Two good studies, opposite conclusions, same treatment

EVIDENCE / WHY THE TRIALS DISAGREE

If you have spent an evening reading about platelet-rich plasma you will have found one trial saying it works and another saying it does not, both apparently competent. That is not a reason to give up on the evidence. It is a reason to know what to look at.

One: follow-up length decides the winner

A trial that stops at eight weeks reports that corticosteroid is superior. The same trial run to twelve months reports the opposite. Both are accurate answers to different questions, and headlines almost never say which was asked.

This is not a statistical artefact. It is the two treatments doing what they do: one suppresses inflammation and works quickly, the other provokes a repair and works slowly. The crossover at around three months is a biological fact, not a disagreement.

Two: PRP is not a standardized product

“PRP” names a category, not a preparation. Between trials, platelet concentration varies several-fold, leukocyte content is frequently unreported, red cell contamination differs, activation methods differ, and injected volume differs.

Pooling those into one meta-analytic estimate averages across interventions that are not the same intervention. The effect gets flattened toward the middle, and a real effect in a well-prepared subgroup disappears into a mediocre average.

Three: the control arm is rarely inert

A saline injection into a degenerate tendon still involves a needle passing repeatedly through the tissue. Needle fenestration converts a chronic non-healing lesion into an acute one, which is itself therapeutic — it is part of why the procedure works at all.

So a trial comparing PRP against saline is frequently comparing two active treatments and reporting the difference as nothing. That is a real finding about the increment, and it is routinely reported as though it were a finding about the treatment.

Four: the excluded patients are the ones who present

Randomized trials routinely exclude diabetes, higher body mass index, inflammatory disease, multiple painful sites and recent injections. Each exclusion is methodologically sensible and together they assemble a population younger, leaner and less complicated than the people who actually walk in.

Those exclusions remove precisely the patients whose tissue is failing for metabolic reasons. A literature built on the least affected produces a modest average, payers read the modest average, and the coverage decision follows. That loop is the subject of the coverage page.

What this does not license

It does not license concluding that the trials are wrong and your case is special. The excluded patients might do better than the literature suggests, because the mechanism is being addressed alongside the driver. They might equally do worse, because glycated collagen and chronic inflammatory load are exactly the conditions in which repair fails.

Both are plausible and the data does not settle it. Any clinic telling you confidently that the trials understate your odds is guessing in the direction that suits them.

How to read a study in four questions

• What was the control? No treatment, saline, or an active comparator.
• How long was follow-up? Under three months tells you little about a biological treatment.
• Was the preparation characterized? Platelet concentration and leukocyte content reported, or assumed.
• Who was excluded? The answer usually tells you whether the result describes you.

Those four questions resolve most of the apparent contradictions in this literature, and they are worth more than any single headline finding.

Why this matters more here than in most of medicine

For a drug with a large effect, trial heterogeneity is noise around a clear signal. For a treatment whose true effect is moderate and whose delivery varies between clinics, heterogeneity can be larger than the effect itself.

That is the honest position on orthobiologics. It is not that the evidence is fake or suppressed. It is that the evidence is genuinely noisy, the noise has identifiable sources, and a clinic that pretends otherwise in either direction — miracle or fraud — is not reading it carefully.

What would change our own position

Stated plainly, because a page criticizing other people’s certainty should say what would move it. Adequately powered randomized trials, with the preparation characterized well enough to reproduce, an injected control arm, follow-up beyond twelve months, and enrollment that includes the metabolically complex patients currently excluded.

Some of that work is underway. None of it helps someone deciding this month, which is why the assessment carries more weight here than the literature does.

What people ask about the evidence

So does PRP work or not?

For chronic degenerative tendon, the twelve-month evidence is reasonable. For cartilage regeneration, no. Where it performs and where it does not.

Why do meta-analyses look unimpressive?

They pool preparations that are not the same intervention. What varies between preparations.

Is a saline-controlled trial the gold standard here?

It is the right control and it understates the effect, because the needle is active. How this plays out against steroid.

Does my insurer read this literature?

It reads the pooled average, which is exactly the problem. Not covered is not the same as unproven.

Related reading

• Not covered is not the same as unproven
• How PRP works
• Am I a candidate?
• Information for referring physicians

The same four questions, applied to a clinic

The reading skills on this page work on marketing as well as on journals, and the questions translate directly.

• What is the comparator? A clinic reporting that most patients improve has not told you improved compared with what.
• Over what period? Six-week satisfaction data describes a flare settling, not a tissue changing.
• What is actually in the syringe? If a clinic cannot state platelet concentration and leukocyte content, it is not measuring what it sells.
• Who did they treat? Results from healthy athletes do not transfer to a sixty-year-old with an A1C of 7.8.

Applied honestly, those questions constrain what we can claim on this website too, which is the reason several pages here tell you not to have an injection. The same discipline decides what a scan is allowed to prove, which is set out on the page about normal scans.

Where we do not claim a benefit

A page about reading trials fairly should name the places where we decline to claim a benefit, and say why.

Midportion Achilles tendinopathy. Loading is the primary treatment here and an injection does not replace it. The comparison that decides the question is not against a needle; it is against a properly executed twelve-week program, and most people arriving for an injection have never had one. The Achilles page.

Cartilage regeneration. No available preparation reliably regrows it. Any before-and-after imaging offered as proof is measuring something else.

Large retracted rotator cuff tears. A structural problem, and biology does not fix structure. When the operation is the right answer.

BMAC over PRP in knee arthritis. Head-to-head comparisons have generally found no significant difference at a year, which matters because BMAC is more invasive and costs several times more. The comparison.

Preparation is real and it is not the largest term.