LED phototherapy: 71 randomised trials, and Cochrane's conclusion is 'no clinically significant effects shown'

A 2018 Cochrane review. It searched the Cochrane Skin Specialised Register, CENTRAL, MEDLINE, Embase, LILACS and ISI Web of Science, plus grey literature (to September 2015). Quality of evidence was assessed with GRADE.

Seventy-one randomised controlled trials with 4,211 participants were included.

The first number to look at is neither 71 nor 4,211.

The median sample size was 31.

That means half of the 71 trials had 31 participants or fewer. The total of 4,211 looks large, but it is 71 very small trials added together. It is the same problem we wrote about on the cysteamine page: many small trials is not several large ones.

And the results.

• 20% ALA photodynamic therapy, activated by blue light, versus vehicle plus blue light: little or no difference in effectiveness on participants' own assessment of improvement (266 people, low quality of evidence)
• MAL photodynamic therapy, activated by red light, versus placebo cream with red light: pooled across three studies, a similar effect on changes in lesion counts (360 people, moderate quality of evidence)
• Yellow light versus placebo or no treatment, infrared versus no treatment, gold microparticle suspension versus vehicle, and clindamycin/benzoyl peroxide plus pulsed dye laser versus clindamycin/benzoyl peroxide alone — none showed any clinically significant effects

One concentration-comparison study (180 people) did find that 20% ALA was no more effective than 15% but better than 10% and 5%.

Adverse effect reporting was inadequate. Scarring was reported as absent, and blistering only in the intense pulsed light, infrared and photodynamic therapy studies. The quality of evidence there was very low.

The authors' conclusion: carefully planned studies using standardised outcome measures and common acne treatments as comparators are needed.

Look again at that last clause. 'Common acne treatments as comparators' means the trials so far have not been measured against treatments already known to work. This site has pages for benzoyl peroxide and adapalene, and the evidence there is far firmer.

Closing this page after the item above would mean seeing half of it. Among those 71 trials is one that is well designed and came out positive.

A 2013 trial in Seoul. Thirty-five patients with mild to moderate acne were randomly assigned to two groups.

• Treatment: a home-use LED device delivering 420 nm blue light and 660 nm red light to the forehead and cheeks, twice daily for 2.5 minutes each, for four weeks
• Control: a sham device

The sham control is what matters. In this site's procedures layer, the trials with an inert control so far are IPL, PRP and deoxycholic acid. A sham keeps the sensation of using a device identical and takes away only the light.

At the final visit at 12 weeks:

• Inflammatory lesions down 77%
• Non-inflammatory lesions down 54%
• No significant change was observed in the control group
• No severe adverse reactions were reported

77% is a large number, and the control group not moving supports it.

But one condition has to be attached to how the abstract is read.

What the abstract reports is change within each group: the treatment group fell significantly and the control group did not change significantly. There is no statistic in the abstract directly comparing the two groups against each other.

Those are not the same thing. 'A changed significantly and B did not' does not automatically mean 'A was significantly better than B'. It is a distinction statisticians raise often, and it bites hardest in small trials like one with 35 people.

And it is 35 people — sitting right at the median of 31 from the item above.

Hence 'early evidence' for this item. The design is good, the size is small, and the between-group comparison cannot be checked from the abstract.

Histology from the same 2013 trial. Not a new study.

Observed in the treatment group:

• Reduced sebum output
• Attenuated inflammatory cell infiltration
• Decreased size of the sebaceous gland
• Reduced immunostaining intensity for interleukin-8, interleukin-1α, matrix metalloproteinase-9, toll-like receptor-2, nuclear factor-κB, insulin-like growth factor-1 receptor and sterol response element binding protein-1
• Messenger RNA expression of SREBP-1c was also decreased

The authors' phrasing was that the experimental results correlated well with the clinical results.

Why biopsy is valuable in this layer. Unlike photo scores or self-ratings, histology looks at what actually changed inside the skin. There is little room for expectation or lighting to enter.

The decrease in sebaceous gland size in particular means something for acne. Sebum output is one of the starting points of acne, and it overlaps with where drugs such as adapalene and benzoyl peroxide act.

Two cautions remain.

The abstract does not say how many people were biopsied. It will not have been all 35 — a biopsy removes a piece of skin, so trials do it in a subset. You have to read this without knowing how many people it represents.

And several molecular markers moving is not itself a clinical benefit. We drew the same distinction on the exosome and PLLA pages. What changed in the tissue and what shows in the mirror are separate questions.

This item reinforces the one above it; it does not replace it.