Mesenchymal Stromal Cell Therapy: An Evidence Review

Evidence reviewed through: July 2026

Scope and limits of this review. This is a narrative evidence review, not a systematic review.

It does not follow PRISMA methodology, it did not use a prespecified search protocol, and it should

not be read as one. Every study cited below has been individually verified against its published

abstract or full text, and every DOI resolves. Where the evidence is weak, contested, or negative,

that is stated rather than omitted.

Key Points

• Where MSC therapy works best, it works well. In steroid-refractory acute graft-versus-host disease, meta-analysis of four RCTs (650 patients) found significantly improved overall response (RR 1.13, 95% CI 1.03–1.23) and complete response (RR 1.43, 95% CI 1.19–1.70), reduced chronic GVHD (HR 0.60) and improved failure-free survival (HR 0.72) — without any increase in adverse events. This is the basis of the first US approval.
• MSC therapy remains investigational for most other indications discussed here. The applications most commonly marketed direct-to-patient are not among the approved ones.
• The most-cited systematic review of stem cell injection for knee osteoarthritis concluded: "In the absence of high-level evidence, we do not recommend stem cell therapy for KOA."
• Transient fever is the one adverse event consistently and significantly associated with MSC infusion across the major safety meta-analyses.
• MSCs are procoagulant. This is the most mechanistically credible serious-harm pathway in the field, it is dose- and source-dependent, and umbilical cord–derived cells express the highest levels of the initiating factor. It has its own section below.
• Long-term safety data is genuinely limited: most follow-up is under five years, and roughly half of early trials did not report how long they followed patients for harms.
• Evidence quality across cardiac, neurological and respiratory applications is dominated by null primary endpoints with favourable subgroup analyses — a pattern that warrants caution, not enthusiasm.

1. Definition and Classification

Mesenchymal stromal cells are multipotent cells capable of differentiating into mesenchymal lineages including osteoblasts, chondrocytes and adipocytes (Pittenger et al., 1999). The International Society for Cellular Therapy established minimal defining criteria: plastic adherence under standard culture conditions; expression of CD73, CD90 and CD105 with absence of CD34, CD45, CD14, CD11b, CD79a, CD19 and HLA-DR; and capacity for trilineage differentiation (Dominici et al., 2006).

Sources

Nomenclature

Friedenstein first identified bone-marrow-derived stromal cells with osteogenic potential in the 1960s (Friedenstein, Piatetzky-Shapiro and Petrakova, 1966). Caplan, who coined "mesenchymal stem cell," later argued the name should change to "medicinal signalling cells" to reflect that their therapeutic action is paracrine rather than by differentiation and replacement (Caplan, 2017). That reframing matters clinically: it is the difference between a therapy that rebuilds tissue and one that changes the signalling environment.

A note on chronology. Claims that MSCs have "50+ years of clinical use" conflate two different dates. Friedenstein's laboratory identification was in the 1960s; clinical use began in the mid-1990s. The distinction matters whenever length of track record is offered as evidence of safety.

2. Biological Mechanisms

2.1 Paracrine signalling

Therapeutic effects are predominantly mediated by secreted factors rather than by direct differentiation and engraftment. The secretome includes VEGF, HGF, TGF-β, IGF-1 and FGF, alongside extracellular vesicles carrying miRNA and protein cargo.

Human autopsy evidence supports the "hit and run" model: donor DNA was detectable in recipient tissues at low levels declining with time since infusion, with no sustained engraftment (von Bahr et al., 2012a).

2.2 Immunomodulation

MSCs modulate innate and adaptive immunity — inducing regulatory T-cells, inhibiting T-cell proliferation, suppressing B-cell antibody production, and promoting M2 macrophage polarisation — through mediators including IDO, PGE2, TGF-β, IL-10 and HLA-G.

2.3 What this does not establish

Mechanistic plausibility is not clinical efficacy. A review of the field's translational record states the position plainly: most clinical-stage MSC therapies "have been unable to meet primary efficacy end points," and the innate therapeutic functions of MSCs administered to humans "are not as robust as demonstrated in preclinical studies" (Levy et al., 2020).

3. Clinical Applications

3.1 Knee osteoarthritis — the largest evidence base, and what it concludes

This is the most studied application, and the most misrepresented.

The most-cited systematic review in this area. Pas et al. searched seven databases and found five RCTs and one non-RCT. All five RCTs reported superior patient-reported outcomes versus controls at final follow-up (range 24–48 months), and superior radiological outcomes were reported. But "all trials were at high risk of bias, resulting in level-3 evidence," and the authors' conclusion was unambiguous:

"Six trials with high risk of bias showed level-3 or level-4 evidence in favour of stem cell injections in KOA. In the absence of high-level evidence, we do not recommend stem cell therapy for KOA." (Pas et al., 2017)

Note also what that review does not contain. It is sometimes cited for durability — that MSC effects persist 12–24 months or longer and outlast hyaluronic acid and corticosteroid — but no such comparison appears anywhere in it. Its included trials ran to a final follow-up of 24–48 months without any head-to-head against another injectable.

Current position. A 2026 meta-analysis of 28 randomised trials found MSC therapies improved pain (ΔVAS mean difference −1.67) and several function domains, while WOMAC, KOOS quality-of-life and Lequesne scores showed no significant difference, and MRI-based WORMS structural scores were non-significant. The authors conclude the role is "primarily symptom-modifying rather than structure-modifying" (Awad et al., 2026).

How much of the benefit is the cells. A 2025 meta-analysis of placebo-controlled trials calculated that contextual (placebo-related) effects accounted for approximately 63% of pain reduction and 61% of functional improvement at 6 months, and about 50% of pain relief at 12 months, with MSCs conferring "only a modest incremental benefit." GRADE certainty: low (Yin et al., 2025).

Society position. The 2025 ESSKA-ORBIT consensus on cell-based therapy supports use in Kellgren-Lawrence grades 1–3 as a second-line option, notes benefits are "inferior" in KL 4, and rated 22 of its 27 statements as low-level evidence or expert opinion (de Girolamo et al., 2025).

Durability. Consistently demonstrated to about 12 months. A 60-month placebo-controlled trial found improvement at 6, 12, 24 and 36 months, then decline "with scores nearing baseline at 60 months" (n=29; Jafar et al., 2026). An uncontrolled 5-year series of 11 patients reported maintained improvement (Kim et al., 2022) — a difference explained by the contextual-effect finding above, since an uncontrolled series would show sustained improvement whether or not the cells were responsible.

3.2 Graft-versus-host disease — the strongest evidence in the field

This is where MSC therapy has performed best, and it deserves to be stated at full strength.

Two 2025 meta-analyses of the same four randomised controlled trials (650 patients) examined MSCs added to second-line therapy for steroid-refractory acute GVHD:

• Overall response rate: RR 1.13 (95% CI 1.03–1.23, P = 0.007) (Xiao et al., 2025); independently RR 1.13 (95% CI 1.04–1.23, P = 0.005) (Khan et al., 2025)
• Complete response: RR 1.43 (95% CI 1.19–1.70, P < 0.001) (Xiao et al., 2025); RR 1.60 (95% CI 1.35–1.90, P < 0.00001) (Khan et al., 2025)
• Reduced incidence of chronic GVHD: HR 0.60 (95% CI 0.42–0.86, P = 0.005)
• Improved failure-free survival: HR 0.72 (95% CI 0.54–0.95, P = 0.022) (Xiao et al., 2025)
• Benefit greatest in grade III–IV disease and in gut, skin or multi-organ involvement
• No increase in adverse events (RR 1.00, P = 0.96) or serious adverse events (RR 1.01, P = 0.79) (Khan et al., 2025)

Overall survival was not improved in either analysis (RR 1.05, 95% CI 0.90–1.23, P = 0.52), and both call for larger standardised trials with longer follow-up.

That profile — significant improvement in response and failure-free survival, in a life-threatening condition with few alternatives, without an increase in harm — is what a working therapy looks like. It is also the basis of the field's first US approval (§7).

3.3 Meniscal injury

A trial often cited as evidence of cartilage regeneration measured meniscal volume after meniscectomy. In a randomised, double-blind controlled study of 55 patients undergoing partial medial meniscectomy, a single injection of 50 × 10⁶ or 150 × 10⁶ allogeneic MSCs produced significantly increased meniscal volume — defined a priori as a 15% threshold — in 24% of the lower-dose group and 6% of the higher-dose group at 12 months, versus none in controls (p = 0.022). No ectopic tissue formation or clinically important safety issues were identified (Vangsness et al., 2014).

Two details are easily lost when this trial is summarised: fewer than a quarter of treated patients met the threshold, and the lower dose outperformed the higher — which does not support a more-is-better dosing model.

3.4 Neurological conditions

Stroke. The MASTERS trial randomised 129 patients with acute ischaemic stroke. It found no difference in global stroke recovery at day 90: odds ratio 1.08 (95% CI 0.55–2.09), p = 0.83. The intervention was safe and well tolerated, with no dose-limiting toxicity (Hess et al., 2017).

Note what was actually given: MASTERS used multipotent adult progenitor cells, not mesenchymal stromal cells. The two are routinely conflated when this trial is cited as stem cell evidence, and they are different products.

Multiple sclerosis. A 2024 systematic review and meta-analysis of 30 studies found improvement in disability scores with intrathecal delivery (weighted mean difference −0.28; 95% CI −0.53 to −0.03), but no significant change with intravenous delivery. Annualised relapse rate did not significantly decrease. The authors conclude effectiveness "remains uncertain due to study limitations and ineffective outcome measures" (Vaheb et al., 2024).

3.5 Cardiac conditions

A Cochrane review of stem/progenitor cell therapy for chronic ischaemic heart disease and heart failure found low-quality evidence of reduced mortality and improved ejection fraction over short- and long-term follow-up, and possible reductions in non-fatal myocardial infarction. But long-term left ventricular ejection fraction measured by MRI showed no evidence of effect (mean difference −1.60; 95% CI −8.70 to 5.50), and the authors state findings "should be interpreted with caution, as event rates were generally low, leading to a lack of precision" (Fisher et al., 2016).

Every outcome in that review was graded low-quality. The review also concerns bone-marrow-derived stem/progenitor cells broadly, not MSCs specifically.

3.6 A recurring pattern worth naming

Across stroke, cardiac and respiratory applications, the dominant pattern is null primary endpoints accompanied by favourable subgroup or secondary analyses. Subgroup findings in trials that missed their primary endpoint are hypothesis-generating, not evidence of efficacy. When a positive result from one of these trials is quoted at you, the question worth asking is whether the trial met its primary endpoint at all.

3.7 Other applications, graded by evidence

Evidence maturity varies enormously across the remaining indications, and lumping them together would misrepresent both the stronger and the weaker cases.

Where randomised evidence shows a significant effect:

• Perianal fistulas in Crohn's disease. Expanded allogeneic adipose MSCs (darvadstrocel) were tested in a phase 3 randomised, double-blind controlled trial and are approved in the European Union for this indication (Panés et al., 2016).
• Ageing frailty. A phase I/II randomised, double-blind, placebo-controlled trial of umbilical cord MSCs reported improved physical component scores from the first post-treatment visit (P = 0.042), improved EQ-VAS (P = 0.002 at 6 months), improved timed-up-and-go (P < 0.05), improved grip strength (P = 0.002 at 6 months), and reduced TNF-α and IL-17 (P = 0.034 and 0.033), with no difference in adverse event incidence (Zhu et al., 2024). A separate phase II RCT reported improved 6-minute walk distance and reduced inflammatory markers (Tompkins et al., 2017). Both are small; neither is confirmatory.
• COPD. A meta-analysis of 11 studies (371 patients) found a significant increase in 6-minute walk distance (52 m, 95% CI 18–87, P < 0.05), a non-significant trend in FEV1 (+71 mL, P = 0.056), and no reduction in hospitalisation risk (Calzetta et al., 2022).

Where trials exist but efficacy is not established: rheumatoid arthritis, lupus, pulmonary fibrosis, Parkinson's disease, peripheral artery disease.

Where controlled outcome evidence is largely absent: skin rejuvenation, hair restoration, and general wellness or anti-ageing indications. These are the most heavily marketed and the least supported.

4. Cell Source Considerations

4.1 What differs between sources

• Donor age affects bone-marrow and adipose-derived cells; perinatal sources avoid this.
• Collection burden differs sharply: marrow aspiration and liposuction are procedures with their own complication profiles; cord tissue requires none from the patient.
• Proliferative capacity is generally higher in perinatal sources.

4.2 What does not follow from that

No definitive clinical advantage of autologous over allogeneic MSCs has been demonstrated (Ankrum, Ong and Karp, 2014) — a statement that cuts against superiority claims in either direction. Claims that young allogeneic MSCs outperform aged autologous cells in most applications, or that patients over 50 should prefer cord-derived cells for superior cell quality, are inferences from laboratory characteristics. Neither is established by clinical outcome data.

4.3 The counterweight that must accompany any source discussion

Meta-analysis has shown bone marrow MSC infusion to be safe. Specialists have explicitly cautioned that "it is not clear if this also applies to diverse new cell products derived from other sources, such as adipose and perinatal tissues," because different products display varying levels of highly procoagulant tissue factor and may trigger the instant blood-mediated inflammatory reaction (Moll et al., 2019).

This is not a minor caveat. See section 6.3.

5. Processing and Quality

5.1 Fresh versus cryopreserved — genuinely contested

The case that thawing impairs cells: freeze-thawed MSCs show reduced responsiveness to inflammatory stimuli, impaired anti-inflammatory mediator production, increased triggering of the instant blood-mediated inflammatory reaction, and strong complement activation immediately after thawing, with faster clearance from circulation (Moll et al., 2014).

The case that it does not matter clinically: a systematic review of 18 comparative preclinical studies — 257 in-vivo efficacy experiments across 101 outcome measures — found only 2.3% (6/257) significantly different, and of those six, two favoured fresh and four favoured cryopreserved. The authors concluded most efficacy and potency outcomes did not differ and that the evidence supports considering cryopreserved products (Dave et al., 2022).

And directly on the manufacturing question: a GMP-grade comparison found that cryopreservation does not alter the main characteristics of clinical-grade MSCs (Luetzkendorf et al., 2015).

Honest position: there is a real, mechanistically documented post-thaw impairment, and there is good evidence it does not translate into worse outcomes. Preparing cells fresh is defensible; claiming demonstrated clinical superiority is not.

5.2 Passage number — an underrated finding

This is one of the more striking clinical observations in the MSC literature and it deserves more prominence than it usually gets. In 31 patients treated for steroid-resistant graft-versus-host disease or haemorrhagic cystitis, one-year survival was 75% in patients who received early-passage cells (passages 1–2) versus 21% with later-passage cells (passages 3–4) (p < .01). The authors concluded that treatment with early-passage MSCs improved survival (von Bahr et al., 2012b).

That is a large difference attributable to a manufacturing variable, in a small and severely ill cohort. It supports asking any provider what passage number their product is.

5.3 Dose

The dose literature does not support "more is better," and is not settled:

• An analysis of 914 registered MSC trials found minimum effective intravenous doses clustering around 100–150 million cells, "with lower and higher doses being less effective" (Kabat et al., 2020).
• In knee osteoarthritis, a network meta-analysis of 11 randomised trials found moderate doses produced the greatest long-term improvement, with high doses producing more adverse events (Xie et al., 2026).
• In the meniscal trial above, the lower dose outperformed the higher (Vangsness et al., 2014).
• Repeated dosing outperformed single dosing at 6 and 12 months in knee OA — but with a higher incidence of adverse events (Deng et al., 2025).

Dose also interacts with the thrombotic risk described below, where procoagulant effects are dose-dependent.

6. Safety

6.1 What the major safety reviews found

Three systematic reviews of intravascular MSC administration in adults form the backbone of the safety literature.

Two findings in these tables are easy to read past. No trial in any of the three reviews was stopped early for safety reasons. And in the 2020 update, *mortality was significantly lower in MSC groups (RR 0.78, 95% CI 0.65–0.94)* — a confidence interval that excludes 1. That result comes from trials in seriously ill populations rather than elective patients, and it was not a prespecified efficacy endpoint, so it should not be read as evidence that MSCs extend life. But it is a positive signal and it belongs on the record alongside the fever finding.

How to read "no association." "Did not detect an association" is a weaker, and different, statement from "no serious adverse events occurred." Serious events, including deaths, did occur in these trials; they were not statistically attributable to the cells, in populations where such events are common from the underlying disease itself. And all three reviews found a statistically significant increase in fever — the treatment's most reproducible adverse effect, not an incidental note.

The authors of the foundational review added a caution worth carrying: because MSC use "may be associated with neoplastic growth long term," it was "difficult to understand why approximately 50% of studies did not report follow up duration for adverse events."

6.2 Adverse events

Consistently and significantly associated: transient fever, typically resolving within 24–48 hours.

Commonly reported: fatigue, headache, acute infusion-related reactions (8% of participants in uncontrolled trials in the foundational review). For intra-articular administration specifically: injection-site pain (RR 2.04) and joint swelling (RR 3.39) versus control, with serious complications uncommon and not significantly increased (Awad et al., 2026).

Reported but not statistically associated in meta-analysis: infection, thrombotic and embolic events, organ complications, death, malignancy.

On serious-adverse-event percentages. Figures such as "uncommon (<1%)" are quoted routinely in this field, and we could not trace them to any source. None of the three safety reviews reports a serious-adverse-event rate; all report effect estimates versus control, which cannot be converted into a bare percentage. A percentage without a denominator, a population and a follow-up window is not a statistic.

6.3 Thrombotic and embolic risk — a signal that requires its own section

This risk is usually dealt with in a passing clause — rare case reports of pulmonary embolism after intravenous infusion of high cell doses. That framing understates what is arguably the most mechanistically credible serious-harm pathway in cell therapy, and one that implicates the route, source and dose most commonly used commercially. It is set out in full here.

MSCs are procoagulant. They express tissue factor, the initiator of the extrinsic coagulation pathway, and phosphatidylserine on their surface. These interact with the coagulation system and can contribute to thrombus formation. Expression is affected by cell dose, tissue source, passage number and culture conditions (Yang et al., 2024).

Tissue factor expression differs by source — and umbilical cord is at the high end. In a study of xeno- and serum-free manufactured cells, adipose tissue and umbilical-cord-derived MSCs expressed the highest levels of tissue factor, followed by dental pulp, with the lowest expression in bone-marrow-derived MSCs. MSCs of diverse origin induced fibrin clots in healthy plasma, partially suppressed by an anti-tissue-factor antibody (Hoang et al., 2024).

There is a measurable human signal, and it is route-dependent. In the same study, patients receiving intravenous umbilical cord MSC infusion showed a transient rise in D-dimer, while this remained stable in patients receiving intrathecal infusion (Hoang et al., 2024).

Tissue factor is not a reliable predictive biomarker. The same authors found no correlation between tissue factor expression and D-dimer or other coagulation indicators, concluding it "cannot be used as a solid biomarker to predict MSC-induced hypercoagulation."

Their recommendation, quoted: "Local administration, prophylactic intervention with anticoagulation drugs, and monitoring of coagulation indicators are useful to prevent thrombogenic events in patients receiving MSCs" (Hoang et al., 2024). This is a matter for treating clinicians, not for patients to act on independently.

Related mechanism. This is the same instant blood-mediated inflammatory reaction described in the product-diversification literature, where variable tissue factor levels across newer cell sources prompted a call for hemocompatibility testing and new clinical guidelines (Moll et al., 2019), and which is also triggered more strongly by freeze-thawed cells (Moll et al., 2014).

How to hold this honestly. In randomised trials, thrombotic and embolic events have not been significantly increased (Thompson et al., 2020; Hum et al., 2026). But those trials are mostly small and not powered to detect rare events, the procoagulant mechanism is demonstrated rather than theoretical, the effect is dose- and source-dependent, and the human biomarker signal is specific to the intravenous route. The correct statement is not "this does not happen" and not "this is a common danger." It is: a real, mechanistically established procoagulant effect exists, it has not produced a detectable excess of clinical events in trials to date, and it is the most plausible route to serious harm — which is why product source, dose, route and coagulation monitoring matter.

6.4 Infection and manufacturing

Trial safety data describes carefully manufactured products given under protocol. It does not describe the commercial marketplace.

A public health investigation examined bacterial infections in 20 patients who received umbilical cord blood–derived products marketed as stem cell treatment for pain, osteoarthritis, rheumatoid arthritis and injury. All but one required hospitalisation. Of unopened, undistributed vials sampled, 65% (22 of 34) were contaminated, with whole-genome sequencing indicating a common source of contamination during processing (Hartnett et al., 2021).

Separately, in the graft-versus-host disease cohort described above, 54% of patients who recovered from acute GVHD died of late infections between 4 months and 2 years after MSC treatment, and the authors concluded that "death from infection was common in MSC-treated patients" (von Bahr et al., 2012b). These were profoundly immunosuppressed transplant patients in whom infection mortality is expected from the underlying disease and its treatment — the study does not establish that MSCs caused those deaths. It is, however, part of what that paper reports, and the paper is frequently cited for its long-term safety findings alone.

6.5 Immunogenicity

MSCs are immune evasive, not immune privileged. Studies describing generation of antibodies against, and immune rejection of, allogeneic donor MSCs indicate they are not truly immune privileged, and patient sensitisation toward donor antigens is a recognised consideration (Ankrum, Ong and Karp, 2014). Studies controlling for MHC matching report both cell-mediated and humoral responses to mismatched MSCs, with clinical implications still unknown (Berglund et al., 2017).

6.6 Tumour risk

No MSC-derived tumours have been documented in the published clinical literature. Supporting evidence: an autopsy series of 18 patients found no ectopic tissue formation and no donor-origin tumours (von Bahr et al., 2012a); a 10-year follow-up of 72 patients treated for knee chondral defects found no tumours or deep infections (Teo et al., 2019); and the large meta-analyses found no association with malignancy.

What this does not establish: the autopsy series is 18 patients, the 10-year cohort is 72, most trial follow-up is under five years, and tumour latency can exceed that. Longer follow-up periods are sometimes claimed — "15 years" or more — but we could not identify a published MSC malignancy-surveillance study of that length.

The accurate position: no documented cases, sound biological rationale for low risk, and follow-up insufficient to exclude rare or late events.

6.7 Contraindications

Active malignancy is the most important contraindication; safety in active cancer is not established. Others include active systemic infection, significant bleeding disorders or anticoagulation, unstable cardiovascular disease, and pregnancy. Poorly controlled diabetes, significant immunosuppression and recent major surgery require individual assessment. Specific numeric thresholds are clinical judgements about individuals and are not published here.

7. Regulatory Landscape

International guidance adopts the standard that stem cell products substantially manipulated or used for non-homologous purposes are drugs or advanced therapy medicinal products — aligned with the US FDA, the European Medicines Agency and Australia's TGA. Cells expanded in culture over weeks fall within that definition. The same guidance issues an updated recommendation to "forcefully caution against the premature commercialization of unproven stem cell-based interventions," and notes that patient registries are not adequate substitutes for randomised controlled trials, with registries used as a form of advertising described as "at best misleading and goes against a duty of care for patients" (Lovell-Badge et al., 2021).

Approved cell therapy products exist for specific narrow indications in specific jurisdictions. The applications most commonly marketed direct-to-patient — osteoarthritis, anti-ageing, systemic wellness — are not among the approved indications in most countries, and are delivered either within clinical trials or outside the approval framework.

8. Limitations of the Current Evidence

• Product heterogeneity is severe. "MSC therapy" describes a family of substantially different products varying by source, expansion, passage, cryopreservation state, dose and route. Pooling across them is problematic, and safety data does not transfer automatically between them.
• Trial quality is the binding constraint. The knee OA systematic review found all included trials at high risk of bias. Most cardiac evidence is graded low quality.
• Follow-up is short. Most trials stop at 12–24 months; roughly half of early trials did not report adverse-event follow-up duration at all.
• Contextual effects are large. Roughly half to two-thirds of symptomatic improvement in knee OA is attributable to placebo and context, which makes uncontrolled series and registry data largely uninformative about efficacy.
• Null-with-favourable-subgroup is the dominant result pattern in the larger, better-controlled trials.
• Economic evidence is unfavourable where it exists. A Markov decision analysis found biologic injection for knee OA not cost-effective, with the limiting factor being "a lack of established clinical efficacy" rather than price (Rajan et al., 2020).

9. Practical Implications

Questions worth asking any provider

• What is the cell source, and what safety data exists for that source by this route?
• What is the passage number? (Early-passage cells outperformed later-passage in the one clinical comparison available.)
• Was the product culture-expanded, and under what regulatory framework was it manufactured?
• Will I see the certificate of analysis — cell count, viability, sterility, endotoxin, mycoplasma — for my batch?
• What coagulation considerations apply given my history, and is monitoring performed?
• What outcome, at what timepoint, would count as success — and what happens if it is not reached?
• What is the strongest trial against this treatment, and how does the provider account for it?

Realistic expectations

Symptom improvement is commonly reported and is genuine; a large share of it is contextual. Structural repair is not demonstrated. Benefit is consistently demonstrated to about 12 months. Response cannot currently be predicted for an individual — no validated biomarkers exist for patient selection.

"Expected improvement" figures of the kind quoted throughout this field — pain reduction of 30–70%, function 20–50%, response rates of 60–80% — are not traceable to a controlled trial, and none of them should be read as a prediction for an individual patient.

10. Conclusion

The evidence for mesenchymal stromal cells is genuinely mixed, and both halves of that deserve stating.

Where it is strong, it is strong. In steroid-refractory acute graft-versus-host disease, randomised evidence pooled across 650 patients shows significantly improved response and complete response rates, reduced chronic GVHD, and improved failure-free survival — with no increase in adverse events. That indication now has an approved product. Randomised trials also show significant benefit for perianal fistulas in Crohn's disease, and early-phase controlled trials in ageing frailty report improvements in physical performance and inflammatory markers. The mechanistic account is coherent, and across three large safety reviews covering more than 6,000 patients no trial was stopped early for safety.

Where it is weak, it is weak. The most-cited systematic review in the largest application area recommends against the therapy. The best-controlled trials in stroke and cardiac disease missed their primary endpoints. Structural repair is not demonstrated. Roughly half to two-thirds of symptomatic benefit in knee osteoarthritis is contextual. Long-term surveillance adequate to exclude rare harms has not been done. And the field's most credible serious-harm mechanism — procoagulant activity — is dose- and source-dependent in a way that specifically implicates the products and routes most used commercially.

The pattern is consistent: MSC therapy performs best in severe, inflammation-driven conditions with few alternatives, and least well in the elective, quality-of-life indications for which it is most heavily marketed. That is the distinction worth carrying away.