Adipose Tissue-Derived Mesenchymal Stem Cells as a Potential Restorative Treatment for Cartilage Defects: A PRISMA Review and Meta-Analysis

Cartilage defects are a predisposing factor for osteoarthritis. Conventional therapies are mostly palliative and there is an interest in developing newer therapies that target the disease’s progression. Mesenchymal stem cells (MSCs) have been suggested as a promising therapy to restore hyaline cartilage to cartilage defects, though the optimal cell source has remained under investigation. A PRISMA systematic review was conducted utilising five databases (MEDLINE, EMBASE, Cochrane Library, Scopus, Web of Science) which identified nineteen human studies that used adipose tissue-derived MSC (AMSC)-based therapies, including culture-expanded AMSCs and stromal vascular fraction, to treat cartilage defects. Clinical, imaging and histological outcomes, as well as other relevant details pertaining to cartilage regeneration, were extracted from each study. Pooled analysis revealed a significant improvement in WOMAC scores (mean difference: −25.52; 95%CI (−30.93, −20.10); p < 0.001), VAS scores (mean difference: −3.30; 95%CI (−3.72, −2.89); p < 0.001), KOOS scores and end point MOCART score (mean: 68.12; 95%CI (62.18, 74.05)), thus showing improvement. The studies in this review demonstrate the safety and efficacy of AMSC-based therapies for cartilage defects. Establishing standardised methods for MSC extraction and delivery, and performing studies with long follow-up should enable future high-quality research to provide the evidence needed to bring AMSC-based therapies into the market.


Introduction
Cartilage, an important element of synovial joints, is composed of chondrocytes that lay down a highly organised extracellular matrix (ECM), consisting of water, type II collagen, glycosaminoglycans (GAGs) and proteoglycans, amalgamated into a dense collagenous network, which is responsible for its unique mechanical properties [1]. As well as crucial shock-absorbing and gliding properties that cartilage is well known for, cartilage ECM also has a role in chondrocyte homeostasis and cartilage phenotypic stability via chemokine signalling [2], and the synthesis of functional components of the skeletal system during embryogenesis [3]. Hence, any insults to cartilage integrity and the surrounding ECM will have a profound impact on chondrocytes, which in turn will lead to ECM composition changes, all in a vicious cycle [4].
Cartilage defect is a risk factor for osteoarthritis, and current therapies aim to relieve symptoms or prevent further degenerative changes to the articular cartilage. Cartilage damage is a key feature in degenerative diseases such as osteoarthritis (OA), with 26.6% of those aged 45 or above having a diagnosis of OA, which is projected to increase in the near

Search Algorithm
This systematic review was conducted in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines [43]. A comprehensive literature search from conception to July 2021 was conducted using the following databases: (1) Ovid MEDLINE(R) (1946 to Present with Daily and Weekly Update, Epub Ahead of Print, In-Process & Other Non-Indexed Citations), (2) Ovid Embase (1910 to Present), (3) SCOPUS, (4) Web of Science and (5) Cochrane Library. References of included studies will be searched as well as studies that cited any of the included studies. Equivalent combinations of the following search terms were used: "adipose tissue-derived" AND "mesenchymal stem cells" AND "cartilage regeneration". A detailed search strategy is available in Table S1. This review was prospectively registered in the International Prospective Register of Systematic Reviews PROSPERO (registration number: CRD42021267382, available from: https://www.crd.york.ac.uk/PROSPERO/display_record.php?RecordID=267382, accessed on 13 August 2021).
Title and abstract screening were independently performed by HM and VL, followed by full-text screening. A third reviewer (WK) was contacted for any unresolvable disagreements. A "snowball search" was performed whereby references of included studies, as well as studies that cited any of the included studies were independently searched by HM and VL.

Inclusion and Exclusion Criteria
Inclusion and exclusion criteria was created using the PICOS model (Population, Intervention, Comparison, Outcome, Study type) [44]. Detailed criteria are shown in Table 1.

Data Extraction
Data extraction was independently performed by HM and VL, with a third reviewer (WK) to resolve disagreements. Data were extracted into data tables created in a standardised excel spreadsheet for evidence synthesis and risk of bias analysis. Data from each study were split into five categories:

1.
Study characteristics, such as study design, level of evidence, outcome measures and duration of follow-up.

2.
Subject information such as subject model, mean age, mean BMI, percentage of female subjects and ethnicity. 3.
Intervention information, including method of AMSC administration, AMSC cell count.
Outcome measures that were pertinent to cartilage regeneration would also be extracted in detail. This would include clinical scores related to cartilage damage, imaging scores and histological cartilage repair scores (Numerical data were extracted corrected to 3 significant figures).

Data Analysis
Comparable quantitative measures were selected for data extraction and meta-analysis. In terms of clinical scores, this includes the Visual Analogue Scale (VAS), a psychometric scale that measures the amount of pain a patient feels across a continuum [45]. We have presented VAS scores from 0 to 10 where 10 represents the most pain. Data pertaining to knee injury and Osteoarthritis Outcome Score (KOOS), which is designed to evaluate the short-and long-term symptoms and function in patients with knee injury that can result in post-traumatic OA or in primary OA [46], were also included. We also included the KOOS score of its five different subgroups (1) Knee-related symptoms (Symptoms); (2) Pain; (3) Function in daily living (ADL); (4) Function in Sports and Recreation (Sport & Rec); (5) and Knee-related quality of Life (QoL). Scores are presented on a 0 to 100 scale where 0 represents knee problems and 100 represents no knee problems. Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) scores were also selected for data extraction and analysis. WOMAC encompasses three main groups including pain, stiffness and function domains and was designed to evaluate OA specifically. WOMAC is expressed on a 0 to 100 scale where 100 means the worst pain, stiffness and physical function [47]. Regarding imaging outcomes, Magnetic Resonance Observation of Cartilage Repair Tissue (MOCART scores) were also selected for meta-analysis since it is a comparable quantitative outcome measure used in multiple studies [48]. MOCART score is an MRI score used for post-operative assessment of repaired cartilage in regards to its adjacent native hyaline cartilage in terms of morphology and signal intensity. It comprises of nine pertinent variables, including (1) the degree of filling of the defect; (2) the integration to the border zone; (3) the description of the surface; (4) description of the structure; (5) the signal intensity; (6) the status of the subchondral lamina; (7) the status of the subchondral bone; (8) the appearance of adhesions; (9) the presence of synovitis. This allows for the quantification of post-operative morphological MRI results and evaluation of repair, comparing the repaired tissue with adjacent cartilage [49]. Final MOCART scores of studies were extracted as the studies conducted MOCART scoring at different time points.
Forest plots were conducted using RStudio (metafor package [50]) with the extracted VAS, KOOS, WOMAC and MOCART results. We provided each subgroup with Cochran's Q and the I 2 statistic as measures of heterogeneity to account for the variability in effect size and proportion of variance attributable to study heterogeneity, p-values for heterogeneity were shown in forest plots, while p-values for overall effects were reported in text. A random effects model is adopted in the assessment of the aforementioned scores since the results are heterogenous (I 2 ≥ 50).
Qualitative results or other results that were not selected for meta-analysis were extracted and presented in a table, as well as discussed in depth in the text. These were divided into clinical outcomes, imaging outcomes, histological outcomes and others.

Assessing Risk of Bias
Quality assessment was carried out independently by HM and VL using Cochrane's RoB 2.0 tool for randomized trials, and Cochrane's ROBINS-I tool for non-randomized trials. Disagreements were consulted with a third reviewer (WK).
Cochrane's RoB 2.0 tool comprises five domains: (1) randomisation process; (2) deviations from intended interventions; (3) missing outcome data; (4) measurement of the outcome; (5) selection of the reported results [51]. Each domain was assessed as low risk, some concerns or high risk. We classified the overall risk of bias of each randomized study according to the RoB protocol.
Cochrane's ROBINS-I tool comprises of seven domains: (1) bias due to confounding; (2) in selection of participants into the study; (3) in classification of interventions; (4) due to deviations from intended deviations; (5) due to missing data; (6) in measurement of outcomes; (7) in selection of reported result [52]. Each domain was assessed as Low, Moderate, Serious and Critical risk of bias. The summary of the results of both RoB and ROBINS-I were presented in a graph showing overall results stratified by their respective domains.

Search Results
We identified 3146 records after our literature search across the MEDLINE, EMBASE, Scopus, Web of Science and the Cochrane library, with 1794 records left after removal of duplicates ( Figure 1). After screening by titles and abstracts with general criteria, 35 records were selected for full-text review. Nineteen records were selected for qualitative analysis after application of selection criteria.
We identified 3146 records after our literature search across the MEDLINE, EMBASE, Scopus, Web of Science and the Cochrane library, with 1794 records left after removal of duplicates ( Figure 1). After screening by titles and abstracts with general criteria, 35 records were selected for full-text review. Nineteen records were selected for qualitative analysis after application of selection criteria.

Characteristics of Selected Studies
Details of study design and intervention are summarized in Table 2. There are six randomized trials and 10 non-randomized trials, including retrospective and prospective cohort studies as well as case series, four were dose-escalating studies. Two studies were on the same cohort of patients and were treated as a single trial [54,55]. Regarding the intervention, two studies utilized AMSCs from the infrapatellar fat pad, while other studies used adipose tissue of diverse origins including from the abdominal and gluteal region. Fifteen studies administered AMSCs via injection, while three studies administered AMSC by implantation. Three studies administered AMSC-based therapy twice [56][57][58]. Five studies used stromal vascular fraction as opposed to culture-expanded AMSCs. In eight studies, AMSC administration was also concurrent with other therapies, such as microfracture, arthroscopic debridement or injection of platelet-rich plasma (PRP) [56,57,[59][60][61][62][63]. Cell count of AMSCs ranges from 10 6 to 10 8 . Participant details of the included studies can be seen in Table 3. Three studies included patients with cartilage defects of the knee for a total of 73 patients, while 15 studies included patients with knee osteoarthritis for a total of 368. Outcome details and results of the included studies are summarized in Table  4. The duration of follow-up ranged from 6 to 36 months. Studies evaluated outcome measures in three major categories, clinical outcomes, imaging outcomes and histological outcomes.

Characteristics of Selected Studies
Details of study design and intervention are summarized in Table 2. There are six randomized trials and 10 non-randomized trials, including retrospective and prospective cohort studies as well as case series, four were dose-escalating studies. Two studies were on the same cohort of patients and were treated as a single trial [54,55]. Regarding the intervention, two studies utilized AMSCs from the infrapatellar fat pad, while other studies used adipose tissue of diverse origins including from the abdominal and gluteal region. Fifteen studies administered AMSCs via injection, while three studies administered AMSC by implantation. Three studies administered AMSC-based therapy twice [56][57][58]. Five studies used stromal vascular fraction as opposed to culture-expanded AMSCs. In eight studies, AMSC administration was also concurrent with other therapies, such as microfracture, arthroscopic debridement or injection of platelet-rich plasma (PRP) [56,57,[59][60][61][62][63]. Cell count of AMSCs ranges from 10 6 to 10 8 . Participant details of the included studies can be seen in Table 3. Three studies included patients with cartilage defects of the knee for a total of 73 patients, while 15 studies included patients with knee osteoarthritis for a total of 368. Outcome details and results of the included studies are summarized in Table 4. The duration of follow-up ranged from 6 to 36 months. Studies evaluated outcome measures in three major categories, clinical outcomes, imaging outcomes and histological outcomes.    The majority of studies reported an improvement in clinical scores assessed using different scoring systems after the administration of AMSCs. In terms of pain symptoms, nine studies reported an improvement as assessed by VAS [54,55,58,59,61,[64][65][66]69,72] and two studies reported improvement assessed by the Numeric Pain Rating Scale (NPRS) [56,57]. Two studies showed an improvement in overall patient health as assessed by the 36-Item Short Form Survey (SF-36) [58,70].
AMSCs were able to improve clinical outcomes as assessed by knee health indices, eight studies reported an improvement in KOOS [54][55][56][57]63,64,[66][67][68], two studies reported an improvement in their Knee Society Score (KSS) [54,55,69], three studies reported an improvement in the International Knee Documentation Committee (IKDC) scores [60,62,64] and a further four studies reported an increase in knee health as assessed by the Tegner Activity Scale and Lysholm Knee Questionnaire [60][61][62]69]. One study reported an improvement in the Hospital for Special Surgery knee score (HSS-KS) [69] and one other study reported an improvement in Range of Motion testing [59].
For Osteoarthritis specific evaluation, 11 studies utilized the WOMAC scale and assessed the condition of osteoarthritic patients, all of which reported improvement after AMSC-based therapy.
Quantitative analyses also corroborate the aforementioned results. Seven studies reported VAS in a comparable manner [54,58,59,61,66,69,72]. The study of Jo et al. was a dose-escalating study and the results are included as different entries [54,55]. In Lu et al., bilateral osteoarthritis patients were included and the results were reported as left and right knee cohorts, these were included separately [58]. A random effects meta-analysis revealed a significant improvement in VAS scores at end point (mean difference: −3.30; 95% CI (−3.72, −2.89), p < 0.001), comparing the baseline of the same cohort of participants ( Figure 2).

Improvement of Clinical Outcomes
The majority of studies reported an improvement in clinical scores asses different scoring systems after the administration of AMSCs. In terms of pain s nine studies reported an improvement as assessed by VAS [54,55,58,59,61,64 and two studies reported improvement assessed by the Numeric Pain Rating Sca [56,57]. Two studies showed an improvement in overall patient health as asses 36-Item Short Form Survey (SF-36) [58,70].
For Osteoarthritis specific evaluation, 11 studies utilized the WOMAC sca sessed the condition of osteoarthritic patients, all of which reported improve AMSC-based therapy.
Quantitative analyses also corroborate the aforementioned results. Seven ported VAS in a comparable manner [54,58,59,61,66,69,72]. The study of Jo e dose-escalating study and the results are included as different entries [54,55]. I bilateral osteoarthritis patients were included and the results were reported a right knee cohorts, these were included separately [58]. A random effects met revealed a significant improvement in VAS scores at end point (mean differe 95% CI (−3.72, −2.89), p < 0.001), comparing the baseline of the same cohort of p ( Figure 2). Four studies reported comparable KOOS scores, and random effects met was conducted on each of the KOOS subgroup results [54,56,57,63]. Again, Jo e dose-escalating study and its three cohorts (low dose, medium dose, high dose cluded separately [54,55]. Pooled analysis shows that there was a significant imp in all five KOOS subgroups (Figure 3). These include KOOS Symptoms (mean  Four studies reported comparable KOOS scores, and random effects meta-analysis was conducted on each of the KOOS subgroup results [54,56,57,63]. Again, Jo et al. was a doseescalating study and its three cohorts (low dose, medium dose, high dose) were included separately [54,55]. Pooled analysis shows that there was a significant improvement in all five KOOS subgroups (  Six studies reported WOMAC results in a comparable method and were eligible for quantitative analysis [54,58,61,66,67,70,72]. Three studies with WOMAC results were dose-escalating studies and were again included in the meta-analysis separately respective to their dosage [54,67,70]. Random effects meta-analysis of the results revealed that after administration of AMSCs, a significant improvement at the end point of the study was found in WOMAC results (mean difference: −25.52; 95% CI (−30.93, −20.10), p < 0.001) when compared with the baseline of the same cohort ( Figure 4). Six studies reported WOMAC results in a comparable method and were eligible for quantitative analysis [54,58,61,66,67,70,72]. Three studies with WOMAC results were doseescalating studies and were again included in the meta-analysis separately respective to their dosage [54,67,70]. Random effects meta-analysis of the results revealed that after administration of AMSCs, a significant improvement at the end point of the study was found in WOMAC results (mean difference: −25.52; 95% CI (−30.93, −20.10), p < 0.001) when compared with the baseline of the same cohort ( Figure 4).  The combined results show that AMSCs are effective in improving clinical of focal cartilage defects, from pain symptoms to knee health scores as well as o tis symptoms.

Improvement of Imaging Outcomes
AMSC amelioration of focal cartilage defects is further supported by the M ing outcomes by 14 studies. From MRI investigations, three studies showed ca generation by increased cartilage volume or depth [58,68,70]. While one othe ported significantly decreased cartilage defect depth [54,55], another study rep significant cartilage defect change with AMSC administration compared with cartilage defect increase in patients without AMSC administration [66]. Standa aging assessment systems were also utilized. Two studies utilized the Whole-O netic Resonance Imaging Score (WORMS), with one reporting significant imp and the other reporting non-significant changes [59,71].
Cartilage quality and composition was also evaluated in an MRI investiga either T1rho mapping, T2 cartilage mapping or with delayed gadolinium-enha of cartilage (dGEMRIC) by four studies [56,57,67,71]. Results showed an increa lage condition from improved glycosaminoglycan/proteoglycan content as we lage maturation.
Seven studies evaluated cartilage regeneration with the MOCART scor studies showing significant progressive improvement or improvement when with control [56,57,59,63,64,69,72]. From the seven studies, five were comparab dom effects meta-analysis [  The combined results show that AMSCs are effective in improving clinical outcomes of focal cartilage defects, from pain symptoms to knee health scores as well as osteoarthritis symptoms.

Improvement of Imaging Outcomes
AMSC amelioration of focal cartilage defects is further supported by the MRI imaging outcomes by 14 studies. From MRI investigations, three studies showed cartilage regeneration by increased cartilage volume or depth [58,68,70]. While one other study reported significantly decreased cartilage defect depth [54,55], another study reported non-significant cartilage defect change with AMSC administration compared with significant cartilage defect increase in patients without AMSC administration [66]. Standardized imaging assessment systems were also utilized. Two studies utilized the Whole-Organ Magnetic Resonance Imaging Score (WORMS), with one reporting significant improvement and the other reporting non-significant changes [59,71].
Cartilage quality and composition was also evaluated in an MRI investigation using either T1rho mapping, T2 cartilage mapping or with delayed gadolinium-enhanced MRI of cartilage (dGEMRIC) by four studies [56,57,67,71]. Results showed an increase in cartilage condition from improved glycosaminoglycan/proteoglycan content as well as cartilage maturation.
Seven studies evaluated cartilage regeneration with the MOCART score, with all studies showing significant progressive improvement or improvement when compared with control [56,57,59,63,64,69,72]. From the seven studies, five were comparable for random effects meta-analysis [57,59,63,69,72]. Pooled results of MOCART scores taken at the end point show the presence of cartilage regeneration (mean: 68.12; 95% CI (62.18, 74.05); p < 0.001) ( Figure 5). Two studies also conducted alternative imaging investigations in the form raphy and plain radiography. Sonographic investigations showed improved c integrity of the soft tissue-cartilage but no increase in cartilage depth [65], w radiography showed neither improvement nor further degradation of cartilage

Improvement of Histological Outcomes
Four studies evaluated cartilage regeneration histologically. Two studies using the International Cartilage Repair Society (ICRS) II histologic score. Thr showed hyaline-like characteristics in the regenerated cartilage, such as the p safranin O and type II collagen in the regenerated cartilage [54,63,64]. One stud stem cell graft-like sheet of stem cells [67].

Other Improvements
One study evaluated the molecular profile of synovial fluid. The catabol profile was altered with decreased metalloproteinase 2 (MMP2) and increased i growth factor type 1 (IGF1). There was also a decrease in pro-inflammatory cyto β, IL6 and IL8) and an increase in anti-inflammatory cytokines (IL10) in the syn [65].

Assessment of Methodological Bias
Overall, the risk of bias in the randomized studies was low (Figure 6a). source of bias was from the measurement of outcome, and this is due to outcome being measured and reported in the form of questionnaires.
For non-randomized studies, overall, there a is moderate risk of bias (Figu major source of bias was from bias from the measurement of outcomes. As the p were not blinded and the measurement outcomes were mostly questionnaires high risk of bias in terms of measurement. Two studies also conducted alternative imaging investigations in the form of sonography and plain radiography. Sonographic investigations showed improved clarity and integrity of the soft tissue-cartilage but no increase in cartilage depth [65], while plain radiography showed neither improvement nor further degradation of cartilage [69].

Improvement of Histological Outcomes
Four studies evaluated cartilage regeneration histologically. Two studies evaluated using the International Cartilage Repair Society (ICRS) II histologic score. Three studies showed hyaline-like characteristics in the regenerated cartilage, such as the presence of safranin O and type II collagen in the regenerated cartilage [54,63,64]. One study found a stem cell graft-like sheet of stem cells [67].

Other Improvements
One study evaluated the molecular profile of synovial fluid. The catabolic/anabolic profile was altered with decreased metalloproteinase 2 (MMP2) and increased insulin-like growth factor type 1 (IGF1). There was also a decrease in pro-inflammatory cytokines (IL1 β, IL6 and IL8) and an increase in anti-inflammatory cytokines (IL10) in the synovial fluid [65].

Assessment of Methodological Bias
Overall, the risk of bias in the randomized studies was low (Figure 6a). The main source of bias was from the measurement of outcome, and this is due to outcome measures being measured and reported in the form of questionnaires.
For non-randomized studies, overall, there a is moderate risk of bias (Figure 6b). The major source of bias was from bias from the measurement of outcomes. As the participants were not blinded and the measurement outcomes were mostly questionnaires, there is a high risk of bias in terms of measurement.
Detailed breakdown of the Risk of Bias analysis is available in Table S2. Detailed breakdown of the Risk of Bias analysis is available in Table S2.

Discussion
MSC-based therapies for cartilage defects induce hyaline-like cartilage regeneratio and, therefore, have the potential to improve clinical outcomes in patients with cartilag defects. This disease-modifying approach is vastly different to conventional palliativ therapies. Our primary findings show that in human studies, after the administration o AMSCs, clinical, histological or imaging outcome measures were improved, althoug there is variation in the type of study included, source and type of MSCs and method o administration. Included studies can be divided into randomized controlled trials an non-randomized trials, such as case series, retrospective studies and single cohort studie

Optimal Source of Stem Cell
AMSCs are an increasingly attractive source for MSCs. A lot of progress on MSC based cell therapy was achieved using bone marrow-derived MSCs (BMSCs), which wa the dominant source of MSCs [73], however, the recent literature has expanded to includ ing MSCs derived from almost all human tissue, including pluripotent stem cells. Amon numerous options, adipose tissue has emerged as a dependable and rich source of MSC with regards to increased quantity, higher yield, lack of ethical issues and ease of harves with minimally invasive procedures such as liposuction [74]. In vivo studies show tha the differentiation potential of AMSCs is less attenuated by age when compared wit BMSCs [75], and AMSCs have better immunosuppressive function [40]. AMSCs also hav higher proliferation potential according to growth curve, cell cycle and telomerase activit analyses [39], although other studies have suggested that the differentiation potential o both types of MSCs are comparable [28,38,76]. Overall, while the exact properties of AM SCs compared with other MSCs have yet to be ascertained, AMSC-based therapies ma

Discussion
MSC-based therapies for cartilage defects induce hyaline-like cartilage regeneration and, therefore, have the potential to improve clinical outcomes in patients with cartilage defects. This disease-modifying approach is vastly different to conventional palliative therapies. Our primary findings show that in human studies, after the administration of AMSCs, clinical, histological or imaging outcome measures were improved, although there is variation in the type of study included, source and type of MSCs and method of administration. Included studies can be divided into randomized controlled trials and non-randomized trials, such as case series, retrospective studies and single cohort studies.

Optimal Source of Stem Cell
AMSCs are an increasingly attractive source for MSCs. A lot of progress on MSC-based cell therapy was achieved using bone marrow-derived MSCs (BMSCs), which was the dominant source of MSCs [73], however, the recent literature has expanded to including MSCs derived from almost all human tissue, including pluripotent stem cells. Among numerous options, adipose tissue has emerged as a dependable and rich source of MSCs, with regards to increased quantity, higher yield, lack of ethical issues and ease of harvest with minimally invasive procedures such as liposuction [74]. In vivo studies show that the differentiation potential of AMSCs is less attenuated by age when compared with BMSCs [75], and AMSCs have better immunosuppressive function [40]. AMSCs also have higher proliferation potential according to growth curve, cell cycle and telomerase activity analyses [39], although other studies have suggested that the differentiation potential of both types of MSCs are comparable [28,38,76]. Overall, while the exact properties of AMSCs compared with other MSCs have yet to be ascertained, AMSC-based therapies may be preferable to other MSC sources due to ease of harvest and having fewer ethical hurdles to overcome, which would aid its expansion in the healthcare market.
In our included studies, AMSCs were either derived from adipose tissue such as gluteal or abdominal adipose tissue or the infrapatellar fat pad (IFPFs). The ideal source for AMSCs is still a subject under research, however, IFPFs are an attractive source compared with adipose tissue since IFPF is usually obtained during the resection of inflamed tissue in knee arthroscopy [77]. IFPF-MSCs also display similar surface markers compared with other cells around the knee joint, and may reduce immunologic rejection [78]. Ding et al. suggested that IFPF-derived AMSCs may have higher proliferative potential than AMSCs derived from abdominal fat in vitro, although in vivo studies are needed to support this finding [78].
AMSCs were administered as culture-expanded AMSC or SVFs. Studies characterized the injected AMSCs by cell count. As mentioned previously, SVFs and AMSCs contain different minimal defining criteria. Research has shown that culture-expanded AMSCs result in better therapeutic potential owing to the larger number of MSCs as well as the higher potential to generate more trophic factors [79,80]. An in vivo study compared culture-expanded AMSCs with SVF regarding its use for osteoarthritis in sheep and reported better imaging, macroscopic and immunohistochemistry outcomes when AMSCs were used [81]. However, the culture expansion of MSCs resulted in lower migration and homing ability [82], and, therefore, culture-expanded AMSCs may require surgery to expose the site of lesion. Furthermore, SVFs that are minimally manipulated may be favoured for economic or regulatory reasons [83]. More comparative studies would be required to define the superior treatment for cartilage defects.
Among our included studies, Zhao et al. evaluated the use of allogeneic stem cells as opposed to autologous stem cells [71]. Allogeneic stem cells were obtained from three healthy donors and preclinical toxicity and chronic tumorigenicity were evaluated in vivo with no adverse events reported. Allogeneic MSC transplants have been previously demonstrated in animals as well as in human clinical trials, again with no adverse events reported [84,85]. This may be due to the low immunogenicity of MSCs as well as the immune-privileged character of cartilage tissue.
No severe adverse events were reported in our studies, and the safety of AMSC-based therapies has been previously reviewed in the literature, showing no permanent adverse effects [86]. Nevertheless, further research could evaluate the risks of harvesting adipose tissue for AMSC-based therapy via liposuction, investigate the interaction of AMSCs with the tumour microenvironment and evaluate the long-term safety of AMSC injection for cartilage defects with large sample size studies. The optimal source of adipose tissue for harvesting AMSCs is still under debate, and further studies are needed to ascertain the best form of AMSC-based therapies, i.e., culture-expanded AMSCs or SVFs, the use of allogeneic or autologous stem cells, as well as long-term safety.

Augmenting the Function of AMSCs
Despite the benefits of AMSCs, the outcomes of advanced clinical trials have sometimes fallen short of expectations [87]. This could be due to the vast dimensions across which MSC heterogeneity is present, such as among donors, tissue sources and even cell subpopulations with the same origin [88], and studies showing that MSCs disappeared after 24 h post-infusion [89]. It is, hence, prudent to ensure that studies utilising AMSCs adhere to well-defined international guidelines such as the International Federation for Adipose Therapeutics and Science (IFATS) [41] while pursuing novel avenues to artificially boost AMSC potency to overcome the shortcomings of naïve AMSCs.
AMSCs can be pre-conditioned in vitro, which prepares it for the harsh microenvironment of the host, enhances its migration to its site of action, or enhances its biological properties. Given that synovial inflammation plays a key role in the pathogenies of osteoarthritis, and is associated with cartilage destruction and pain, enhancing the immunomodulatory function of AMSCs has gained a lot of attention [90]. Pre-treating AMSCs with pro-inflammatory stimuli increases their immunosuppressive and anti-inflammatory potential by reducing NF-κB activity and promoting macrophage differentiation into the M2 anti-inflammatory phenotype [91]. Studies have shown that AMSCs express toll-like receptors (TLRs), which can be exploited via specific TLF-agonist engagement to affect their migration, secretion of immune modulating factors and induce a change in cell fate [92]. Furthermore, IFNγ-stimulated MSCs can enhance chondrogenesis [91]. However, the degree to which differentiation is affected depends on the MSC source, despite ligation of the same TLR, is under investigation [93], and future, in-depth comparative studies are needed to determine the best MSC source for chondrogenic differentiation in this scenario.
AMSCs are also amenable to genetic modification and may be adequate vehicles for gene delivery [94]. Targeted overexpression of microRNAs have produced favourable clinical outcomes [95,96]. miR-302 transfection increased proliferation and inhibited oxidantinduced cell death in AMSCs, and can be used to enhance the therapeutic efficacy of AMSCs in vivo [97]. However, studies have shown that genetic manipulation can affect differentiation potential, although results are conflicting and the effect on chondrogenic differentiation is unknown [98]. As described above, the immunomodulatory potential of AMSCs to promote cartilage repair is attractive, and genetic engineering has the potential to improve immunomodulation. CTLA4Ig-overexpressing AMSCs have been shown in a mouse model to protect against cartilage destruction and ameliorated severe rheumatoid arthritis [99]. Nevertheless, there is a risk of tumorigenicity and immunogenicity [100], and studies that examine the long-term clinical outcomes of using genetically modified AMSCs at a local and systemic level have yet to be published.
The niche microenvironment is crucial for stem cell integrity, fate and behaviour, and the regulation between an active and quiescent state [101]. Culture conditions must be optimised and can increase the biological potency of AMSCs. Conventional practice involves growing MSCs on a two-dimensional system as monolayers. This is artificial and lacks the key cell-cell and cell-extracellular matrix contact present in vivo. Three-dimensional cultured MSCs have shown superior expansive and differentiation potential, such as undergoing large-scale in vitro chondrogenic differentiation and enhanced in vivo cartilage formation in an animal model [102]. MSC aggregation into spheroids can also augment their immunomodulatory ability, as shown in a mouse model of cartilage damage [103]. Nevertheless, with the increased complexity of three-dimensional culture, it is necessary to regulate the time duration of the culture and size of spheroids formed [104]. Furthermore, studies have reported that AMSCs cultured under hypoxic (2% oxygen) conditions enhanced early chondrogenic differentiation, while decreasing osteogenesis, thus, favoured chondrocyte formation [105].
Finally, AMSC therapy can be combined with concomitant application after other bioactive molecules, or performed with other conventional procedures. Two studies injected AMSCs or SVF with PRP [61,68]. In vitro research has suggested that PRP optimizes MSC-based therapy by stimulating MSC proliferation, migration and immune modulation without affecting differentiation potential [106]. PRP may also optimize MSC-based therapy for cartilage regeneration by enhancing chondrogenic differentiation [107]. However these studies did not investigate the effect of PRP alone, making it hard to determine the actual role PRP has on chondrogenesis when administered together with AMSCs. Koh et al. compared AMSCs concurrent with microfracture therapy with just microfracture therapy [63] and reported improved radiologic and KOOS pain and function scores when AMSCs were administered together with microfracture. Nevertheless, only large cartilage defects (≥3 cm 2 ) were investigated, and future research is needed to investigate the effects of AMSC and microfracture co-therapy for lesions of different sizes.

Methods of AMSC Administration
Methods of AMSC administration varied greatly in the included studies, though most studies favoured an intra-articular injection of AMSCs or SVF. Three studies investigated the use of implantation of AMSCs, none of which reported any adverse events [60,62,64].  [60], suggesting that in the implantation group, ICRS scores and clinical scores such as IKDC and Tegner scores were higher than those in the injection group. The article suggests that the cartilage regenerated by implantation has greater durability than that of injection. This may also be due to better cell retention and survival at the lesion site in implantation compared with injection.
Some studies also compared the use of AMSC therapy and other pre-existing therapies for focal cartilage defects. Hong et al. compared SVF with the injection of hyaluronic acid [59], while Lu et al. compared the AMSC injection against the injection of hyaluronic acid [58]. Both studies reported significant improvement in the AMSC or SVF group when compared with the control group with only the injection of hyaluronic acid. However, no study compared the efficacy of AMSC with SVF. Studies investigating cellular therapies for knee osteoarthritis suggested that iatrogenic complication may be higher in SVF groups, with more frequent knee effusion (SVF 8%, AMSC 2%) [108] and minor complications related to the fat harvest site (SVF 34%, AMSC 5%). However, clinical outcomes such as pain VAS improved earlier and to a greater degree in the AMSC group than VAS group.
Two included studies utilised arthroscopies prior to AMSC injection. Jo et al. only used arthroscopy to examine the patient and guide AMSC administration [54,55]. However, Zhao et al. performed arthroscopic debridement, which could lead to bias [71]. Arthroscopic debridement removes inflammatory synovial fluid, which impedes AMSC adhesion but also enhances the immunomodulatory profile of the AMSC secretome. Further research is needed to determine the effect of arthroscopy on the outcomes of AMSC treatment.

Improved Clinical Outcomes
All studies concluded that AMSC administration correlates with an improvement in pain and functional outcomes. Pooled analyses of WOMAC scores showed a statistically significant improvement in all studies across all follow-up times, regardless of the dose of SVF or AMSC used. This suggests that disease modification is long term, rather than simply acting as short-term analgesics, hence avoiding the need for the repeated administration of MSC therapy. However, a meta-analysis that directly compares SVF and AMSC treatments therapies could not be performed due to the low number of studies; furthermore, heterogeneity in the studies precluded any subgroup analysis of AMSC or SVF therapy.
Four studies segregated their cohort to receive AMSCs at different doses, namely low, medium and high. Three studies reported a dose-dependent effect on clinical outcomes, with the group receiving a high AMSC dose having the greatest reduction in cartilage defects and best clinical improvements [54,70,71]. Despite this suggesting a relationship between the number of AMSCs administered and therapeutic effect, Pers et al. found that only the group that received a low AMSC dose experienced significant improvements in pain levels and function compared to baseline [67]. This could be explained by the fact that significant synovial inflammation was present in the low dose cohort, and, as stated previously, MSCs can be primed by an inflammatory environment to exert their immunomodulatory effects [91]. This could lead to lower costs and the increased speed of AMSC harvesting, increasing their appeal in clinical practice. Due to the low number of studies investigating the outcomes of AMSC doses, subgroup analyses were not performed. Further research is needed on the dose-dependent relationship between AMSC therapy and long-term clinical outcomes, and, perhaps, studies that investigate the difference between one low dose and multiple low dose AMSC therapy, rather than giving one large bolus dose.
Joint pain and loss of functionality are the major symptoms of osteoarthritis. In the present study, VAS scores and other pain-related outcomes have been shown to improve after AMSC administration. There is an improvement in KOOS scores as well, which suggests an improvement in both symptoms and functions and also in the patient's quality of life. Improvement in such clinical parameters suggests the therapeutic potential of using AMSCs over conventional therapies for focal cartilage defects. The therapeutic potential of MSCs on osteoarthritis have been previously described [109,110]. In the present study, AMSCs are also shown to improve cartilage regeneration as well as OA symptom scores. Despite macroscopic appearance scoring being predictive of histological scoring [111], more studies should be performed to evaluate the correlation between macroscopic and histological appearance with functional improvement.
Most studies did not stratify their patients based on the severity of osteoarthritis. Given that an inflammatory environment encourages AMSCs to bring out their immunomodulatory effects, they may be most effective during end stage osteoarthritis, given that inflammation levels are highest as the disease progresses. The current literature remains divided, with Nguyen et al. suggesting better efficacy in patients with less severe osteoarthritis [112], while Tran et al. reported greatly reduced WOMAC scores 24 months post-treatment in patients with Kellgren-Lawrence (K-L) grade 3 than K-L grade 2 [113].
Additionally, studies could have stratified patients based on age or BMI. Studies have shown that ageing negatively impacted AMSC proliferation and reduced its chondrogenic differentiation ability in favour of adipogenic differentiation [114]. Obesity has been shown to be a key risk factor for osteoarthritis, however, the biological role of adipose-derived inflammation on MSC efficacy have yet to be investigated [115].

Hyaline-Like Cartilage Regeneration
The underlying mechanism of AMSCs amelioration of cartilage defects lie in the potential of AMSCs to generate hyaline-like cartilage.
Previous in vitro studies have exhibited the chondrogenic differentiation ability of AMSCs with successful differentiation indicated by the immunohistological staining of type II collagen or expression of glycosaminoglycan in the regenerated tissue similar to hyaline cartilage [116][117][118]. Animal studies have also corroborated with in vitro results and evaluated the safety of the administration of MSCs [109,119]. Bone marrow-derived MSCs and synovium-derived MSCs have been shown to induce hyaline-like cartilage regeneration confirmed by biochemical results such as improved GAG content and histological results related to type II collagen expression and integration. Early human AMSC studies, such as Pak et al., showed probable cartilage regeneration in the knee joint after AMSC administration [120]. In our included studies, histological and imaging outcomes as well as biomarker analysis shows the presence of hyaline-like cartilage regeneration. Koh et al. performed histological staining and showed that participants receiving AMSC administration in concurrence with microfracture exhibited a higher degree of staining for safranin O and type II collagen than patients who received microfracture alone, suggesting the presence of hyaline cartilage regeneration after AMSC administration [63]. For imaging analysis, the MOCART tool assesses regenerated cartilage compared to its similarity in terms of morphology and signal intensity with the surrounding native hyaline cartilage. The improved MOCART score post AMSC administration therefore suggests the presence of hyaline-like properties of the regenerated cartilage. Aside from MOCART, Freitag et al. utilized additional T2-weighted mapping to evaluate the quality of the regenerated cartilage and showed progressive cartilage maturation over time [56,57]. These results point towards the ability of AMSCs to generate hyaline-like cartilage in humans.
Nevertheless, with an average follow-up time of 18.3 months, the long-term effects of AMSC therapy on cartilage repair are unknown. Jo et al. reported that cartilage degeneration occurred after two years post-treatment, perhaps due to desensitisation of the knee joint to AMSC therapy, or simply the need for another AMSC injection [54]. Park et al. provided more optimistic results, with all cartilage regeneration parameters remaining stable over a seven year follow-up period [121]. There is a need in the literature for more studies with long follow-up times to determine the long-lasting effects of AMSC therapy.

Strengths and Limitations
Our study has several strengths, including an extensive search strategy, robust inclusion and exclusion criteria, and thorough data extraction in the form of both quantitative measures and qualitative outcomes, with meta-analysis of quantitative outcomes. However, the included studies in the present review show a high degree of heterogeneity from different methods of administration, harvesting, type of MSC-based therapy, duration of follow-up as well as the disease state of participants. Furthermore, the different methods of reporting mean that we were only able to conduct meta-analyses on quantifiable outcomes reported in a comparable manner, creating a bias towards studies with similar quantifiable outcomes. WOMAC and VAS are non-specific scoring systems, and it may be better to create a novel patient-reported outcome measure specifically for patients treated with MSC-based therapies. Risk of bias analysis also shows a high risk of measurement bias due to most of the clinical outcome measures being in the form of self-reported questionnaires. There is an especially high risk of bias in the measurement of the outcomes in non-randomized and non-blinded studies.

Conclusions
MSC-based therapies hold high potential for restorative treatment of cartilage defects, and, in recent years, there has been an interest in the use of AMSCs due to their ease of collection and abundance. This systematic review and meta-analysis compiled the findings in human studies for the administration of AMSC-based therapies for amelioration of cartilage defects. The evidence suggests that there is an improvement in clinical, imaging and histological outcomes after AMSC or SVF administration with no severe adverse events reported. Despite heterogeneity in studies included, there is evidence supporting the use of AMSCs or SVFs for focal cartilage defects. We recommend researchers establish the roles of biochemical components that stimulate cartilage repair after AMSC therapy, as well as making chondrocyte gene expression, cartilage macroscopic appearance and histological scores important outcome measures, in addition to functional and clinical outcomes. Establishing the most efficient and safest method for MSC extraction, culture and delivery, and performing studies with long follow-up times to determine the lasting implications should enable future high-quality research to provide the evidence needed to bring AMSC-based therapies into the market to tackle major public health challenges such as OA.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/ph14121280/s1, Table S1: Detailed search strategy, Table S2A: Breakdown of risk of bias analysis using RoB 2.0 tool of randomized studies, Table S2B: Breakdown of risk of bias analysis using ROBINS-I tool of non-randomized studies Table S3: PRISMA checklist.

Conflicts of Interest:
The authors declare no conflict of interest.