Platelet-rich plasma (PRP)

PRP concentrates platelets from your own blood and reinjects them into the joint, where degranulation releases growth factors — PDGF, TGF-β, VEGF, IGF-1 — that promote chondrocyte proliferation and calm synovial inflammation in the lab. Whether a transient growth-factor bolus changes cartilage biology in a loaded human knee is unproven: the one RCT that measured cartilage on MRI found no structural effect.

The RESTORE trial

RESTORE (JAMA, 2021) is the most rigorous test PRP has faced: 288 patients, three weekly leukocyte-poor injections versus saline, double-blind, with 93% retention and blinding the trial measured and found intact. At 12 months it was null on both primary endpoints — pain change −2.1 versus −1.8 (P=.17), medial tibial cartilage volume −1.4% versus −1.2% (P=.81) — and the pain improvement cleared the 1.8-point threshold for a clinically important change in both arms. Two of 31 secondary outcomes separated, both of them self-rated global improvement favouring PRP. No structural secondary favoured PRP, and inside one of them more PRP participants had three or more areas of cartilage thinning: 17.1% versus 6.8%.

The wider placebo-controlled literature is kinder but not by much. A meta-analysis of 11 saline-controlled RCTs (1,616 patients) found a small pain benefit at 3 and 6 months that disappears by 12; the authors call its clinical relevance debatable. The standout positive trial is Patel 2013, where a roughly 10× concentration, WBC-filtered preparation beat saline on every WOMAC parameter — with one injection as good as two and benefit fading after 6 months.

Against hyaluronic acid, and the preparation question

Against hyaluronic acid, PRP wins consistently: across 18 level-1 RCTs, PRP patients improved 44.7% on WOMAC versus 12.6% with HA, and a 5-year double-blind follow-up found fewer reinterventions with PRP at 24 months. The catch: HA itself performs near placebo (see the hyaluronic acid entry), and intra-articular saline outperforms oral placebo in network meta-analysis — beating HA is a low bar.

Preparation seems to matter. In network meta-analysis, leukocyte-poor PRP improved WOMAC versus both HA and placebo while leukocyte-rich PRP separated from neither — though this rests entirely on indirect comparison, and RESTORE used leukocyte-poor PRP and was still null.

The high-dose question

Dose is where the live question sits, and it is why this appears on the evidence map as "high-dose PRP". What counts is absolute platelet count per course, not the concentration multiple. A review of 29 RCTs found positive trials averaged ~5.5 billion platelets versus ~2.3 billion in null trials; a dose meta-analysis puts the favorable threshold near 10 billion, clearer for function than pain. Dose advocates note that RESTORE's commercial kit is rated at 1.6 to 5 times whole-blood platelet concentration and call its platelet dose subtherapeutic.

The status of that argument: the dose-response evidence is ecological — comparisons across trials, not within them. RESTORE also assayed the preparation it gave, and the healing growth factors and cytokines the dose argument runs on were elevated in it. The only trial designed around an absolute dose (Bansal 2021, single 10-billion-platelet injection, benefit sustained at 12 months while the comparator faded) used HA as the comparator, not saline, and carries a published correction. A three-injection protocol beat single PRP in early OA, which fits cumulative dose but is confounded with injection frequency. Until a randomized dose-ranging trial against saline exists, "high dose fixes the null trials" is a hypothesis, not a finding.

Practical notes

Trials used 1–3 injections, usually weekly; benefit, where present, peaks at 3–6 months and wanes by 12. If following the dose hypothesis, the literature's target is ≥5–10 billion platelets per course. Costs are out-of-pocket in most systems, and AAOS gives PRP only a limited recommendation.

Safety

Autologous blood, so the main events are transient post-injection pain and swelling — more frequent with leukocyte-rich preparations (adverse-reaction odds ratio 5.63 versus HA, mostly self-limited). No serious safety signal in RCTs. Avoid NSAIDs around the injection window given theoretical platelet inhibition, and use caution with active infection, platelet disorders, or anticoagulation.