Do PRP injections work for knee osteoarthritis? Only against other injections

Whether PRP works depends entirely on what you measure it against, which explains most of the confusion around it.

Against hyaluronic acid, the gel injection that has been the standard alternative for years, PRP wins. Pooling 18 randomised trials covering 1,608 adults of average age 58, where patients were assigned to PRP or the gel by chance, knee scores for pain, stiffness and everyday function improved by 44.7% after PRP against 12.6% after the gel [4]. A separate pooling of 20 randomised trials put PRP ahead of the gel on pain at six and twelve months, and on function from the first month onward [5].

It beats the steroid injection too. Eight trials covering 648 adults of average age 59 found PRP better for pain, stiffness and function at three, six and nine months, with the gap widest at six and nine months [6].

Put PRP up against a plain saltwater injection, though, and the advantage disappears. The RESTORE trial gave 288 Australian adults aged 50 and over either three weekly PRP injections or three injections of saltwater, and after twelve months pain had fallen by 2.1 points out of 10 with PRP and 1.8 points with saltwater [1]. The researchers had decided in advance that anything under 1.8 points was too small for a patient to notice, and the gap between the groups was 0.4 points. Of 31 other measurements, 29 showed no difference [1].

A second trial injected 120 knees with steroid, gel, PRP or saltwater and found none of the three active injections beat the saltwater over six months, though those patients started with relatively little pain, which leaves less room to improve [2].

One trial using a saltwater comparison does report a benefit, and it followed people longest. It gave 237 adults one or three injections of PRP or saltwater, and the PRP groups scored better at three, six and twelve months, with three injections beating one [3]. By the two-year mark every group had converged and no difference was left [3].

PRP does not rebuild the joint. Researchers pooled 14 studies covering 1,099 people who had their knees scanned before and after injection and found no measurable gain in cartilage [7]. RESTORE measured the same thing directly: cartilage in the inner knee thinned by 1.4% with PRP and 1.2% with saltwater over a year, which is to say it carried on thinning in both groups [1].

Side effects look unremarkable. Across 20 randomised trials PRP caused no more problems than the gel injection [5], and a review of the randomised evidence described it as safe [16]. That safety record has not moved the guidelines. The American Academy of Orthopaedic Surgeons assessed 54 studies and marked where the evidence ran out alongside the potential harms and costs [10], and its 2022 guideline named PRP as an area needing better research [9]. A 2026 expert panel from the American Academy of Physical Medicine and Rehabilitation issued five evidence-based recommendations and eleven consensus best practices, while concluding that trials testing dose properly are still needed [8].

The working theory is that PRP calms the joint rather than resurfaces it. Platelets are the blood cells that clump together at a wound, and concentrated into an injection they release growth factors, the signalling proteins that tell surrounding tissue to repair itself, which act on the inflammation and tissue breakdown going on inside an arthritic knee [14].

One trial measured that directly. In 150 adults assigned by chance to PRP or the gel injection and followed for a year, two of the proteins that drive joint inflammation fell in the PRP group within a month and did not in the gel group [12]. The same trial found PRP patients could walk 120 metres further in six minutes without pain after a year, against 4 metres further for the gel group [12].

Whether PRP also pushes cartilage cells to multiply is contested, and results from laboratory dishes and animals have not carried over to people [14]. That is the most likely explanation for pain scores improving while the scans stay flat [7][14].

There is no standard recipe, which is why one clinic's PRP is not the same product as another's. Platelet concentration, whether the white blood cells are filtered out, and whether the platelets are activated first all vary between preparations, and that variation is a large part of why the trials are hard to compare [15].

The preparation may also explain the split in the results. RESTORE, which found nothing, used a shop-bought preparation with the white blood cells filtered out [1]. The two-year trial that did find a benefit used one with the white blood cells left in [3]. Pulling the other way, the pooled trial evidence scored the filtered version higher than the unfiltered one [4]. Nobody has tested the two head to head at a matched dose, which is the gap the 2026 expert panel said needs filling [8].

The big pooled averages and the best single trials disagree, and the pooled averages are built on the weaker studies. Reviewers describe the underlying trials as low to moderate quality, meaning their design leaves room for the result to be flattered, and they use preparations that differ from study to study [15][16]. "PRP" names a process, not a defined product [15].

The two largest poolings both compare PRP against the gel injection rather than against a dummy injection [4][5]. That shows PRP beats that particular alternative. It does not separate how much of the relief comes from the platelets and how much from the needle and the appointment around it, and when a trial does separate those, as RESTORE did, the advantage stops showing up [1].

Pain scores in these trials are what patients report, so they move with expectation. That is the main argument for weighting the saltwater comparisons above the gel ones.

Who this evidence does not cover:

• People with severe, bone-on-bone arthritis. Every trial here enrolled mild to moderate wear [1][13], and reviewers suggest younger patients with less advanced wear may be the ones who respond [16].
• Anyone hoping for structural repair. The scanning evidence says no [1][7].
• Other joints and other uses. This is knee osteoarthritis. Hip evidence amounted to three randomised trials as of 2017 [16], and nothing here speaks to shoulders, tendons or scalps.
• Anyone asking about year three. The longest follow-up here stops at two years [3], and most trials stop at twelve months [4][16].

No trial has shown that PRP slows the arthritis itself [11].

Longevity and regenerative clinics list PRP among their joint services, as do sports medicine and orthopaedic practices. A first visit looks the same everywhere: blood is drawn from the arm, spun to concentrate the platelets, and injected into the knee [12][14]. Single shots are not what the trials used. The four-arm trial gave two injections three weeks apart [13], RESTORE gave three a week apart [1], and the trial comparing one against three found three worked better [3].

If a clinic quotes a success rate, it is worth asking what the comparison was. A figure drawn from the gel-injection trials describes something real but flattering. The saltwater trials are the harder test, and PRP has not passed it [1][2][4].

Two other questions separate the trials that reported a year of benefit from the ones that reported nothing: how concentrated the platelets are, and whether the white blood cells are filtered out [4][12][15]. The trial that held the platelet count at 10 billion platelets per injection saw its benefit last a full year [12].