the safety and efficacy of recombinant human fibroblast growth factor 18 (rhFGF18, also known as sprifermin) 80. Single-dose and multiple-dose regimens were trialled in 180 patients, 168 of whom were evaluated for the primary endpoint of changes in cartilage thickness at 6 months and 12 months. No significant difference was seen in serious adverse events or acute inflammatory reactions between the treatment and placebo groups, and patients treated with sprifermin had statistically significant dose-dependent improvements in several secondary outcomes 80. Post hoc analysis from this trial and the preliminary results of the ongoing phase II trial are discussed in detail in the intra-articular therapy pipeline section of this Review.

Cell therapies

Autologous point-of-care cell therapies

A lack of study comparability, as well as both methodological and intrinsic limitations, makes the efficacy of point-of-care cell therapies difficult to critically appraise. A comprehensive review of these treatments is beyond the scope of this Review; however, a basic understanding of point-of-care cell therapies is crucial given their widespread clinical use in the United States. In the following sections, we provide a general overview of the most commonly administered and/or studied point-of-care cell therapies used to treat knee OA (Fig.  2).

Platelet-rich plasma

Platelet-rich plasma (PRP) is autologous blood that has been centrifuged to concentrate the platelets to a level above that normally found in serum 81; however, the term PRP has been used to describe a range of different treatments 82. PRP contains a complex and diverse milieu of chemical mediators that interact with endogenous cells within the joint 83. This treatment was originally cleared by the FDA for use in enhancing the handling properties of bone graft materials 84 or, in the case of PRP gel, to ‘maintain moisture’ in a wound 85. Although intra-articular injections of PRP can be legally offered off-label in the United States to patients with OA in the clinic 86, they are not approved by the FDA for this indication and are not covered by most health insurance companies 8789.
Despite being the most studied point-of-care cell therapy for OA, a lack of well-powered trials and confusion resulting from both the biological complexity and lack of standardization between different PRP protocols 90 make it difficult to draw conclusions about the efficacy of PRP. However, several meta-analyses have attempted to broadly shed light on this topic 9193. A 2013 quantitative synthesis of data from RCTs showed that patients treated with sequential intra-articular PRP injections had significant improvements in WOMAC scores at 6 months compared with those who had been treated with injections of saline or HA (mean difference 18.0; 95% CI 8.3 to 28.8) 92. A 2016 systematic review suggested similar improvements in WOMAC scores up to 1 year post-intervention (mean difference −15.4; 95% CI −28.6 to −2.3) 93. The magnitude of these differences is notable 9496, but additional standardized, high-quality studies are needed before preliminary conclusions about efficacy can be drawn.

Bone marrow aspirate concentrate

The methods used to prepare bone marrow aspirate concentrate (BMAC) are similar to those used to prepare PRP. As such, BMAC and PRP share many of the same limitations. For example, the methods used to prepare BMAC are highly variable 97,98, which limits the comparability of treatments and studies. Intra-articular injections of BMAC for knee OA are not covered by health insurance in the United States. As with PRP, BMAC is a source of potentially beneficial anabolic and anti-inflammatory mediators 99; however, extracting bone marrow is more costly and invasive than extracting peripheral blood. Moreover, BMAC injections are frequently marketed as ‘stem cell’ or regenerative treatments 100 despite the facts that only 0.001–0.01% of the cellular content of BMAC is stromal cells 101 and that no clear regenerative benefits have been demonstrated to date 97.

Adipose tissue injections

Autologous adipose tissue injections (also known as fat grafts) are obtained and processed at the point-of-care from lipoaspirate by mechanical means without the use of enzymatic digestion 102. In accordance with the 2015 FDA guidelines for human cells, tissues, cellular products and tissue-based products 103, proponents of these treatments claim that mechanical processing of adipose tissue does not alter the original relevant characteristics of the tissue relating to its utility for reconstruction, repair or replacement, and that the injections merely provide cushioning and support. These 2017 claims 104 amount to careful restatements of the 2015 FDA’s criteria for ‘minimal manipulation’ and ‘homologous use’, which must be met for a product to be regulated solely under Section 361 of the Public Health Service Act 105. Despite being widely available throughout the United States, few human studies have been conducted to date that have demonstrated the effectiveness of these ‘minimally manipulated’ treatments 98.

Stromal vascular fraction

Stromal vascular fraction (SVF) is treated differently than the aforementioned mechanically derived point-of-care cell therapies. The FDA has stated that they consider SVF to be beholden to the provisions for investigational new drugs (INDs) because enzymatic digestion is needed to dissociate and isolate the stromal elements from the surrounding connective tissue 104. The final cell product, which is prepared by centrifugation of the enzymatically digested tissue, is a distinct component of lipoaspirate that contains a population of progenitor cells 106. However, SVF is highly heterogeneous 107 and only ~15–30% of the cellular content is stromal cells 108. Moreover, although adipose-derived stromal cells can be purified from SVF 106, SVF should not be confused with adipose-derived stromal cells 102,108. Few clinical studies have been performed to investigate the use of SVF to treat cartilage pathology 109,110, but SVF is not commonly used as an intra-articular agent 111 and the few studies that have been published have only provided preliminary safety data 112,113.

‘Stem’ or stromal cell therapies

Mesenchymal ‘stem’ (or stromal) cell (MSC)-based therapies are not a homogeneous class of cellular treatments 114. Because clinical trials to date have been limited to low-powered safety studies (that often did not include a control), the authors of systematic reviews that have attempted to critically appraise the use of intra-articular MSCs have failed to reach clinically applicable conclusions 115118. Nevertheless, given the increase in orthopaedic clinics marketing ‘stem cell’ treatments in the United States 100, these therapies cannot be ignored. A comprehensive overview of studies investigating MSC injections for knee OA will not be provided here, as numerous reviews of this topic exist 116119. However, certain aspects of the field are often overlooked and require clarification (Box  1).

Stem cells or stromal cells?

The term MSC was first coined in the early 1990s and was taken to mean mesenchymal stem cell 120. Since that time a number of different definitions have been associated with the term 121. The most pervasive and widely accepted definition (mesenchymal stromal cell) was established in 2006 by the International Society for Cellular Therapy (ISCT) 122 1 year after they released a position statement that attempted to retain the ‘MSC’ abbreviation while separating MSCs from the ‘stem cell’ label 123. According to the ISCT criteria, MSCs must be plastic-adherent, express or lack specific cell surface markers and be capable of trilineage differentiation in vitro into osteoblasts, adipocytes and chondrocytes 122.
The term mesenchymal ‘stem’ cell conveys assumptions that were not included in the original concept 120 or supported by direct experimental evidence 124. However, with respect to their clinical use as an intra-articular therapy for OA, MSCs are plastic-adherent (or prospectively isolated) populations of stromal cells that can be obtained from any tissue, and that express specific cell surface markers and are capable of trilineage differentiation in vitro 108,122. This definition clearly distinguishes MSC treatments from the various cell concentrates discussed previously without overtly conflicting with the original ISCT criteria (Fig.  2).

Mechanism of action

Although the in vivo mechanisms of therapeutically used MSCs are still unclear, the release of chemical mediators is thought to be important 125. The conditions currently treated with MSCs fall into two broad categories: immune or inflammatory conditions and tissue repair or regeneration 126. However, it is important to stress that besides in a few well-established indications, the assertion that MSCs have an intrinsic capacity to sense and address whatever is needed for the repair and regeneration of cartilaginous tissue in the joint is not based on scientific evidence 127.

Are MSC therapies safe to use?

The safety profile of MSC therapies depends on the type of cell that is used, as well as the methods that are used to isolate and process it. For example, autologous cells administered at point-of-care are likely to have a lower risk of tumorigenicity than cells that have been expanded in vitro 128, as culture conditions can modulate the mechanisms by which therapeutic cell products operate in vivo 67,126. Although no major adverse events have been reported that were attributable to autologous or allogeneic MSCs administered via intra-articular injection 115,117, malignant transformation remains a potential risk for any cell therapy. According to the results of a 2017 systematic review, over one third of studies failed to clearly describe their method of assessing safety 129. Additionally, the majority of studies have focused on the use of autologous cells; only two studies have been published on intra-articular allogeneic MSC treatments for knee OA or knee cartilage repair 130,131. Moreover, even in the absence of serious adverse events, the use of expensive, unproven cell therapies could delay or hinder a patient’s access to well-established surgical treatment options 105.

Clinical use of MSCs and progenitor cells

With a few exceptions 115,116, literature reviews have tended to take a positive view of MSCs as a promising potential treatment alternative for OA and cartilage repair 117119,132,133. However, an apparent disconnect exists between the results of in vitro studies, preclinical studies and human studies 134, and the highly heterogeneous nature and poor quality of studies published to date precludes quantitative synthesis 118. Moreover, among the RCTs that have been published, intra-group improvements and/or inter-group sub-score differences are often highlighted, whereas intra-group comparisons fail to show improvement or are left unreported 135. Overall, the limited availability of strong clinical data suggests that the generally positive efficacy conclusions concerning MSC therapy for knee cartilage pathology are premature and might be overstated. Nevertheless, broad efficacy conclusions are of little practical utility given the complexity and intrinsic lack of comparability between different MSC treatments. Just as each biologic therapy must be evaluated on a case-by-case basis, so too must each MSC therapy.

Box 1 Controversies surrounding mesenchymal stem (or stromal) cell therapies

The term MSC can be used to describe two different types of cells: mesenchymal stem cells and mesenchymal stromal cells. Stem cells are multipotent and can self-renew in vivo 195. By contrast, stromal cells are loosely defined as plastic-adherent cells that express and/or lack specific cell surface markers and are capable of trilineage differentiation into osteoblasts, adipocytes and chondrocytes in vitro 108,122.
The terms marrow stromal cell, multipotent stromal cell, adipose-derived stromal cell and medicinal signalling cell have all been used to describe passaged, plastic-adherent adult multipotent mesenchymal cells 121. This inconsistent nomenclature reflects the assumed, and not yet fully understood, immunomodulatory and/or immunosuppressive properties that are associated with stromal cell populations 126, as well as the confusion surrounding the dynamic interactions between cellular niches that help to determine cell fate 196.
Regulation for use as a therapy
With respect to their use as an intra-articular therapy in clinical research, true MSCs are classed as drugs that require federal regulatory approval before they can be administered to patients 104. Although some treatments containing multipotent stromal cell populations might claim to be exempt from FDA regulation (such as so-called minimally manipulated cell therapies and certain blood-derived products), these treatments should not be confused with true MSC therapies, as they tend to contain more heterogeneous cell populations than those in true MSC therapies and have effects that are not primarily attributed to their medicinal immunomodulatory and/or immunosuppressive capacity or differentiation potential 197.
Therapeutic use
Direct-to-consumer marketing of MSC therapies is widespread in the United States 100 despite the fact that few high-quality studies have been published 118. Moreover, the authors of many of the randomized controlled trials that evaluated the efficacy of MSC therapy for knee cartilage pathology have reported intra-group differences and/or inter-group sub-score differences, but have failed to report the original intended outcomes of the studies 135, suggesting a need for prudence moving forward. Although scientific optimism and enthusiasm is warranted, it is important that patients interested in these treatments understand that a lack of robust data exist to support their use.

The intra-articular therapy pipeline

Several intra-articular treatments for OA are currently in clinical development (Fig.  3). In the following section, we discuss some of the most anticipated and promising of these treatments and critically appraise the data that are currently available. Therapies in this section have all had clinical data published or presented and are progressing towards FDA approval. As such, promising therapeutic candidates that are currently in early clinical or preclinical development, such as RCGD 423 (ref. 136) and UBX0101 (refs 137,138), are not discussed. Treatments for which clinical data have been published, but for which there are no current plans for further development, such as co-injected Tr14 (Traumeel) and Ze14 (Zeel) 139, have also been excluded from this Review.
One of the notable intra-articular therapies that is not included in the discussion below is AmnioFix, a mixture containing human amniotic and chorionic membranes that have been dehydrated, micronized and suspended in saline 140,141. This treatment deserves mention because, before 2013, a number of clinics in the United States were using AmnioFix to treat knee OA under the assumption that the treatment was ‘minimally manipulated’. However, MiMedx (the company that produces AmnioFix) was forced to recant their claims when the FDA warned that the treatment could not be offered to patients until an IND application was obtained 142. In 2017, MiMedx reported that the FDA had allowed their IND-approved, phase IIb study 143 to proceed. However, no clinical data have been released and only one animal study investigating the effects of AmnioFix for knee cartilage pathology has been published 144.

LMWF-5A (Ampion)

LMWF-5A (Ampion) is an injectable, low molecular weight fraction of 5% human serum albumin that is currently being developed by Ampio Pharmaceuticals (Englewood, FL, USA). The primary constituent of LMWF-5A, artyl-alanyl diketopiperazine, modulates the inflammatory immune response in vitro via a pathway in which T cells are implicated 145, and several clinical studies have been published in which the effects of LMWF-5A on knee OA were investigated 146148.
To date, the most notable study 146 to investigate the use of LMWF-5A to treat knee OA was published in 2014. In this study, 329 patients were randomized at a ratio of 1:1:1:1 to receive a single 4 ml or 10 ml intra-articular injection of either LMWF-5A or saline. The WOMAC pain scores of patients treated with LMWF-5A were significantly better than those treated with placebo at week 12 (estimated difference from control −0.25, P = 0.004) 146. However, the effect size was roughly one quarter of what was assumed and such a small difference is not likely to be clinically meaningful to an appreciable number of patients 32. Nevertheless, Ampio Pharmaceuticals announced that their phase III clinical trial met its primary endpoint, with 71% of patients in the treatment group meeting the Outcome Measures in Rheumatology (OMERACT)–OARSI responder criteria 149. Overall, the available data suggest that the short-term effects of LMWF-5A might be non-inferior (although they are not likely to be superior) to currently used intra-articular treatment modalities. The long-term effects of LMWF-5A have yet to be determined, but the results of in vitro studies 145,150 and of a post hoc analysis 148 are encouraging. An open label, phase III extension study to evaluate the long-term safety of LMWF-5A is currently underway 151.

HA–triamcinolone hexacetonide (Cingal)

Cingal is an HA-triamcinolone hexacetonide combination drug that was developed by Anika Therapeutics, Inc. (Bedford, MA, USA). A 2017 double-blind, saline-controlled trial compared the use of Cingal, HA (the same as that used in Cingal) and saline in 149, 150 and 69 patients with knee OA, respectively 152. In this study, Cingal provided better symptomatic relief than placebo, as measured by the WOMAC pain score at 26 weeks. However, Cingal only produced statistically significant benefits at week 1 and week 3 when compared with HA alone. Although the authors of this study 152 claim that the rapid response seen in patients demonstrates the ‘clinical significance’ of Cingal, the report does not mention the MID and the short-term improvements compared with HA alone were modest (<10% difference) 152. A single-group follow-on study in which the safety of repeated Cingal injections was investigated was completed in 2015, but the results have yet to be announced or published 153. Overall, the currently available data do not support the use of Cingal over HA, although the results from an ongoing phase III trial ( n = 576) in which Cingal is compared with HA alone and triamcinolone hexacetonide alone 154 are needed before firm efficacy conclusions can be drawn.

Autologous protein solution

The nSTRIDE APS kit (Zimmer Biomet, Warsaw, IN, USA) is a single-use device that produces a ‘cell concentrate’ from autologous blood (Fig.  2), which the manufacturer refers to as an autologous protein solution (APS). Conceptually, APS and PRP (both of which contain white blood cells) are very similar. However, unlike traditional PRP systems, the nSTRIDE APS kit passes the concentrated plasma though a dried polyacrylamide gel that preferentially concentrates anti-inflammatory cytokines such as IL-1 receptor antagonist and TNF receptor inhibitor 155. Zimmer Biomet have also obtained an Investigational Device Exemption for the nSTRIDE APS kit 156, which means that, unlike PRP, if approved, the intra-articular administration of APS to patients with OA would not be considered off-label use.
APS has demonstrated preliminary feasibility in preclinical models 157,158 and safety in patients with OA 156,159. The results of the phase II trial 160 showed a statistically significant improvement in WOMAC pain scores at 12 months for those treated with APS compared with those treated with placebo, which the authors claim became apparent between 6 and 12 months. However, in this study 160, outcomes were not assessed between 6 and 12 months, the WOMAC was not a pre-specified outcome of interest and no statistically significant differences between the WOMAC pain scores for APS and placebo were found at week 2, month 1, month 3 or month 6. Additionally, the authors of the study failed to report that the primary outcome (VAS at month 6) was not met and that statistically significant differences were not found in the responder rate, VAS, quality-of-life, patient/clinical global impression of change or KOOS at any timepoint 161. Nevertheless, the generally positive safety profile of APS supports its continued clinical development, and the results from an ongoing phase III clinical trial ( n = 246) in which the nSTRIDE APS kit is compared with saline 162 are needed before preliminary efficacy conclusions can be drawn.


SM04690 is a novel small-molecule Wnt–β-catenin signalling pathway inhibitor that is currently being developed by Samumed LLC (San Diego, CA, USA) 163. Excessive activation of β-catenin-dependent signalling pathways can severely inhibit cartilage formation, as well as growth plate organization and function 164. Additionally, inhibition of β-catenin-dependent signalling pathways induces chondrogenesis and inhibits joint destruction in rats 165.
The exact mechanism of action of SM04690 is still under investigation 163, and previous attempts in mice to inhibit Wnt signalling pathways have failed 166,167. Nevertheless, the results from the first-in-human, 24-week, phase I RCT revealed SM04690 to be safe and well-tolerated, and showed no evidence of exposure outside of the injected joint 168. Additionally, all three treatment cohorts had reduced joint-space narrowing compared with the placebo group at 24 weeks. A phase IIb study to evaluate the safety and efficacy of SM04690 is still ongoing 169, but a phase IIa trial 170 in which 0.03 mg, 0.07 mg and 0.23 mg doses of SM04690 were compared with placebo was completed in 2017 ( n = 455). Results from this trial 170 have yet to be formally published, but preliminary findings have been announced by Samumed 171. The improvements in clinical outcomes and joint-space width (JSW) in the intention-to-treat population were not statistically significant, but a subpopulation of unilateral symptomatic patients in the mid-dose cohort demonstrated better improvements in WOMAC and medial JSW than patients in the control group at week 52 (ref. 171). This subpopulation was not a pre-specified group of interest for this trial 170, but unilateral symptomatic patients are listed as a subgroup of interest in the ongoing phase IIb trial 169.

rhFGF18 (sprifermin)

rhFGF18 (sprifermin) is a growth factor therapy being developed by EMD Serono (Darmstadt, Germany and Billerica, MA, USA). As mentioned in the section on growth factor therapy, the largely successful phase I trial showed preliminary safety and efficacy 80. Post hoc analyses of the phase I data showed slight improvements in cartilage at the patellofemoral joint 172, as well as a minor increase in total cartilage thickness and a small reduction in total cartilage loss 173. A 5-year randomized, placebo-controlled, phase II study in which three, weekly, intra-articular injections of placebo or sprifermin administered in cycles of once every 6 months or once every 12 months are compared ( n = 549), is currently in its final year 174. In the 2-year results, which were reported at the ACR annual meeting in 2017, the primary endpoint of a change in total tibiofemoral joint cartilage thickness from baseline was met 175. Patients who were treated with 100 µg of sprifermin every 6 months or every 12 months had a significantly greater increase in total tibiofemoral joint cartilage thickness than patients in the control group (+0.03 mm and +0.02 mm versus -0.02 mm, respectively; P < 0.001) 175. Although clinical outcome data showing improvements in pain and function compared with placebo are still needed to validate the utility of the reported structural modifications, the data released to date seem to suggest a disease-modifying benefit.


CNTX-4975 is an injectable, high-purity trans-capsaicin that is currently being developed by Centrexion Therapeutics (Boston, MA, USA). This treatment targets the capsaicin receptor (transient receptor potential cation channel subfamily V member 1), which contributes to the detection and integration of pain-producing stimuli 176. The analgesic effects of capsaicin-based treatments have been attributed to several different mechanisms (collectively referred to as the ‘defunctionalization’ of nociceptive fibres), including the transient retraction of nerve fibre terminals 177,178. Before the technology was acquired by Centrexion Therapeutics, a less-purified version of CNTX-4975 (ALRGX-4975) was clinically investigated by Anesiva Inc. as a treatment for postoperative pain under the name Adlea 179. The results of the studies conducted by Anesiva Inc. were never published or released, and it was not until Centrexion Therapeutics acquired the technology that the treatment was first administered via intra-articular injection to treat moderate-to-severe OA-related pain.
The results of the phase IIb study in which CNTX-4975 was compared with placebo were reported at international meetings in 2017 (refs 180,181). Patients treated with 1 mg CNTX-4975 ( n = 70) had improved WOMAC A1 scores (10-point VAS) at week 12 (least squares mean differences (LSMD) −1.6) and at week 24 (LSMD −1.35) compared with patients who received placebo ( n = 69). With respect to intra-articular therapies currently in clinical use, the magnitude of difference and duration of benefits for CNTX-4975 is very encouraging, particularly given the inclusion of morbidly obese patients and patients with severe OA (Kellgren-Lawrence grade II–IV). The most notable limitation of CNTX-4975 is that it does not halt or reverse the course of OA pathogenesis. Clinical data have been presented at a number of scientific congresses, but the preliminary data have yet to be published. Nevertheless, Centrexion Therapeutics has begun recruiting participants for a phase III trial ( n = 325) 182, and this treatment was granted Fast Track designation by the FDA in 2018.

TissueGene C (Invossa)

TissueGene C is a cell therapy being developed by Kolon Tissue Gene, Inc. (Rockville, MD, USA). The therapy, which is currently available in South Korea under the trade name Invossa, is a mixture containing chondrocytes that have been transduced with a viral vector containing TGFB1. In preclinical studies, the genetically modified chondrocytes demonstrated long-term transforming growth factor-β secretion and constitutive type II collagen expression in vitro, as well as the ability to form cartilage in vivo 183. Following the successful completion of a phase I safety study 184, a phase II, multicentre, double-blinded, placebo-controlled RCT was conducted in which TissueGene C ( n = 67) was compared with saline ( n = 35) 185. Although no difference was found between TissueGene C and saline in the International Knee Documentation Committee (IKDC) score or VAS of patients at week 4 or week 24, statistically significant differences between the groups were observed at week 12, week 52 and overall 185. The magnitude of the overall difference was notable (12-week VAS (LSMD -13.8; CI -25.0 to -2.6), 52-week VAS (LSMD -13.1; CI -25.1 to -1.1)), particularly for a phase II trial. However, the analysis was not intent-to-treat and 16 patients in the treatment group (24.2%) and 6 patients in the control group (18.8%) were lost to follow-up, which could have introduced selection bias. A 2017 publication in which the structural effects of TissueGene C were detailed provided unclear evidence of disease modification 186.
Data from a multicentre, double-blind, clinical trial conducted in South Korea revealed that TissueGene C provided long-term clinical benefits compared with placebo 187. In this study, 163 patients with Kellgren-Lawrence grade III OA were randomly assigned to receive a single intra-articular injection of TissueGene C or saline. The study met both of its primary outcomes by showing statistically significant improvements in IKDC scores (+15 versus +5; P < 0.001) and VAS (-25 points versus -10 points; P < 0.001) at 52 weeks compared with control 187. The IKDC score and VAS were also improved at week 26 and week 39, and statistically significant improvements were seen in secondary clinical outcomes (WOMAC and KOOS) at week 52 compared with control 187. Although significant differences on radiographs (changes in JSW) or whole-organ MRI scoring were not observed, post hoc analysis of MRI data suggested possible (albeit small) improvements in cartilage thickness 186. Even in the absence of disease modification, the relatively robust and consistent long-term improvements of patients treated with TissueGene C across different clinical outcomes is promising. A follow-up phase III study ( n = 510) has been announced 188 that is powered to determine if the treatment provides a disease-modifying effect.

FX006 (Zilretta)

FX006 (Zilretta) is an intra-articular microsphere-based formulation of triamcinolone acetonide developed by Flexion Therapeutics (Burlington, MA, USA) that has been approved by the FDA for use in treating OA-related knee pain 189. This extended-release glucocorticoid uses a proprietary matrix that is designed to prolong the analgesic benefits of triamcinolone acetonide in patients with knee OA 190. A phase IIa trial in which FX006 was compared with triamcinolone acetonide failed to show a statistically significant improvement in pain from baseline at 12 weeks, but did demonstrate significant improvements at 8 weeks and 10 weeks 191 (all three timepoints were pre-specified as primary outcomes 192). The promising results of this trial prompted a follow-up phase IIb trial to compare 32 mg FX006 ( n = 104), 16 mg FX006 ( n = 102) and saline ( n = 100) 193. Similar to the phase IIa trial, the follow-up phase IIb trial failed to show statistically significant improvements in average daily pain (ADP) at 12 weeks compared with saline (primary outcome), but did show a statistically significant difference in ADP between FX006 and saline at weeks 1–11 and at week 13 (secondary outcomes) 193.
In a 24-week, phase III, multicentre, double-blinded RCT, patients with knee OA were treated with a single intra-articular injection of FX006 ( n = 161), placebo ( n = 162) or triamcinolone acetonide ( n = 161) 194. Although patients treated with FX006 had a greater improvement in the ADP at week 12 than those treated with placebo (-3.12 versus -2.14), an improvement in ADP compared with patients treated with triamcinolone acetonide was not seen 194. The failure to find a statistically significant difference between triamcinolone acetonide and FX006 might have resulted from the large assumed effect size, which was considerably larger than can be justified by the results of the phase IIb study 193. Moreover, although statistically significant improvements were seen in several exploratory endpoints (such as WOMAC subscales and KOOS quality-of-life) compared with triamcinolone acetonide, the differences were very small and not likely to be clinically significant.
Overall, FX006 seems to provide better pain relief than intra-articular saline; however, the advantages over traditional formulations of triamcinolone acetonide are unclear. Although it is possible that FX006 provides better pain relief than traditional triamcinolone acetonide formulations, particularly in the short-term (<12 weeks), additional appropriately powered studies that compare traditional triamcinolone acetonide formulations and FX006 are still needed, and it seems unlikely that the magnitude of the difference would be large enough to be clinically important to an appreciable number of patients. Given the lack of long-term data, the potentially harmful long-term effects of glucocorticoids and the questionable clinical benefits of FX006 compared with traditional triamcinolone acetonide, clinicians should be particularly careful when prescribing FX006.


For most of the 21st century, HA and glucocorticoids have been the standard intra-articular treatments for the management of knee OA in patients who fail to respond to non-pharmacological therapy, NSAIDs or analgesics. Although the prospect of new, non-surgical treatments for knee OA is likely to cause excitement in both clinicians and their patients, the benefits of new treatments must be carefully weighed against their cost and potential risks. It should be remembered that a strong placebo response exists towards agents administered via intra-articular injection, and that new intra-articular treatments might not be appropriate for every patient.


The work of the authors was supported by grants UL1TR001855 and UL1TR000130 from the National Center for Advancing Translational Science (NCATS) of the US National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Reviewer information

Nature Reviews Rheumatology thanks C. Evans, D. Hunter and A. Migliore for their contribution to the peer review of this work.

Competing interests

C.T.V.Jr. declares that he holds shares in CarthroniX Inc. and in Parcus Medical. The other authors declare no competing interests.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Related topics