Do Stem Cell Knee Injections Last? A Placebo-Controlled Trial Followed Patients for Five Years

Stem cell injections for knee osteoarthritis are sold on a promise of repair. The implicit claim is that unlike a steroid injection, which quietens a joint for a few weeks, or a course of platelet-rich plasma, which may last some months, cell therapy does something structural and therefore lasting. It is an appealing idea, and the price attached to it in private clinics reflects how appealing.

What has been missing is time. The randomised trials in this area are mostly six or twelve months long, which is long enough to see whether pain improves and nowhere near long enough to see whether it stays improved. A patient weighing up a four-figure or five-figure procedure is not really asking whether their knee will feel better next spring. They are asking whether it will still feel better in five years.

A trial published in April 2026 in Stem Cells Translational Medicine is one of the few that can speak to that. A team at the Cell Therapy Center at the University of Jordan ran a double-blind, placebo-controlled trial of allogeneic adipose-derived mesenchymal stromal cells in stage III knee osteoarthritis, then kept following the treated patients for 60 months. The answer it produced is more interesting, and less flattering, than either enthusiasts or sceptics would predict.

Patients with symptomatic stage III knee osteoarthritis were randomly assigned to one of two arms. Arm A received two injections of expanded allogeneic adipose-derived mesenchymal stromal cells, given two weeks apart under ultrasound guidance, at a mean total dose of about 69.58 million cells. Arm B received two equal-volume injections of normal saline. Allogeneic means the cells came from a donor rather than from the patient's own fat, which is how the doses could be expanded in a laboratory in advance and standardised across participants.

Neither patients nor assessors knew which arm they were in for the first 12 months. Outcomes were measured with two standard, validated knee questionnaires, the Knee Injury and Osteoarthritis Outcome Score and the Western Ontario and McMaster Universities Osteoarthritis Index, at 6, 12, 24, 36, 48 and 60 months. MRI scans were taken at baseline and again at 12 months to look for structural change in the joint rather than only for symptom change.

Twenty-nine people completed the study, 21 in the cell arm and 8 in the saline arm. After the 12-month blinded phase ended, follow-up continued for the cell arm alone, out to five years. That design decision is the single most important thing to hold on to when reading the results, and the next two sections explain why.

Within the blinded period the treatment performed well. Patients in the cell arm showed significant improvements in their clinical scores compared with baseline, and the differences were highly significant at 6, 12, 24 and 36 months, reported at a p value below 0.0001. In a condition where people typically expect slow deterioration, three years of better pain and function scores is not a trivial finding.

The MRI results add something that symptom scores cannot. At 12 months the imaging showed structural improvement in the treated knees, reported at a p value below 0.02. Imaging change is the closest thing this field has to evidence that something happened in the joint itself rather than only in how the joint felt, which is why it carries weight even from a small trial.

Safety was the least complicated part of the study. Adverse events in the cell arm were mild and transient, consisting mainly of localised pain and swelling after the injections. Nothing in five years of follow-up suggested a serious safety problem with the procedure as performed here, which matters because allogeneic donor cells raise reasonable questions about immune reactions.

Then the effect wore off. The improvements declined steadily after the 36-month mark, and by 60 months the scores had returned to a level close to where the patients had started. The authors state this plainly and draw the corresponding conclusion: the therapy appears safe and provides sustained clinical and structural benefit for up to three years, rather than indefinitely.

That is a genuinely useful result, and a more honest one than a shorter trial could have produced. A study that had stopped at 12 or 24 months would have reported an unambiguous success and left readers to assume the benefit continued. Following patients to five years converted an apparent repair into something closer to a long-acting treatment with a measurable half-life.

It is worth being careful about what the fade does and does not mean. Osteoarthritis is a progressive condition, so returning to baseline after five years is not the same as the treatment having done nothing: three years of meaningfully better function is three years of meaningfully better function, and by definition the untreated knee would have been expected to decline over that period rather than hold steady. But it does argue against the idea that a single course of cell injections permanently changes the trajectory of the joint, which is the idea most often used to justify the price.

This trial is small. Twenty-nine completing patients is a sample from which only large effects can be detected reliably, and the split between arms makes it smaller than it looks: 8 people in the placebo group is a very thin comparison. Small trials with unbalanced arms tend to produce unstable estimates, and a single centre running a single protocol cannot tell you how the treatment performs in other hands.

The bigger issue is the loss of the control group. Blinding was maintained for 12 months and follow-up beyond that point covered only the treated arm. Everything from 24 months onward is therefore a comparison against the patients' own baseline rather than against a placebo group, which is a much weaker form of evidence. Knee pain fluctuates, people change their activity and their painkillers, and expectation effects do not vanish when a trial stops being blind. Without a control arm running alongside, there is no way to separate the treatment's contribution from those influences during years two to five.

One more boundary is worth stating explicitly. This was allogeneic adipose-derived mesenchymal stromal cells, laboratory-expanded, injected into the joint under ultrasound guidance, in stage III osteoarthritis. It says nothing about same-day point-of-care preparations made from a patient's own fat or bone marrow in a clinic room, nothing about intravenous stem cell infusions, and nothing about exosome products. Those are different interventions and they need their own trials.

A single small trial should never be read alone, and here the surrounding literature is informative. A systematic review and meta-analysis published in Clinical Rheumatology in 2026 pooled 28 randomised controlled trials of intra-articular stem cell-based therapies in knee osteoarthritis. It found statistically significant improvements in several pain measures and in some function domains, with benefits appearing more consistent for laboratory-expanded preparations, bone marrow sources and ultrasound-guided injections. Crucially, it found no consistent structural benefit: MRI-based joint scores were not significantly improved. The reviewers concluded that these therapies serve a primarily symptom-modifying rather than structure-modifying role. They also found higher rates of injection-site pain and joint swelling than with control injections, though serious complications such as infection were uncommon.

A second meta-analysis, published in Frontiers in Medicine in 2025, asked a sharper question: how much of the improvement is the cells and how much is everything else? It pooled 8 randomised trials comparing stem cell injections against inert placebo in 467 patients and calculated the share of the treatment effect attributable to contextual factors, meaning the placebo response, the attention of the clinical team, the ritual of the procedure and the expectation of benefit. At six months contextual factors accounted for roughly 63 percent of the pain reduction and 61 percent of the functional improvement. At twelve months they explained about half of the pain relief and around two thirds of the functional gain. The authors rated the certainty of that evidence as low using the GRADE framework, and concluded that the cells themselves confer only a modest incremental benefit.

Put the three studies side by side and a coherent picture emerges. Something real happens when someone has this procedure, the effect on symptoms is genuine but mostly not attributable to the cells, the structural claim is not supported once trials are pooled, and whatever benefit exists appears to fade over a period of years rather than persisting.

The Irish context is small. Gerovia lists two providers offering stem cell therapy, both in Dublin, and anyone considering the procedure here is likely to be choosing between a short list of private clinics rather than a public pathway. That makes the questions you ask at the consultation more important than usual.

Four are worth writing down. What exactly will be injected, and is it a laboratory-expanded cell product or a same-day preparation made from your own tissue in the clinic? Under what regulatory basis is that product prepared and supplied, given that cell-based medicines in the European Union are tightly regulated and a clinic should be able to answer this without hesitation? Will the injection be ultrasound-guided, since the pooled analysis found benefits more consistent when it was? And what does the clinic tell you about duration of effect, because a clinic quoting permanence is quoting something the best long-term trial available does not support.

It is also fair to ask how this compares with the cheaper option. Platelet-rich plasma has a larger evidence base in knee osteoarthritis and costs a fraction as much, and the 2026 meta-analysis found that stem cell benefits were more consistent than uniform across preparations. Neither treatment is a repair. Both are best framed as symptom management for a progressive condition, sitting alongside the interventions with the strongest evidence in knee osteoarthritis, which remain weight management, structured exercise and strength work.

Rated on the strength of the evidence rather than the appeal of the finding, this sits at mixed. The trial is well designed for its blinded year, uses a placebo control and MRI, and reports five-year data that almost nobody else in the field has. Those are real virtues. But 29 completing patients, 8 of them in the placebo arm, and no control group after month 12 mean the long-term findings are observations rather than proof, and the wider meta-analyses point towards a symptom-modifying treatment whose effect is substantially contextual.

The most valuable thing this trial contributes is the shape of the curve. Benefit builds, holds for around three years, and then decays back towards baseline. Anyone selling a one-off procedure as a permanent fix for an arthritic knee is selling something this evidence does not show, and anyone dismissing cell therapy as doing nothing at all is ignoring three years of significant improvement and a structural signal on imaging.

As always, this is general information rather than medical advice. A painful knee deserves a proper assessment, imaging where indicated, and a conversation about the full range of options with a doctor who has no financial interest in which one you choose.