MSC Therapy for Arthritis: Real Knee OA Signal, Big Gaps

Thirty-six trials do not answer four arthritis questions

Mesenchymal stromal cell (MSC) therapy has a real short-term pain signal in knee osteoarthritis (OA). It does not yet have comparably measurable evidence for rheumatoid arthritis (RA), ankylosing spondylitis (AS), or systemic sclerosis (SSc). That distinction gets buried beneath the headline numbers in a 2024 systematic review: 36 randomized controlled trials (RCTs) and 2,076 participants across four conditions.

The review found statistically significant reductions in OA pain, stratified by cell source. But it reported no pooled effect sizes, confidence intervals, or p-values for common RA, AS, or SSc disease measures. "Potential therapeutic effect" is not a substitute for a quantified result. For someone considering an expensive cell-therapy clinic, the condition being treated matters more than the total number of patients in the paper.

MSC therapy is therefore an evidence-supported option to investigate for knee OA pain, not a validated umbrella treatment for autoimmune or rheumatic disease.

What the largest review actually found

Zeng et al.'s 2024 meta-analysis pooled 36 RCTs involving 2,076 participants with RA, OA, AS, or SSc. Its quantifiable results concentrate in OA. On the visual analog scale (VAS) for pain, the pooled standardized mean differences favored MSC therapy for all three sources:

• Bone marrow-derived MSCs (BM-MSCs): SMD −0.95; 95% CI, −1.55 to −0.36; P = 0.002.
• Adipose tissue-derived MSCs (AT-MSCs): SMD −1.26; 95% CI, −1.99 to −0.52; P = 0.0009.
• Umbilical cord-derived MSCs (UC-MSCs): SMD −2.03; 95% CI, −2.99 to −1.07; P < 0.0001.

By the usual convention for standardized mean differences, each estimate exceeds the 0.8 threshold often labeled a large effect. AT-MSCs also significantly improved WOMAC pain (P = 0.0001), physical function (P = 0.001), and total WOMAC score (P = 0.0003). Those are patient-relevant outcomes, although statistical significance does not tell us how much improvement any individual patient should expect.

Independent reviews focused on knee OA broadly support a pain-and-function signal. A 2021 network meta-analysis by Zhao et al. found adipose and bone marrow MSC injections performed better than hyaluronic acid and saline. A 2024 network meta-analysis by Chen et al. likewise found source-dependent differences among BM-MSC, AT-MSC, and UC-MSC therapies, consistent with the ranking in Zeng's analysis. Tan et al. also reviewed MSC monotherapy in knee OA, avoiding the added ambiguity of protocols that combine cells with other treatments.

This repetition across reviews makes the OA signal more credible than a single favorable study would. It does not turn a pooled standardized effect into proof that every commercial MSC product, dose, processing method, or injection schedule works.

The biggest UC-MSC estimate deserves restraint

The UC-MSC pain estimate, −2.03 standard deviations, is strikingly large. Its 95% confidence interval runs from −2.99 to −1.07, a span of almost two standard deviations. That degree of imprecision and variation within a cell-source subgroup should temper confidence in the point estimate.

Zeng et al. searched Chinese- and English-language databases, including CNKI and Wanfang. Such broad searching can identify trials missed by English-only reviews, but it can also add small studies with uneven reporting and trial methods. The review's abstract does not provide the subgroup heterogeneity statistic needed to quantify that problem. Treat the UC-MSC number as a possible upper-bound estimate, not as a reasonable promise of two-standard-deviation pain relief.

The best-described individual UC-MSC study in this group is a 2025 randomized, controlled, double-blind pilot trial by Pico et al. of a single cryopreserved UC-MSC dose for knee OA. Blinding matters for pain outcomes, especially when patients know they received a novel procedure. Yet its pilot design limits what it can establish about reproducibility, uncommon harms, or long-term durability.

Beating placebo is not the clinical finish line

The most useful comparison is not MSC versus saline alone. It is MSC versus treatments patients would plausibly receive, including corticosteroid injection, hyaluronic acid, and exercise-based care.

That is where the picture becomes less tidy. A 2025 RCT-based meta-analysis by Han et al. concluded that stromal vascular fraction and adipose-derived MSC injections had variable efficacy and higher potential complications than corticosteroid and hyaluronic acid for knee OA. This directly qualifies the broader claim that MSC therapy is relatively safe. The conceptually important difference is the comparator: Han's analysis assessed active standard injection options, whereas Zhao et al.'s favorable 2021 network analysis compared MSCs with hyaluronic acid and saline but did not include corticosteroids.

A 2023 review by Kim et al. also found heterogeneous and mixed results for autologous adipose-derived stem-cell and stromal vascular fraction injections in knee OA RCTs. Variation in donors, cell preparation, dose, and protocol can produce different results under the same marketing label. "Stem cells" is a category, not a standardized drug.

Care also has context outside the syringe. A 2023 network meta-analysis by Liao et al. found that injection plus physical therapy outperformed injection alone across modalities in knee OA. An MSC trial using injection-only care as its comparator may therefore overstate the practical advantage over a well-delivered multimodal program.

There is no comparable conclusion for RA, AS, or SSc

Despite their inclusion in the 36-trial review, RA, AS, and SSc do not receive reportable pooled estimates for DAS28, BASDAI, ASDAS, modified Rodnan skin score, or organ outcomes in the 2024 publication's available abstract. No independent disease-specific MSC review appears alongside the OA-focused reviews cited here.

A 2025 companion review by Zeng et al. addresses autoimmune and rheumatic diseases more broadly, but the available abstracted information does not provide the quantitative outcomes needed to resolve this gap. The responsible conclusion is insufficient evidence, not a negative verdict and certainly not an extrapolation from knee OA. These diseases differ in biology, treatment goals, and the outcomes that matter. Pain relief in an osteoarthritic knee does not demonstrate control of autoimmune inflammation, spinal disease activity, skin fibrosis, or internal-organ involvement.

What remains unknown

Trials in the pooled literature vary across cell source, likely cell dose, passage number, delivery route, and whether patients receive one or several injections. Pooling biologically different interventions can generate an average that fits few real protocols. A 2025 dose-response meta-analysis by Lohrasbi et al. addresses dose and safety for BM-MSCs in OA, but comparable dose-response evidence is unavailable for UC-MSCs, AT-MSCs alone, or the non-OA conditions.

Durability is another open question. The available studies emphasize short-term pain and function, while large prespecified trials of structural outcomes such as 12-month MRI cartilage thickness or joint-space width do not appear in this evidence base. MSC injections may help symptoms without rebuilding cartilage; the current evidence cannot settle that distinction.

Verdict

For knee OA, intra-articular MSC therapy has a replicated, statistically significant short-term pain signal. BM-MSCs and AT-MSCs have the most credible source-specific support in this evidence base, while the exceptionally large UC-MSC estimate needs more skeptical interpretation. The treatment has not clearly established risk-adjusted superiority over corticosteroid injection, and the newest active-comparator meta-analysis raises a potential complication signal.

For RA, AS, and SSc, the evidence remains too thinly quantified to guide treatment decisions. A clinic offering MSCs for those conditions should be able to show disease-specific controlled-trial data, not point to a pooled arthritis headline. Until then, proven disease-modifying care and structured rehabilitation remain the reference standard patients should expect.