Mesenchymal Stromal Cell-Based Therapy—An Alternative to Arthroplasty for the Treatment of Osteoarthritis? A State of the Art Review of Clinical Trials

Osteoarthritis (OA) is the most common degenerative joint disorder worldwide and to date no regenerative treatment has been established in clinical practice. This review evaluates the current literature on the clinical translation of mesenchymal stromal cell (MSC)-based therapy in OA management with a focus on safety, outcomes and procedural specifics. PubMed, Cochrane Library and clinicaltrials.gov were searched for clinical studies using MSCs for OA treatment. 290 articles were initially identified and 42 articles of interest, including a total of 1325 patients, remained for further examination. Most of the included studies used adipose tissue-derived MSCs or bone-marrow-derived MSCs to treat patients suffering from knee OA. MSC-based therapy for knee OA appears to be safe and presumably effective in selected parameters. Yet, a direct comparison between studies was difficult due to a pronounced variance regarding methodology, assessed outcomes and evidence levels. Intensive scientific engagement is needed to identify the most effective source and dosage of MSCs for OA treatment in the future. Consent on outcome measures has to be reached and eventually patient sub-populations need to be identified that will profit most from MSC-based treatment for OA.


Osteoarthritis
With a steadily increasing prevalence in past decades, 14 million US individuals are currently estimated to suffer from symptomatic knee OA [1]. A constantly aging society will lead to a dramatic rise in affected patients and concomitant joint replacement surgery in the coming years [2].
Underlying pathomechanisms of OA are versatile and include both, intrinsic and extrinsic factors. Intrinsic factors include age, gender, menopause, genetics, nutrition and bone density, and extrinsic or mechanical factors include body weight, body mass index (BMI), injuries, previous surgeries and joint deformities, which are all directly linked to an increased susceptibility to OA [3,4]. Mostly a (1) The initially favoured theory, that MSCs mainly differentiate into cells of a specific mesodermal lineage and replace damaged or missing cells has lost its popularity in recent years. However, MSCs seem to share the ability to induce tissue regeneration through the stimulation of local endogenous cells [13]. (2) Today, the major potential of MSCs is seen in their secretion of paracrine factors ("bystander effect"), that allows an immunomodulation of the local pro-inflammatory environment, which plays a key role in cartilage degeneration in OA [14,15]. Whether paracrine effects are long-lasting or merely a "hit and run" phenomenon is currently debated [16].

First MSC Trials and Cartilage Repair
The origin of today's cell-based therapy concepts dates back to the 1960s, when Friedenstein characterized mesenchymal stem cells or MSCs [17]. MSCs are multipotent precursor cells of mesodermal origin. They are able to differentiate into tissue cells of the mesodermal lineage like chondrocytes, osteoblasts, adipocytes and myocytes. Even a differentiation into neuronal cell types has been described [18]. Due to their differentiation potential, MSCs raised the hope to be exploitable in the treatment of degenerative diseases such as OA through the regeneration of damaged cartilage. Ashton et al. were the first to report of the chondrogenic potential of MSCs in vivo in 1980 [19]. 30 years later, Mokbel et al. showed that the repair of damaged articular cartilage tissue happened through homing, engraftment and production of cartilage matrix by MSCs [20]. Further animal experiments using MSCs for the treatment of OA followed, demonstrating reduced articular cartilage degeneration, osteophytic remodelling, and subchondral sclerosis after intra-articular injection of MSCs [13,21,22].
Early in vitro studies demonstrated the capacity of MSCs to form cartilage-like tissue, evident by the production of type II collagen and hypertrophic chondrocytes after growth-factor stimulation [23]. Yet, the quality of tissue-engineered cartilage was reported to be compromised, indicated by low collagen contents with a presumably negative impact on the load-carrying properties of cartilage [24]. New cell-carrier scaffolds, made from natural cartilage extra-cellular matrix, promoted proliferation and chondrogenic differentiation of bone-marrow-derived MSCs (BMMSCs) without outside stimulations in an in vivo animal model [25].
A 2002 study published by Wakitani et al. marked one of the first OA trials of BMMSCs used in humans. Twenty four patients suffering from knee OA were either treated with a high tibial osteotomy (HTO) combined with intra-articular autologous cultured BMMSC injections or with an HTO alone as a control group. At the 16-month follow-up, pain scores, function and muscle strength had significantly improved in the treatment compared to the control group [26].
With the rise of this new cellular-based therapy, concerns about the quality of autologous MSCs from patients suffering from OA arose. In 2002 Murphy et al. described a compromised ability of MSCs from OA patients to differentiate into chondrocytes [27]. A more recent study however found that MSCs from OA patients did in fact have the same chondrogenic differentiation potential as age-matched patients with femoral fractures [28]. These contradicting results may be explained by the fact that the compared patient collectives in the study by Murphy et al. [27] were not strictly age-matched and an isolated impact of age may not be disregarded, as MSCs were described to lose their chondrogenic differentiation potential with age [29][30][31]. The authors further discussed whether the reduced differentiation potential of progenitor cells may even be a cause for OA, rather than a result of it [27].
An increased chondrocyte proliferation was observed when chondrocytes were co-cultured with MSCs and proliferation was further associated with the ability to synthesize cartilaginous extracellular matrix in vitro [32,33]. In turn, anti-proliferative effects have been described, when MSCs were challenged with "negative" signals, like interferon-γ or bacterial stimuli like lipopolysaccharide [22,34].

OA as an Inflammatory Disease and Immunomodulatory Properties of MSCs
For a long time, OA was classified as a non-inflammatory "wear and tear" disease. This changed through the works of Goldring and Scanzello et al., who helped to categorize OA as a stress-induced and pro-inflammatory disease in 2011 [35,36].
Numerous studies followed, concluding the same and showing a direct role of pro-inflammatory factors, such as tumor necrosis factor (TNF)-α and interleukin (IL)-1β in OA pathogenesis [37,38]. Since then inflammation and synovitis have been fundamentally included in the theoretical framework of the OA pathomechanism and are also considered targets for new treatment modalities such as MSC-based therapy [16,39]. In addition to the differentiation potential of MSCs towards chondrocytes and the recruitment of neighbouring progenitor cells, it was proven that MSCs have potent anti-inflammatory as well as immunomodulatory properties [40,41].
Immunomodulatory effects of MSCs were reported for adipose tissue-derived MSCs (ATMSCs) from different adipose tissues (abdominal fat, infrapatellar Hoffa fad pad, subcutaneous hip fat), as they show a reduction of inflammatory factors like TNF-α, IL-1β and CCL3/macrophage inflammatory protein-α, when co-cultured with chondrocytes and synoviocytes from OA patients [42].
The immunomodulatory effects of ATMSCs and BMMSCs show some similarities, as both cell types reduce T-cell numbers and subsequent pro-inflammatory cytokines [43,44].
The mechanism behind the regulation of local inflammatory environments, is in part explained by the secretion of growth factors, chemokines, cytokines, micro-RNAs and extracellular vesicles (EVs) by MSCs. MSC-secreted factors affect macrophages, dendritic cells, T-and B-cells, neutrophils and natural killer cells [45]. Further, ATMSCs and BMMSCs decrease antibody production and B-cell chemotaxis by inhibition of B-cell proliferation through cell-cycle arrests and apoptosis in a T-cell independent manner [46,47].
Keeping in mind that inflammation plays a key role in osteoarthritic joints, immunomodulatory properties have been used to explain partial joint tissue healing through MSCs. The deployment of a suitable environment for tissue regeneration by MSCs may therefore either be achieved by direct secretion of bioactive materials or by controlling and containing cytokine and growth factor production from effector cells [44]. Anti-inflammatory and immunomodulatory effects of MSCs can further be modulated and enhanced by ex vivo "priming" of MSCs with cytokines, hypoxia, chemical agents, biomaterials, and receptor challenging. A comprehensive literature overview regarding priming approaches for MSCs was recently published by Noronha NC et al. [48] and a summary of growth factors, chemokines, cytokines, micro-RNAs and EVs, secreted by MSCs was provided by Lin et al. [45].

Oncological Safety Profile of MSCs
Due to the high proliferative capacities of MSCs and the potential of tumour-derived cytokines to attract MSCs to tumour sites, oncological side effects must be discussed when assessing the safety profiles of cell-based therapies. MSCs have shown to exert potential tumorigenicity in several murine and in vitro models, yet no novel cancers or cancer recurrences have been diagnosed in clinical trials to date, that would originate from administered MSCs as summarized in recent reviews by Lee et al. [49] and Ridge et al. [50].
Toyserkani et al. conducted a concise systematic review about the safety of ATMSCs in 2017, assessing 70 studies with over 1400 patients, who received ATMSCs for various conditions. Thromboembolic complications were registered for the systemic application of MSCs, yet the local treatment appeared to be safe, especially regarding oncological side effects. One case of breast cancer recurrence was identified in 121 patients with a known history of previous breast cancer. Although the authors came to the conclusion that ATMSCs have a favourable oncological safety profile, they stressed the low quality of reporting of adverse events (AEs) and the short follow-up period of 12 months [51]. Standardized reporting systems for AEs and long-term follow-ups of several years are therefore needed to fully evaluate the oncological safety of MSC-based therapy in the future.

Preparation of MSCs
MSCs can be harvested from various tissues of mesenchymal origin. These include e.g., bone-marrow, adipose tissue, synovial fluid, skeletal muscle and the placenta. MSC-based therapy for OA follows the basic principle of a local application of harvested cells into the affected joint, after ex vivo isolation, concentration or cell expansion. Delivery of MSCs is either achieved through an intra-articular injection or an arthroscopically assisted implantation of MSCs. The distinction between autologous and allogenic sources stems from the origin of MSCs. Allogenic cells are harvested from a suitable donor, whereas autologous cells originate from the patients themselves. In general, autologous MSCs are often seen as the safe choice, as concerns about unwanted immune-responses or transplant rejections can be disregarded [52]. The complex autologous MSC preparation procedure that may include cell culture and expansion, makes autologous solutions however a more complicated and expensive method opposed to allogenic alternatives. Allogenic MSCs are cheaper and logistically more convenient, as they can be obtained as an off-the-shelf product [53].
In the context of ex vivo expansion, xenogenic elements (e.g., remains of fetal bovine serum) should, if possible, be replaced by humanized material or eliminated as far as possible to reduce post-transplantation risks such as transmission of microbes (viruses, bacteria, prions), cytotoxicity, uncontrolled and unpredictable growth patterns as well as the theoretical risk of anaphylactic reactions [54].
Further, some patients' comorbidities (e.g., anaemia, severe osteoporosis) do not allow harvesting of autologous MSCs and certain underlying autoimmune conditions like rheumatoid arthritis [55,56] and type 2 diabetes [57], may impair the immunosuppressive capacities of MSCs, which in those cases leaves allogenic MSC-based therapies as the more feasible choice. Generally, three types of altered autologous MSC products can be distinguished: (1) MSC concentrates are the product of harvested fluids or tissues, which have been concentrated in order to increase the number of MSCs per unit of applied suspension. Bone-marrow aspirate concentrates (BMAC) include high concentrations of growth factors and high levels of anti-inflammatory IL-1 receptor antagonist next to MSCs [58]. (2) Culture-expanded MSCs can be administered without scaffolds or seeded onto scaffolds after having been isolated and cultured. Traditionally two-dimensional (2-D) plastic culture plates are in use for cell expansion, but three-dimensional (3-D) techniques have evolved in recent years. 3-D cultures are able to mimic in vivo conditions and provide high density and expansion potential [59]. New additions like highly elastic culture dishes, automatically adapting the dishes' surface to growing cell numbers, optimize the expansion process before MSCs are ready for application [60]. Increased complexity and non-standardized expansion protocols, alongside higher costs should be taken into consideration, when applying 3-D expansion techniques. However, reports about increased immunomodulatory and chondrogenic potential of MSCs, cultured in 3-D compared to conventional 2-D cultures, point towards promising technical developments in the field of tissue engineering for OA treatment [61,62]. (3) The stromal vascular fraction (SVF) is a combination of ATMSCs, endothelial cells, growth factors, precursor cells, macrophages, t-regulatory cells, lymphocytes and others. The SVF is derived from lipoaspirates by mechanical or enzymatic isolation [63]. The SVF can be injected into the joint, after a purification process, often within the same visit, as there is no need for expansion or culturing of cells.
As with many novel therapeutic strategies, a standard is not yet available and different approaches are present in the current literature regarding cell origin, harvesting technique, cell dose, culturing and application. Moreover, controversy prevails regarding outcome measurements and follow-up times in current study designs. In this review we analysed the relevant clinical literature on MSC-based treatment for OA with special regards to safety and efficacy. As the field of regenerative medicine and especially the use of MSCs in degenerative joint disorders has become more and more relevant for the medical and public community, we hope to shed some light on this valuable and fast-evolving treatment concept.

Methods
First step: A literature search was performed to identify all relevant articles, involving MSC-based treatment for OA. Pubmed (MEDLINE), The Cochrane Library and clincialtrials.gov were searched until the 13 of April 2020. For Pubmed and The Cochrane Library searches, MeSH (Medical Subject Headings)-terms were used as shown below.
We included all clinical studies, that used MSCs to treat OA in humans, with a recruitment of at least seven participants. Articles from any country were included but limited to those published in English language. We excluded any preclinical studies, case reports, review articles and studies addressing isolated, focal chondral defects not clearly associated with OA. Further excluded were all articles older than ten years, to ensure actuality of this review.
Third step: Reference lists of included articles were scanned regarding potentially missed studies. Missed studies were then identified and added by hand.
Fourth step: Included articles were analysed regarding number of patients, follow-ups, treatment details, outcome measures, study type, control arm and site of OA (see Table 1).
The initial literature search of PubMed yielded a total of 290 articles. The search of The Cochrane Library database yielded 16 results, of which all had already been covered by the previous Pubmed search. After removal of duplicates 290 articles were scanned and full-text evaluation followed. 34 articles were identified as relevant to this review. The reference lists of the 34 selected articles were scanned regarding missed studies and eight studies were added by hand. A total of 42 studies, including 1325 patients, were examined in detail (see Figure 1 and Table 2). J. Clin. Med. 2020, 9, x FOR PEER REVIEW 6 of 30 English language. We excluded any preclinical studies, case reports, review articles and studies addressing isolated, focal chondral defects not clearly associated with OA. Further excluded were all articles older than ten years, to ensure actuality of this review. Third step: Reference lists of included articles were scanned regarding potentially missed studies. Missed studies were then identified and added by hand.
Fourth step: Included articles were analysed regarding number of patients, follow-ups, treatment details, outcome measures, study type, control arm and site of OA (see Table 1).
The initial literature search of PubMed yielded a total of 290 articles. The search of The Cochrane Library database yielded 16 results, of which all had already been covered by the previous Pubmed search. After removal of duplicates 290 articles were scanned and full-text evaluation followed. 34 articles were identified as relevant to this review. The reference lists of the 34 selected articles were scanned regarding missed studies and eight studies were added by hand. A total of 42 studies, including 1325 patients, were examined in detail (see Figure 1). As of April 13th 2020, the search on clinicaltrials.gov yielded 96 results. 25 trials were currently recruiting OA patients for MSC treatment (see Table 1) and two additional trials were currently recruiting, but not addressing MSC treatment in OA directly (acetabular labrum, "synovial brushing" study). 34 trials were already completed, 11 were not yet recruiting and 24 were either "withdrawn", "unknown", "suspended", "terminated" or "active, not recruiting".                 As of April 13th 2020, the search on clinicaltrials.gov yielded 96 results. 25 trials were currently recruiting OA patients for MSC treatment (see Table 1) and two additional trials were currently recruiting, but not addressing MSC treatment in OA directly (acetabular labrum, "synovial brushing" study). 34 trials were already completed, 11 were not yet recruiting and 24 were either "withdrawn", "unknown", "suspended", "terminated" or "active, not recruiting".

Safety
Although not all of the included studies evaluated the safety of MSC-based treatment for OA, all reported treatment-related serious adverse events (SAEs) and AEs were described to have subsided without sequelae. Treatment-related AEs included either injection site complications like local knee pain and swelling or harvesting site complications like low back pain after iliac bone harvesting and haematoma after liposuction.
Gupta et al. reported one treatment-related SAE (synovial effusion) and nine treatment-related AEs in a study of 60 patients with knee OA after intra-articular injections of allogenic BMMSCs. No sequelae of the SAE were reported [78]. Bastos et al. reported two treatment-related SAEs (intense knee pain) and eight treatment-related AEs in 18 knee OA patients, treated with autologous BMMSC injections. SAEs were treated with analgesics and resolved without lasting impairments [70].
Mild to severe AEs were reported in 80-100% of patients receiving either single or repeated ATMSC injections for knee OA in a trial by Freitag et al. No significant difference was observed between treatment groups. Discomfort and swelling at the injection site were the most commonly reported AEs [83].
Soler et al. reported of 14 mild and two moderate treatment-related AEs ranging from arthralgia and joint swelling to knee locking and back pain in a study of 15 knee OA patients treated with autologous BMMSC injections [74]. Orozco et al. reported transient and mild injection site pain for a few days in 50% of patients in a study treating 12 knee OA patients with autologous BMMSCs [76]. Song et al. reported 20 treatment-related AEs in 18 patients, receiving different intra-articular cell doses of autologous ATMSCs for OA of the knee. Transient pain and joint swelling occurred equally distributed among groups and were all spontaneously relieved [98]. Chahal et al. treated 12 patients with BMMSCs for knee OA and reported of four patients with minor transient AEs (local pain and swelling), which all subsided without intervention [72].
None of the authors reported any lasting impairments resulting from treatment-related SAEs or AEs. All reported cases resolved within a few days and no patient had to terminate any of the trials preliminary due to safety issues. Moreover, reported AEs were similar to those previously reported during clinical trials, evaluating intra-articular injections of HA or GC [111,112]. It can therefore be assumed that treatment-related AEs are of procedural nature and that intra-articular MSC injections are safe, regardless of tissue origin and applied cell dose. Long-term follow up studies of seven and four years, conducted by Park et al. [104] and Lamo-Espinosa et al. [15] reported no further SAEs and AEs that persisted over the acute post-injection phase.
Almost all included studies recorded clinical improvements and cartilage regeneration for the first 12 months after the treatment. Studies with longer follow-up periods reported further improvements over time [85,86,[88][89][90]95], plateau phases with maintained clinical improvements [100,104], or deterioration of previous improvements after 30 months [73]. A four-year follow-up study, conducted by Lamo-Espinosa et al., showed clinical and functional improvements four years after a single intra-articular injection of BMMSCs [65]. Larger randomized controlled trials with extended follow-up visits will be needed to gain further insights into the question of how long locally injected MSCs can exert their effects on the joint. So far, relevant clinical improvements have been recorded for as long as six to 12 months in most published studies.

Quality of Life and Mental Health
The Quality of Life Short Form-36 (SF-36) questionnaire evaluates quality of life, using eight sub-scores, concerned with the physical and mental status of the patient. It is widely used to evaluate therapy success with a focus on the patient's physical and emotional state during or after an intervention. Interestingly, while most evaluated clinical (e.g., pain visual analogue scale (VAS) score) and structural (e.g., Whole-Organ Magnetic Resonance Imaging Score (WORMS)) parameters of included studies improved over time, results of the SF-36 did often not correlate [74,76,101,103]. The SF-36 questionnaire is often criticized for being too generic and less sensitive than for example the Western Ontario and McMaster Universities OA Index (WOMAC) score, which was specifically developed to assess lower extremity arthritis [113]. The SF-36 questionnaire should therefore be interpreted with caution and may be replaced if possible, by more specific alternatives.
MSC-induced cartilage improvements, visible in MRI, are an ongoing topic of debate and contradicting results are continuously reported. While some authors demonstrate cartilage regeneration in accordance with clinical improvements [76,77,82,87,97,99], others could not confirm changes in cartilage structure, although clinical improvements were observed [95,103,115].
In 2008 Centeno et al. reported one of the first intra-articular injections of autologous BMMSCs in a patient with OA of the knee. The group found that six months after treatment the patient had already shown statistically significant growth of cartilage and meniscus in the conducted MRI [114]. Soler et al. and Vega et al. could both confirm improved cartilage quality at the 12-month follow-up for knee OA patients, treated with BMMSCs [74,79]. Wong et al. showed a significantly improved Magnetic Resonance Observation of Cartilage Repair Tissue (MOCART) score one year after HTO, microfracturing and intra-articular injection of autologous BMMSCs mixed with HA in 28 knee OA patients compared to the control group, comprised of 28 OA patients who underwent the same procedure without MSC injection [67]. In 2013, Koh et al. reported that improved MRI scores were positively correlated with the number of injected MSCs [97].
Pintal et al. treated 19 knee OA patients with autologous ATMSCs and although they observed a significant improvement in clinical scores at the six-and 12-month follow-up, no significant differences in cartilage constitution were observed in the conduced MRI [96]. Gupta et al. treated 60 knee OA patients with varying doses of allogenic BMMSCs against a control group. While pain VAS and clinical scores improved in all treatment groups, no cartilage changes were detected [78]. In accordance, Matas et al. found significant pain and functional improvements in knee OA patients, treated with UCMSCs, yet no structural changes were evident in conducted MRI [103].
As an alternative more invasive outcome measure, arthroscopic evaluations can be conducted, which are mostly performed as a second-look intervention at a scheduled follow-up.
In 2018 Kim and Koh performed a comparative matched-pair analysis, comparing outcomes of HTO-treated with HTO + SVF-treated knee OA patients. The authors reported significant cartilage improvements in the latter, according to ICRS grades, assessed 12 months after the intervention. ICRS grades correlated further with clinical outcome measures [86].
Koh et al. compared HTO + PRP + autologous ATMSC-treated (n = 21) with HTO + PRP-treated knee OA patients (n = 23) and second-look arthroscopy after a mean of 19.8 months showed significantly more cartilage healing in the HTO + PRP + ATMSC group [81]. According to the Kanamiya grading system [116], 50% of patients in the HTO + PRP + MSC group showed partial or fibrocartilage cover, whereas this was only shown for 10% in the control group [81].
Most of the aforementioned studies need to be interpreted with caution as they showed a low level of evidence. Results from a recent meta-analysis that assessed five randomized controlled trials on MSC-based therapy for knee OA [64,67,78,79,81] found that limited evidence for pain relief and functional outcomes does exist, yet there is a lack of evidence to support the claim that MSC-based therapy facilitates cartilage repair [115].

Cell Dose
The evaluation of cell dosing is important in order to find the minimum dose of MSCs, which is safe and still provides best overall outcomes. Given the limited amount of data and protocols on MSC therapy, most of the included studies were in fact dose-finding phase I or II trials.
The used cell doses ranged from the lowest mean dose of 1 × 10 6 autologous BMMSCs [72] to a mean dose of 1.5 × 10 8 allogenic BMMSCs [78]. The lowest mean dose was used by Chahal et al. in a dose-finding trial from 2019. Twelve knee OA patients were treated with escalating cell doses. Best clinical and radiological results were obtained in the high dose group of 5 × 10 7 cells [72].
Gupta et al. compared four doses of allogenic BMMSCs (2.5 × 10 7 , 5 × 10 7 , 7.5 × 10 7 and 1.5 × 10 8 BMMSCs mixed with 2 or 4 mL of PLASMA-LYTE A followed by an injection of HA (20 mg) in 60 knee OA patients in a randomized, double-blinded, multicentric, placebo-controlled study. They were able to show that the lowest dose of 2.5 × 10 7 cells provided the maximum pain reduction in all subjective parameters (WOMAC and pain VAS score), yet the observed improvement was not significant, when compared to placebo [78].
Pers et al. compared three cell doses (2 × 10 6 , 1 × 10 7 and 5 × 10 7 autologous ATMSCs in the form of SVF) and showed that patients treated with the lowest dose experienced significant improvements in pain and function WOMAC sub-scores, compared to baseline values. Of note, patients in the low-dose group had higher pain and function WOMAC sub-scores at baseline, compared to those receiving higher doses, which may have put a bias on the reported results [101].
In contrast, other authors yielded the best results for much higher cell doses in their studies [64,99,100,114]. Jo et al. compared three doses of autologous ATMSCs (1 × 10 7 , 5 × 10 7 and 1 × 10 8 ATMSCs) for the treatment of 18 knee OA patients. Improvements in the WOMAC score and improved cartilage volumes in MRI were observed in the high-dose group after six months. Second-look arthroscopy revealed a significant reduction of cartilage defects and a reduction of the International Cartilage Research Society (ICRS) grade in the high-dose group [99].
Jo et al. continued evaluating patients in a second phase of the study for a two-year follow-up. Improvements in WOMAC score, the Knee Society Score (KSS), the Knee injury and Osteoarthritis Outcome Score (KOOS) and reduced knee pain for up to 24 months were observed in both, the lowand the high-dose group. However, statistical significance was reached mainly in the high-dose group. Clinical outcomes and MRI scores declined after 12 months in the low-and mid-dose group, whereas a plateau was observed in the high-dose group until the 24-month follow-up [100]. A 2019 study by Freitag et al. found no differences between patients who were treated with a single injection of 1 × 10 8 cells or those who were treated with repeated injections of the same dose, six months apart (each time 1 × 10 8 cells) [83].
A retrospective analysis of 373 patients, who received autologous BMAC and PRP for 424 osteoarthritic knee joints, was able to demonstrate that patients receiving > 4 × 10 8 cells showed significantly lower post-treatment numeric pain scale values than patients receiving <4 × 10 8 cells. However, improved function according to the Lower Extremity Functional Scale and the International Knee Documentation Committee (IKDC) score were seen in all patients, regardless of dose. The preference of the relatively high cell dose in this study was biased by the cell count, which included all nucleated cells in the BMAC (BMMSCs, haematopoietic stem cells, monocyte precursor cells, macrophages, B-and T-cells, etc.) and not only MSCs [117].
The analysis of the existing data underlines that only limited evidence exists on the most efficient cell dose in OA treatment to date. The range of the ideal therapeutic doses is still open for debate and discussed vividly among physicians and scientists. Yet, data shows that a dose range from a mean of 1 × 10 6 autologous BMMSCs [72] to a mean of 1.5 × 10 8 allogenic BMMSCs [78] appears to be safe and effective in selected parameters.
Finally, a meta-analysis by Kim et. al. could not find a recommended cell dose, due to a great variation of concentrations, applied in different randomized controlled trials [115].
Further dose-finding, randomized, controlled, blinded and preferably matched trials with larger numbers of participants are needed in order to avoid biased results.
ATMSCs can easily be harvested through a simple and minimally invasive liposuction procedure. Liposuctions can be performed repeatedly and harvested cells show rapid expansion potential when cultured [118]. Adipose tissue contains a substantially higher number of MSCs than bone-marrow and ATMSCs are less affected by age and morbidity of patients. Further, ATMSCs maintain their differentiation potential even at later stages of life [119]. ATMSCs also exhibit the strongest anti-inflammatory potential when compared with other MSC sources [120].
Numerous clinical studies have proven the safety and efficacy of SVF therapy for OA. The isolation process and injection can usually be performed during the same visit, which makes SVF treatment an attractive alternative to pure ATMSC injections. Despite these advantages, the use of autologous ATMSCs has limitations. Cell manipulation due to ex vivo preparation and a lack of standardized harvesting protocols may impact the quality and quantity of cells. In this context, the MSC secretome, a valuable source of EVs, growth factors, and cytokines, has emerged as a possible cell-free therapeutic alternative [121]. Tofiño-Vian et al. were able to show that ATMSCs could serve as a source for EV exploitation [122], yet OA treatment, using the isolated secretome, has so far only been conducted in pre-clinical studies [121,123].
Currently recruiting clinical trials (see Figure 1) show a growing trend towards the use of allogenic UCMSCs and Wharton Jelly-derived MSCs (WJMSCs). Park et al. used UCMSCs in the treatment of seven knee OA patients and performed an extended follow-up of seven years. The study reported improved cartilage tissue at the 12-week follow-up and significantly improved pain VAS and IKDC scores at the 24-week follow-up. Improved clinical scores remained stable without significant deterioration for up to seven years. MRI findings showed preserved regenerated cartilage after three years [104]. A recently published study supports the favourable safety and efficacy profile of UCMSCs for knee OA treatment in 23 patients [103]. PLMSCs were also reported to be safe and effective regarding pain relief and improved range of motion in OA patients. Significant changes in the PLMSC-treated group were however only detected until eight weeks after treatment, when compared to the saline-treated control group [105]. Further studies with bigger patient samples are required to assess the safety and efficacy of UCMSCs and PLMSCs in depth. For the inter-study comparison, the group evaluated the efficacy by taking pain relief divided by the initial pain score, a method described by Huskisson et al. in 1974 to provide a tool for assessing general treatment efficacy [124]. The results showed the best efficacy of 0.75 for the autologous BMMSC treatment (Orozco et al. [76]) with an effect size of 1.29 versus an efficacy of 0.39 and an effect size of 0.96 for the autologous ATMSC treatment (Jo et al. [99]). A slightly worse efficacy of 0.36 was determined for the allogenic BMMSC treatment with an effect size of 1.07 (Vega et al. [79]).
Due to the differing patient numbers, control arms, follow-ups and injection protocols of MSC injections, a comparison of studies like this has limited relevance. This was also concluded by Shariatzadeh et al. in a recent review comparing the level of efficiency of different MSC sources for knee OA treatment. The authors concluded that exact protocols of MSC characteristics, culture, dosage, and clinical application are necessary for a final evaluation and efficient comparison of different studies [39].
Vega at al. suggested future studies that directly compare autologous and allogenic MSCs within the same study [79]. Applying similar treatment protocols to both study arms would help to determine the most suitable source of MSCs, and analogous to dose-finding trials, "source-finding trials" would improve knowledge on safety and efficacy of different MSC tissue origins.

Independent Outcome Predictors
Identifying factors that are directly associated with clinical outcomes of a treatment is valuable as it allows prognostic statements about the efficacy of a potential therapy. Schiavone Panni et al. could recently demonstrate that patients with a baseline pain VAS score > 8 showed significantly greater improvements in clinical and functional outcomes after SVF-based treatment, when compared to patients with pain VAS scores < 8 [90].
Koh et al. observed that overweight patients (BMI > 27.5 kg/m 2 ) and patients with a large cartilage lesion size (≥5.4 cm 2 ) showed significantly worse clinical (IKDC score and Tegner activity scale) and arthroscopic (ICRS grade) outcomes in ATMSC-treated knee OA patients. Other factors, including age and sex, did not have a significant influence on outcomes [92]. The negative effect of morbid obesity on ATMSCs' proliferation potential and multilineage differentiation capacity was proven in previous animal and human studies and may be linked to the high secretion of inflammatory cytokines in overweight patients [125,126].
Kim et al. [91] confirmed these findings in a retrospective study from 2015, analysing 49 knee OA patients after intra-articular autologous ATMSC treatment. The group could show that patient age and cartilage lesion size were independent predictors of clinical failure. Patients older than 60 years and patients with cartilage lesions larger than 6.0 cm 2 were at risk of poor clinical outcomes (IKDC score and Tegner activity scale). The study also showed a statistically significant association between age and cartilage lesion size as well as between BMI and cartilage lesion size. Sex, side of involvement (left or right), lesion location and BMI could not predict clinical outcomes independently [91].
These findings were contradicted by a 2016 study from the same group, treating 20 knee OA patients with an arthroscopic autologous ATMSC implantation. The study concluded that sex, BMI, size and location of the cartilage lesion were no independent risk factors for poor clinical outcomes [94]. Soler et al. described patients younger than 65 years with mild to moderate OA of the knee as the ideal candidates for intra-articular MSC application. In their trial from 2016, the group treated 15 patients with a mean age of 52 years with autologous BMMSCs and showed significant improvements in clinical scores and cartilage repair (MRI) after 12 months [74].
Koh et al. examined 30 knee OA patients, older than 65 years of age, who received intra-articular autologous ATMSC injections in 2015. Almost all patients showed significant improvements in KOOS, pain VAS and Lysholm scores at the final two-year follow-up. Cartilage status improved or was maintained in 87.5% of the patients two years after treatment and none of the patients underwent total knee arthroplasty during the two-year follow-up period [93]. These findings question the cut off for intra-articular MSC therapy at the age of 60, suggested by Kim at al. [91]. Finally, Chahal et al. showed that anti-inflammatory markers, assessed from autologous BMMSCs after harvesting, were strongly predictive of clinical outcomes, including WOMAC and KOOS scores [72]. This points towards potential valuable screening panels, that may be applied in the future. Such panels could help identify patients, who would benefit most from MSC-based treatment, and those who would not.

Conclusions and Outlook
MSC therapies have been successfully realized in a limited number of clinical trials, treating patients with OA. First impressions show that locally applied MSCs seem to halt disease progression, partially regenerate cartilage and alleviate pain. Yet, the existing clinical studies vary strongly in treatment protocols, levels of evidence and follow-up periods. Interpretation of these data demands a differentiated view and more scientific engagement is undoubtedly required to provide further evidence for the specific efficacy of MSC treatment in OA. Thus, we suggest keeping a substantial critical view on these innovative new therapeutics and propose to establish standards in documentation, clinical trial management and handling to ensure comparability across trials.
The vast majority of MSC studies for OA target the knee joint, most likely due to the high prevalence amongst all forms of OA. Yet, results from included studies do not allow a direct translation to other joints. Shoulder and hip joints for example have different biomechanics compared to the knee joint and therefore have to be analysed from that view. Whereas we are hoping for more randomized, double-blinded, multicentric, placebo-controlled phase III trials in knee OA, there is also a big need for phase I and II studies, examining other forms of OA.
According to the trials included in this review, MSC treatment for knee OA has shown only a few adverse effects, which were all likely due to the procedural nature of intra-articular injections and MSC harvesting procedures. However, the majority of safety studies provided only a short follow-up period or a limited sample size, and thus final safety conclusions remain difficult. Results indicate some efficacy in selected parameters over a duration of at least several months. The biggest drawback of the current clinical evidence for MSC-based therapy is that most of the included studies did not have resilient control groups, which reduces the level of evidence and makes results less reliable. Therefore, a detailed analysis and comparison of the costs and therapeutic alternatives for the daily clinical practice seem necessary. Most clinical trials include patients with mild to moderate OA, where MSCs are believed to work best by targeting the persisting low-grade inflammation. Patients with more severe OA proved not to be the ideal candidates for MSC-based treatments, as osteophytes, subchondral sclerosis and cysts are unlikely to be altered by MSCs. Therefore, MSCs cannot be understood as an alternative to arthroplasty surgery, but potentially as an additional therapy for early onset and mild to moderate cases of OA.
Although the mechanisms by which MSCs exert their effects on affected joints are not fully understood, it is generally accepted that immunomodulatory and anti-inflammatory properties on the one hand and repair and restoration mechanisms on the other hand, have a synergistic or additive effect on the osteoarthritic joint. Future research should put a critical focus on the emerging role and potential clinical application of the MSC secretome and its biologically active factors. The enhancement and modulation of MSCs' paracrine effects by priming may further increase their capacities to adequately alleviate pain and potentially restore joint integrity and function in OA-troubled patients.