The study
A systematic review and meta-analysis of randomised controlled trials of platelet-rich plasma (PRP) for knee osteoarthritis, published in Archives of Orthopaedic and Trauma Surgery on 22 October 2026 by authors from Brazil.
It included only trials that measured joint structure on imaging. Pain and function weren't analysed. That gave 12 RCTs and 1,200 patients (572 received PRP), with a mean age of 59 and follow-up from 3 to 24 months. The review was registered on PROSPERO, assessed risk of bias with the Cochrane RoB 2 tool and rated certainty with GRADE.
Reading the numbers
Most results are standardised mean differences (SMD), which put different measurements onto one scale so they can be pooled. As a rough guide, 0.2 is a small difference, 0.5 moderate and 0.8 large.
The 95% confidence interval (CI) is the range the true average plausibly sits in. If it crosses zero for an SMD, or 1.0 for a risk ratio, no effect is possible.
I² describes how much the trials disagree with each other, from 0% (they agree) to 100% (they disagree a lot).
What they found
Patellofemoral cartilage thickness or volume: SMD 0.94 (95% CI 0.53 to 1.34), I² 0%. Low certainty.
Bone marrow oedema: SMD −0.40 (−0.71 to −0.09), I² 0%, driven by two trials. Low certainty.
Medial tibial cartilage thickness or volume: SMD 0.47 (−0.57 to 1.52), I² 92%. Very low certainty.
Cartilage defect progression: risk ratio 1.06 (0.47 to 2.38). Very low certainty.
Synovitis: SMD −0.22 (−0.43 to 0.00), with the upper limit sitting exactly on zero. Removing one trial made it significant in PRP's favour. Low certainty.
Osteophytes: mean difference −0.11 (−0.70 to 0.47) on the WORMS scale. Low certainty.
Subchondral cysts: mean difference −0.64 (−2.15 to 0.88) on the WORMS scale. Low certainty.
Seven trials were rated at low risk of bias and five had some concerns. None was rated high risk.
Some results couldn't be pooled and were described trial by trial. The largest saline-controlled trial was RESTORE, run in Melbourne and Sydney (144 people per arm, 12 months). It used leukocyte-poor PRP from a commercial Regen Lab kit, spun once at 1500g for 5 minutes, with 5 mL injected under ultrasound weekly for three weeks. Its main cartilage outcome, medial tibial cartilage volume, didn't differ from saline (−1.4% against −1.2%). As one of 31 secondary outcomes, more PRP patients had cartilage thinning in three or more areas: 17.1% against 6.8% with saline (risk ratio 2.71, 95% CI 1.16 to 6.34).
In a hyaluronic acid-controlled trial, 48.3% of PRP patients improved by at least one MRI grade against 8% of controls. Total WORMS score, a whole-joint MRI score, came from a single trial and didn't differ between groups.
What PRP was compared with
Three trials used saline as the only comparator, and four used hyaluronic acid. The rest used a mix: hyaluronic acid or an anti-inflammatory, PRP plus hyaluronic acid against hyaluronic acid alone, PRP plus exercise against exercise, microfragmented fat, and one trial with hyaluronic acid, saline and steroid arms.
How the PRP was made
The review doesn't tabulate leukocyte content, number of centrifuge spins, kit, volume or number of injections. The authors say these "varied among the included studies, often without complete reporting." One included trial compared single-spin and double-spin PRP within itself, and another used a single leukocyte-poor injection.
Funding and conflicts
The review had no funding. Brazil's CAPES paid the open-access fee, and the authors declared no competing interests. The review didn't report who funded each included trial.
What it means for your practice
The one consistent cartilage benefit was in the patellofemoral joint, and it pooled two different measures. Thickness and volume were combined into one result, which the authors list as a limitation. It's low-certainty evidence across mixed comparators. The tibiofemoral result was too inconsistent to read (I² 92%).
The authors read the bone marrow oedema result as a sign of less inflammation. They describe it as a marker of reduced inflammatory activity and say it shouldn't be read as evidence of cartilage preservation.
The largest, best-described saline-controlled trial found no cartilage benefit, and more thinning with PRP on a secondary measure. RESTORE reported its preparation in full, and its main cartilage outcome didn't move. The thinning result is a single secondary outcome, and it sits directly against the idea that PRP rebuilds cartilage.
Most trials didn't report enough about the PRP itself to tie a result to a preparation. Leukocyte-rich and leukocyte-poor PRP, one injection or three, are different treatments. Pooling them gives an average across products we can't name.
Follow-up was 3 to 24 months. Osteoarthritis progresses over years, so even the positive MRI findings can't say what happens to the joint long term.
The review measured structure only. Pain and function are a separate body of evidence, and this paper doesn't add to it.
Bottom line
Across 12 randomised trials, PRP improved patellofemoral cartilage measures and bone marrow oedema at low certainty, and showed no consistent effect on tibial cartilage, defect progression, synovitis, osteophytes or cysts. On this evidence, I wouldn't describe PRP to a patient as rebuilding cartilage.