Stem Cell Injections for Knee OA: What Evidence Shows

Stem cell injections for knee osteoarthritis can reduce pain and improve function, but the trade-off depends heavily on what is injected. In the most detailed safety review available, adverse events occurred in 8.1% of stromal vascular fraction (SVF) recipients and 10.7% of bone marrow-derived MSC recipients, versus 29.5% for cultured adipose-derived MSCs and 51.7% for umbilical cord MSCs. The reported events were generally temporary swelling, pain flares, or effusions—not infections, cancer, or death—but a transient complication still counts when the treatment's benefit is modest and delayed.

The evidence does not support cartilage regrowth claims. Trials have not shown robust MRI-confirmed cartilage regeneration as a primary randomized endpoint. The likely clinical effect is anti-inflammatory and paracrine: cells release signals that may alter the joint environment rather than rebuild worn cartilage. That may help symptoms; it does not reverse established structural osteoarthritis.

The safety gap is real, but the estimates are uneven

A 2025 systematic review of complications by Riggle et al. included 48 studies and 1,152 patients receiving isolated stem cell-based knee injections. It found 28 adverse events among 346 SVF recipients (8.1%), 57 among 533 bone marrow MSC recipients (10.7%), 72 among 244 cultured adipose-MSC recipients (29.5%), and 15 among 29 umbilical cord MSC recipients (51.7%). Cultured adipose and umbilical cord preparations had significantly higher adverse-event rates than bone marrow MSCs and SVF (P<.01).

No severe adverse events appeared in that review. Most reported problems resolved in about four weeks. That is reassuring, but it does not prove equal safety across products. The umbilical cord estimate comes from only 29 patients in two studies, so its confidence interval is necessarily wide even though the observed rate is high. The review also found inconsistent adverse-event reporting: its included studies were level II–IV evidence, and most did not use validated grading scales or report duration consistently.

A 2024 network meta-analysis by Tang et al., spanning 16 randomized controlled trials and 1,005 participants, reached a similar safety ranking. Placebo ranked safest, while adipose-derived MSCs ranked least safe, with a P-score of 13.3. In that analysis, adding hyaluronic acid was associated with a lower adverse-event risk (RR 0.77, 95% CI 0.61–0.97). This is an association within a mixed treatment network, not proof that hyaluronic acid makes every MSC product safer.

The clinically relevant comparison is not only between cell sources. A 2025 randomized-trial meta-analysis by Han et al. found variable efficacy and higher potential complications for SVF and MSC injections than for corticosteroid and hyaluronic acid injections. For someone considering an expensive procedure, a low rate of severe complications is not enough; the injection must also improve outcomes enough to justify its added burden over established care.

Which source works best? The answer changes when co-treatments disappear

Broad network meta-analyses can make adipose-derived and umbilical cord cells look like the leaders. Tang et al. ranked adipose-derived MSCs first for pain on the visual analog scale (P-score 85.3; standardized mean difference versus placebo 0.97, 95% CI 0.37–1.57), IKDC score (P-score 88; SMD −0.71, 95% CI −1.20 to −0.21), and KOOS (P-score 87). Umbilical cord MSCs ranked first for WOMAC (P-score 91.4; SMD versus placebo 1.65, 95% CI 0.27–3.03).

Those rankings do not isolate the biology of the cells. Many adipose-MSC trial arms also used platelet-rich plasma or hyaluronic acid. When Tan et al. removed adjuvant-containing treatments, the result reversed. Their 2021 systematic review and meta-analysis included 19 studies and 440 knees treated with MSC injections alone. Bone marrow-derived MSCs outperformed adipose-derived MSCs by 1.50 VAS points (95% CI 0.04–2.96; P=.04) and 23.2 WOMAC points (95% CI 10.0–36.4; P<.01).

That does not establish bone marrow MSCs as universally superior. The subgroup dataset was modest, and the source-comparison confidence interval for pain was wide. It does show why source rankings from adjuvant-inclusive analyses deserve caution: platelet-rich plasma and hyaluronic acid can affect symptoms and adverse events, so they can inflate or obscure the apparent effect of the cell source. A 2022 source-comparison meta-analysis by Jeyaraman et al. also found that source-dependent results changed under more standardized treatment conditions.

Preparation may matter more than tissue source. In the adjuvant-free Tan analysis, cultured MSCs outperformed uncultured preparations by 2.19 VAS points (95% CI 0.57–3.81; P<.01) and 14.4 WOMAC points (95% CI 1.21–27.5; P=.03). Culture expansion may raise the efficacy ceiling, but the safety review places cultured adipose preparations in a higher-adverse-event category. The strongest symptomatic signal and the cleanest risk profile do not currently belong to the same product.

MSCs alone show a delayed, modest symptom benefit

The cleanest overall efficacy estimate comes from a 2025 RCT-only meta-analysis by Cao et al.. It included eight trials and 502 participants receiving standalone MSC injections without adjuvants or surgical co-intervention. MSC treatment significantly improved WOMAC at six and 12 months and also improved VAS and KOOS scores. The small number of eligible trials limits confidence in source and dose subgroups, but the design avoids crediting MSCs for a co-injected treatment.

Time point matters. A 2022 network meta-analysis by Ding et al. of 13 randomized trials found no significant WOMAC reduction at six months (95% CI −0.90 to 0.18; P=.19). At 12 months, it found improved KOOS scores (SMD 0.68, 95% CI 0.07–1.30; P=.029) and pain relief (SMD −1.05, 95% CI −1.46 to −0.64; P<.001). Cao's later analysis found a significant six-month WOMAC result, so the evidence is not perfectly uniform. Across analyses, the more consistent and interpretable signal appears at 12 months.

Statistical significance does not settle clinical value. Studies use different symptom scales, cell processing methods, osteoarthritis severity grades, and injection schedules. Cell doses range from 2×10⁶ to 1×10⁸, and trials used one to three injections over periods ranging from three to 48 weeks. These differences make a pooled standardized effect hard to translate into a promised improvement for one knee.

Dose and durability remain unresolved

Higher-dose adipose MSCs have looked better in post-hoc long-term subgroups in the Cao and Ding analyses, but neither establishes a minimum effective dose. A 2024 Phase I dose-escalation trial of umbilical cord MSCs by Matas et al. found safety across the tested doses without identifying a maximum tolerated dose. Phase I trials are designed primarily for safety, not to determine the dose that reliably improves pain or function.

Evidence beyond one year is thinner. Three studies in a 2024 systematic review maintained 24-month follow-up, while a three-year SVF cohort reported sustained functional gains. Neither observation can establish long-term treatment efficacy without adequately powered controlled follow-up. Symptoms can change for many reasons over three years, including exercise, weight change, regression to the mean, and additional treatment.

The practical verdict

MSC injections are not first- or second-line treatment for knee osteoarthritis on current evidence. Exercise-based physical therapy, weight management where relevant, and established injection options have a clearer place in routine care; a 2023 network meta-analysis of injections and physical therapy found physical therapy and combined approaches competitive with injection-only treatment for some outcomes.

For a well-informed person with moderate knee OA who has exhausted standard options, bone marrow-derived MSCs or uncultured SVF offer the most defensible risk-adjusted case. Their observed adverse-event rates were 10.7% and 8.1%, respectively, with no severe events in the largest safety review. Bone marrow MSCs also performed better than adipose MSCs in the adjuvant-free source comparison. Cultured adipose and umbilical cord products may produce larger symptom signals, but their complication rates require explicit discussion, especially because durable superiority over standard care remains unproven. Evaluate any benefit at 12 months, track pain and function before treatment, and treat cartilage-regeneration promises as marketing until randomized imaging evidence catches up.