Vascular laser: a meta-analysis of 25 randomised trials put a different device above it

A 2022 single-blind randomised controlled trial at Yonsei University in 27 patients clinically diagnosed with rosacea.

• The two sides of each face were randomly assigned
• One side: long-pulsed alexandrite laser plus low-fluence Nd:YAG
• The other: PDL
• Four treatments at monthly intervals
• At every visit, erythema index (EI) was measured by skin analysis systems, and two independent dermatologists graded photographs on a five-point scale

One month after the fourth treatment (visit 5):

• Alexandrite side: 366.5 → 295.8 (P < .001)
• PDL side: 369.0 → 302.7 (P < .001)

Three months after the fourth treatment (visit 6):

• Alexandrite side: 360.3 → 282.0 (P < .001)
• PDL side: 364.3 → 281.6 (P < .001)

Both sides improved significantly, and the improvement held to three months.

But comparing the two sides against each other, there was no difference.

• Percentage EI reduction at visit 5: alexandrite 18.7%, PDL 16.4% (P = .501)
• At visit 6: alexandrite 21.7%, PDL 21.9% (P = .943)
• The physicians' grading and patient satisfaction were comparable between the two sides
• No serious adverse events occurred on either side

If you have been reading this site, the shape of that conclusion will be familiar. As the hyaluronic acid filler page sets out, almost every randomised trial in aesthetic medicine is device A versus device B, ending in 'no difference'.

There is something that design cannot answer. Whether both sides improved because of the lasers, or because rosacea itself waxes and wanes over four months — with no untreated side, the two cannot be separated.

Rosacea is by nature a condition that comes and goes. That makes the distinction matter more here, not less.

Note also that this was 27 people. The authors' conclusion was that alexandrite could be an alternative on the practical grounds that it requires no consumables — not a claim that it works better.

A 2026 network meta-analysis. It searched MEDLINE, CENTRAL and Web of Science for randomised controlled trials of lasers and energy-based devices in rosacea, with additional searches for ongoing trials. Twenty-five were included for qualitative analysis, and subsets fed the network meta-analyses.

Briefly, what a network meta-analysis is. An ordinary meta-analysis pools only trials that compared A with B directly. A network meta-analysis links A-versus-B and B-versus-C trials to estimate A versus C indirectly. Its strength is that it can line up devices that have never met head to head; its weakness is that the estimate depends on the quality of the links.

The results.

• Radiofrequency microneedling was more effective than PDL on patient satisfaction (mean difference −1.32, 95% CI −1.89 to −0.76) and erythema (mean difference −1.44, 95% CI −1.96 to −0.91)
• Oxymetazoline combined with PDL appeared superior for telangiectasia (mean difference −0.58, 95% CI −1.03 to −0.14)
• Adverse events and discontinuation rates were comparable across treatments

Neither confidence interval crosses zero, so statistically these are clear differences.

How to read this.

PDL is the device that comes to mind first for facial redness, in Korea as elsewhere. And yet when the randomised trials in this field were gathered and lined up, another device came out above it.

That other device already has a page here: microneedle radiofrequency. We built that page around scars and elasticity; that it also produced this result in rosacea is something we verified only now.

But a ranking is not a recommendation. Oxymetazoline is a topical drug, and the combination of it with PDL came out ahead for telangiectasia. The answer splits depending on what you are aiming at — redness, visible vessels, or satisfaction.

And the next item is the condition attached to this whole ranking.

Before taking the ranking above at face value, read what the same paper wrote about its own材料 — about its own material.

The authors' wording:

• Most studies exhibited unclear or high risk of bias.
• A slight publication bias was observed.

What each of those means.

High risk of bias means that things like whether randomisation was done properly, whether assessors were blinded, and how dropouts were handled were unclear or poorly handled. It does not mean the results are wrong; it means there is room for results to have come out better than reality.

Publication bias is more direct. It refers to the tendency for trials that found no effect to stay in the drawer rather than get published. Where that happens, a ranking built only from what was published is more optimistic than the whole.

There is a reason we made this a separate item.

Meta-analysis sits high on the ladder of evidence. 'A network meta-analysis of 25 randomised trials' sounds weighty, and it is. But a meta-analysis cannot be better than the trials it gathers. Pool 25 weak trials and you get a weak conclusion.

The authors wrote that down themselves, and we write their sentence at the same size as the ranking.

So this item is 'not established'. What is not established is 'which device is best'. What the two items above showed is that several devices lowered erythema index, and beyond that the data in this field cannot yet carry the weight.

One last thing: all the trials above were in rosacea. For other indications such as scars or port-wine stains, this page has nothing we have yet verified at source.