Start with the treatment that was actually studied
If you are reading about an injection for knee osteoarthritis, the first question is simple: what was in it?
The studies below tested several different preparations. Some used cells grown from a person's own fat tissue. Others used bone-marrow concentrate, a mixture of cells from fat, or cells from donated umbilical-cord tissue. Extracellular vesicles, or EVs, are small particles released by cells. An EV preparation is a different intervention again. [1] [3] [5]
That distinction matters when someone points to a study as evidence for a proposed treatment. A result belongs to the preparation, patients and procedure that were tested. It does not automatically transfer to another product with a similar name.
This article looks at selected human studies that make those differences clear. It is not a complete review of every knee osteoarthritis trial.
Cell trials have not all reached the same answer
The 2023 MILES trial randomized 480 people and reported its primary analysis in 440. It compared three cell preparations with corticosteroid injections. At 12 months, none was superior for the two main pain outcomes. The preparations were bone-marrow aspirate concentrate, adipose stromal vascular fraction and donor umbilical-cord-tissue mesenchymal stromal cells (MSCs). This was a comparison with an active treatment, not saline placebo. [1]
A published correction later restored two missing baseline rows in a table. The notice did not report a changed result or conclusion. [2]
A 2026 trial enrolled people with moderate-to-severe knee osteoarthritis. Researchers took a mixture of cells called stromal vascular fraction from each treated participant's own fat tissue and injected it into one knee that day. At one year, 43 of 83 people in the treatment group (52%) and 9 of 31 in the placebo group (29%) met the study's threshold for improvement. The measure combined pain and function scores from both knees, although only one knee received the injection. [7]
The trial estimates about 23 more responders for every 100 people treated. With only 31 people in the placebo group, that estimate is imprecise. The 95% confidence interval ranged from about 4 to 42 more responders per 100. This trial compared the injection with placebo. MILES compared its own fat-derived preparation with a corticosteroid. [7] [1]
A 2023 phase III paper reported enrolling 261 people with grade 3 knee osteoarthritis. The trial compared culture-expanded cells from each person's own fat tissue with placebo. At six months, pain and function improved more in the cell group. Follow-up data were available for 252 participants. MRI showed no significant difference between groups in the change in cartilage defects. That supports a symptom finding over six months, without establishing cartilage repair or lasting benefit. [3]
The 2025 ADIPOA2 trial also tested expanded cells from participants' own fat tissue, but found a different result. It randomized 135 people. Of those, 99 received treatment and 97 were analyzed. At six months, 47.3% of the cell group and 54.8% of the saline group met its predefined threshold for substantial pain or function improvement. The difference was not statistically significant. Secondary outcomes, including cartilage thickness, also showed no significant group differences. [4]
These trials differed in their preparations, participants and methods. Their results give us reason to ask more precise questions. They do not tell us, by themselves, why the findings differed.
What the controlled EV study found
A 2024 trial tested EVs derived from placental mesenchymal stromal cells. It enrolled 31 people with osteoarthritis in both knees. One knee received EVs and the other saline. The final analysis included 29 people, or 58 knees. At two and six months, the researchers found no significant difference in symptom scores. MRI findings at six months also did not differ. [5]
The paired-knee design had a practical weakness. Participants found it difficult to report walking and daily function separately for each knee. The study was also small and short. Those limitations matter, but they do not turn its negative result into evidence of benefit. [5]
Why an uncontrolled improvement needs a different reading
An October 2026 report reviewed 47 patients treated with two different EV products in one practice. Only 33 had paired scores for the main six-week comparison. Their average knee symptom score improved. [6]
There was no control group, and results from the two products were combined. Missing follow-up and other treatments further limit interpretation. The report can describe what was recorded after treatment. It cannot establish that the injections caused the improvement or that either product is effective. [6]
Pain relief, cartilage change and safety are separate questions
Feeling better matters. It is still a different outcome from showing a structural change in cartilage. A pain questionnaire cannot establish that cartilage regrew, and an MRI finding needs its own comparison and follow-up.
Safety also depends on how a study looked for problems. MILES reported no study-related serious adverse events, while recording problems including swelling and events associated with tissue collection. A small study without observed serious events cannot rule out uncommon or delayed harms. Retrospective records are more limited again because an event may not have been documented. [1] [6]
In a clinical conversation, ask to see the evidence for the specific preparation being proposed. Ask what the main outcome was, whether the comparison was placebo or another treatment, and how many participants actually completed follow-up. Those details help make an honest discussion possible.
Questions people ask
Do these studies show that injections regrow knee cartilage?
They do not establish that a proposed injection will regrow your cartilage. Symptom improvement and structural change must be evaluated separately. The positive 2023 adipose-cell trial found no significant between-group difference in cartilage-defect change at six months. [3]
Does the EV case series outweigh the negative controlled trial?
No. The products differed, and the case series had no comparison group. Its improvements cannot establish a treatment effect or cancel the controlled trial's findings. [5] [6]
What should I bring to a consultation?
Bring the exact product name and the original study being used to support it. Ask how closely its participants, preparation and follow-up match your situation, and what remains uncertain.
Sources and further reading
Sources checked November 9, 2026. Original full texts were checked for MILES, its correction, the placental EV trial and the case series. The Kim, ADIPOA2 and Garza summaries use their published abstracts. The Garza trial registry identifies the placebo injection and sponsor.
This article is educational and is not personal medical advice. A qualified treating clinician should assess your condition and discuss suitable options, expected benefits and risks. Any treatment decision should be made with a qualified physician. Individual results vary.