Discussion

This meta-analysis of 38 preclinical studies provides robust quantitative evidence that mesenchymal stem cell-derived extracellular vesicles (MSC-EVs) substantially attenuate osteoarthritic cartilage degeneration. In human-derived EVs, the pooled MD for OARSI histological scores was -3.27 (95 % CI: -4.66 to -1.88). Animal-derived EVs demonstrated an even more profound structural rescue, yielding an MD of -5.58 (95 % CI: -7.13 to -4.03). Importantly, the magnitude of these effects is clinically meaningful when interpreted against the global burden of disease: 595 million individuals (≈ 7.6 % of the world’s population) lived with osteoarthritis (OA) in 2020, with case numbers increasing by approximately 132 % since 1990 [1]. Rigorous publication-bias assessments support the validity of our pooled estimates: Trim-and-Fill analyses estimated zero missing studies across both datasets, and Rosenthal Fail-Safe N values of 2058 (human) and 813 (animal) establish that these therapeutic effect sizes are highly resistant to the file-drawer problem.

Despite these robust efficacy signals, the pervasive heterogeneity observed across studies (_I_2 > 84 %) presents a critical translational barrier that must be examined through a pharmacokinetic (PK) and ADMET lens. EV therapies in preclinical OA models currently suffer from unstandardized dosing regimens arbitrarily defined by either protein concentration (ranging from 0.25 to 500 μg mL-1) or particle count (8×107 to 1010 particles). As our meta-regression indicated, arbitrary dosage metrics alone fail to significantly predict treatment efficacy. This is consistent with established intra-articular (IA) PK behaviour: small-molecule drugs and protein biologics injected into the synovial cavity are typically cleared with half-lives of only a few hours, owing to rapid synovial-fluid turnover and extensive lymphatic and venous drainage [37]. Free MSC-EVs delivered IA face the same clearance pressures, compounded by macrophage uptake within the synovial lining. Reporting a dose without defining the joint retention time, cumulative exposure (AUC), or biodistribution is therefore pharmacologically incomplete and obscures meaningful comparisons between studies.

Subgroup analyses revealed distinct pharmacodynamic profiles based on EV cellular origin. While bone marrow-derived EVs (BMSC-EVs) provided highly reproducible benefits across species, adipose-derived EVs (ADSC-EVs) exhibited particularly potent structural rescue, yielding MDs of -3.01 and -8.70 in human and animal subgroups, respectively. This differential efficacy is plausibly driven by variation in EV miRNA cargo (Table S5). Among MSC-derived EV miRNAs, miR-140-5p has been most directly linked to chondroprotection: synovial-MSC exosomes engineered to overexpress miR-140-5p enhanced articular chondrocyte proliferation and migration and prevented OA progression in a rat model, indicating cargo-specific potency rather than a generic vesicle effect [11]. The frequently reported chondroprotective miRNAs (e.g. miR-100-5p, miR-140-5p, miR-26a-5p) converge on extracellular-matrix metabolism and mTOR/autophagy pathways. However, in the absence of standardized potency assays mapping specific cargo concentrations to target engagement, the precise dose-response relationship for any individual miRNA cargo remains elusive across the included studies.

Methodological rigor across the included studies remains a significant vulnerability and likely contributes substantially to the heterogeneity observed in our pooled estimates. Although most studies adhered to the basic characterization triad endorsed by MISEV 2023, morphology by transmission electron microscopy, particle quantification by nanoparticle tracking analysis, and a small panel of tetraspanin markers, MISEV 2023 emphasizes that adequate EV characterization now requires multi-modal verification across morphology, particle metrics, and protein/lipid composition, together with full reporting of separation conditions [7]. Our formal MISEV 2023 adherence assessment of all 38 included studies (Table S2) revealed that no study achieved Complete adherence, and only 18 % reached Substantial adherence; the remaining 82 % were classified as Partially or Limitedly adherent. The two most pronounced reporting deficits, numeric size distribution data (adequate in only 11 % of studies) and procedural reporting clarity covering centrifugation force, temperature, and equipment (adequate in only 13 %), directly compromise reproducibility and prevent meaningful between-study comparison of EV preparations. The omission of exact ultracentrifugation speeds, temperatures, and critical solution compositions in many original reports threatens reproducibility, and the lumping of distinct vesicular populations (small EVs versus larger microvesicles) due to unstandardized isolation inevitably confounds efficacy readouts and downstream regulatory oversight. These findings provide quantitative support for the field-wide call to make MISEV 2023 compliance a standard prerequisite for publication of EV therapeutic studies.

To transition MSC-EVs from promising experimental biologics to regulated advanced therapy medicinal products (ATMPs), the field must pivot toward rigorous ADMET compliance and harmonized manufacturing. EV therapeutics already fall under the regulatory umbrella of established biologics frameworks: in the United States, EV products require Investigational New Drug filings to FDA’s CBER or CDER, and in the European Union the EMA classifies EV-based medicines as ATMPs under Regulation (EC) No 1394/2007, with sponsors filing Clinical Trial Applications via the centralised Clinical Trials Information System and obtaining Committee for Advanced Therapies classification [49-51]. Meeting these standards requires moving beyond simple efficacy readouts toward in vivo imaging that quantifies intra-articular EV half-life and biodistribution, evaluation of hydrogel or scaffolding delivery systems to prolong the therapeutic window, and the establishment of standardized particle-to-biomolecule ratios for dosing. Only through methodological harmonization and integrated PK/PD modelling can MSC-EVs achieve the clinical reliability required for true disease modification in osteoarthritis.