PRP: the masked raters found no difference; the participants did

A randomised split-face trial published in JAMA Dermatology in 2018. The design is worth setting out in full.

• 27 participants enrolled, 19 analysed, with photoaged skin
• 3 mL of PRP injected intradermally into one cheek
• Sterile normal saline into the other
• Both the participants and the raters were blinded to which side was which

Comparing the two halves of one face automatically controls for age, skin condition and lifestyle. And the comparator was not another product with ingredients in it, but saline, which does nothing.

Blinded dermatologists graded four things — fine lines, mottled pigmentation, roughness and sallowness — at two weeks, three months and six months.

The result: no significant difference, on all four measures, at all three time points.

How to read this.

Nineteen is a small number. A small real difference could have been missed (a type II error). So the trial did not prove that 'PRP does not work'.

But it does show one thing. On the other procedure pages you will have read, repeatedly, that 'product A and product B improved equally'. In that design you cannot tell whether both sides improved because of the treatment, or because a needle went in and time passed. Here the needle went into both sides equally, and only one side had platelets in it. And there was no difference.

Worth noting too: 27 enrolled, 19 analysed. Where the other eight went matters when reading the result.

This is the second result from the same trial. Not a new study — the same 19 people, the same faces.

At six months, participants rated the PRP side as significantly more improved.

The blinded doctors found no difference; the people themselves did. Same face, same two cheeks.

This is why this site labels the endpoint separately. The badge on the item above says 'graded by blinded assessors'; the badge here says 'self-reported'. Same trial, same people, opposite conclusions.

Which should you believe? We do not say either is false. But we do say plainly that the two endpoints measure different things.

• A blinded rater's score measures change that is distinguishable in a photograph.
• A self-rating measures what the person perceives. That includes expectation, memory and the day's lighting, along with the skin.

That the participants were also masked matters. It rules out the simple explanation that 'the expensive side looks better'. They did not know which side was PRP.

Which leaves two explanations: either there is real change that a doctor's eye does not catch but the person does, or over six months of comparing your own two cheeks, people tend to settle on a side. This trial cannot separate the two.

We record it as a result that makes you think about what satisfaction with an aesthetic procedure is actually made of.

There are two systematic reviews of PRP, and both are more favourable than the trial above.

The 2020 review searched the Cochrane Library, MEDLINE, EMBASE and Scopus from inception to March 2019 for prospective trials and case series with ten or more patients. Twenty-four studies qualified, eight of them randomised controlled trials, representing 480 patients who received PRP.

The conclusion: based on physician global assessment, PRP monotherapy by injection was shown to induce at least temporary, modest improvement in facial skin appearance, texture and lines. Periorbital fine lines and pigmentation may also benefit. The authors chose the words 'at least temporarily' and 'modest'.

The 2021 review and meta-analysis confined itself to the periorbital area, gathering nineteen studies, and meta-analysed patient satisfaction among the randomised trials. The authors describe it as the first PRISMA-compliant meta-analysis of periorbital PRP.

Now set that beside the item above.

The reviews are more favourable because of the character of the studies they gathered. Sixteen of the 24 were not randomised, and even among the eight randomised trials, inert comparators such as saline are rare. Compare before and after without a control and the passage of time, expectation and even the season all get credited to the procedure.

Note also that the endpoint the 2021 meta-analysis pooled was patient satisfaction — the very endpoint that, in the item above, pointed the opposite way from the blinded raters.

That is why this item sits at 'moderate'. There is a fair amount of human evidence, but the results are split, and they are not split at random: they split systematically by design and by endpoint. A large body of evidence is not the same as a body of evidence that agrees.

Same procedure, different answers on the face and the scalp. That is why this entry stands apart.

The 2016 trial (Dermatol Surg) was a randomised, placebo-controlled, double-blind half-head study in 25 people with androgenetic alopecia. PRP went into one half of the head and placebo into the other, three times, one month apart.

Six months after the first session:

• anagen hairs, telogen hairs, hair density and terminal hair density all differed significantly from baseline (p < .05)
• hair density was also higher on the PRP side than on the control side (p < .05)

The 2021 trial (J Cosmet Dermatol) was a randomised, placebo-controlled, double-blind half-head study at Southern Medical University in China, with 52 people who had androgenetic alopecia. At six months, mean hair count, density, diameter and the proportion of anagen hairs were all significantly better than on the control side.

So why nothing on the face and something on the scalp?

We will not claim to know. But one place where the designs part is plain to see.

• The facial trial's endpoint was a score given by masked doctors looking at photographs — fine lines, pigmentation, roughness, dullness.
• The scalp trials' endpoint was a count — hairs per cm², and their thickness.

The distinction this site keeps returning to shows up here intact. What can be counted registers small differences; a score given by eye does not. Whether the difference seen on the scalp was absent on the face, or present and unmeasured, these two pairs of trials cannot settle.

Why we did not attach 'strong evidence'.

By the rule as written, the conditions are met: two randomised controlled trials, an endpoint that is not self-assessment, no manufacturer paying. We held it down anyway.

• 25 people and 52. The trials behind that same badge on minoxidil had 352 and 381.
• Both are half-head designs. What was injected came out of platelets, so there is no guarantee the other side was untouched. If it was, the gap between the sides comes out smaller than it really is — which runs against PRP here, but the design is still not clean.
• Above all, the single word 'PRP' points at different things in different studies. Platelet concentration, spin conditions and activation method are not standardised.

Sometimes the rule would let something through and we hold it down anyway. This is one of those places.