Red Light for Myopia: Best-Supported Wavelength, With Limits

Repeated low-level red-light therapy has the best evidence among wavelength-specific light treatments for slowing myopia progression in children who already have myopia. In a 2024 Bayesian network meta-analysis, 650 nm red light was the only evaluated wavelength linked to statistically significant slowing of both axial elongation and refractive progression. That finding deserves attention. It does not prove that violet or full-spectrum light is biologically inferior, and it does not establish that red light has beaten orthokeratology in a fair head-to-head trial.

The distinction matters because myopia control aims to slow eye growth over years, not win a single measurement. Red light has a credible short-term signal. The unanswered questions concern the optimal protocol, durability, safety over childhood, and whether results from Chinese pediatric studies apply elsewhere.

Red light leads the wavelength ranking—for now

A 2024 Bayesian network meta-analysis by Zaabaar et al. pooled 12 clinical studies, including eight randomized controlled trials, of light-exposure therapies for pediatric myopia control. In the analysis that included all study designs, repeated low-level red light was the only wavelength with statistically significant results for both major outcomes.

• For axial length, red light reduced elongation by a mean difference of 0.38 mm per year versus control (95% credible interval, −0.59 to −0.16 mm per year).
• For spherical equivalent refraction, it slowed progression by 0.72 diopters per year (95% credible interval, 0.35 to 1.10 D per year).

Axial length is especially important because increasing eye length drives the structural change behind myopia. A reduction of a few tenths of a millimeter per year can be clinically meaningful. Still, the network meta-analysis did not show that violet or full-spectrum light fails biologically. Those approaches had fewer contributing studies and smaller total samples, leaving their estimates too imprecise to reach statistical significance. A sparse evidence base can make a treatment look like a loser when the honest answer is that researchers have not run enough good studies.

Red light therefore ranks first on current evidence, not on a settled map of retinal biology. Wavelength alone may not explain the result; irradiance, tissue penetration, session duration, and adherence could also matter.

The more conservative estimate is 0.28 mm per year

The same Zaabaar analysis contains the most useful brake on the headline number. When the authors limited the network meta-analysis to the eight randomized trials, the axial-length effect fell from −0.38 to −0.28 mm per year (95% credible interval, −0.40 to −0.15). The refractive effect fell from +0.72 to +0.57 D per year (95% credible interval, 0.22 to 0.92).

Red light remained statistically significant in the randomized-only analysis, which supports a real treatment signal. But excluding non-randomized studies reduced the estimated axial-length benefit by about 26% and the refractive benefit by about 21%. For families and clinicians considering the treatment, the randomized-trial estimate—about 0.28 mm less axial elongation per year—is the more credible benchmark than the larger mixed-design figure.

Several independent reviews reach the same general conclusion. Meta-analyses published in 2023 through 2025, including reviews by Wang et al., Tang et al., Deng et al., and Fan et al., reported pooled reductions in axial elongation of roughly 0.20 to 0.38 mm per year and slower spherical-equivalent progression of roughly 0.50 to 0.72 D per year. The range reflects differences in included trials and follow-up, not a menu of interchangeable promises.

How does it compare with atropine and orthokeratology?

A larger 2025 network meta-analysis by Zheng et al., drawing on 41 randomized trials, ranked repeated low-level red light above low-dose atropine and orthokeratology for reducing axial elongation and refractive progression. That is meaningful corroboration because it compares more active treatments and rests on a much larger randomized evidence base than the wavelength-specific analysis.

Yet network meta-analysis is not a direct contest. It estimates comparisons through a connected web of different trials, populations, and protocols. If one trial compares red light with a control and another compares orthokeratology with a control, the red-light-versus-orthokeratology estimate is indirect. Such estimates can guide decisions, but they cannot substitute for a randomized head-to-head trial.

The only direct comparison described here is a 2024 retrospective Chinese cohort study by Sun et al. of 138 children aged 6 to 14 years. Over two years, children receiving red light had mean axial growth of 0.10 mm per year (95% CI, 0.07 to 0.14), compared with 0.25 mm per year (95% CI, 0.24 to 0.27) among children using orthokeratology lenses. The aggregate difference was 0.15 mm per year in favor of red light (95% CI, −0.17 to −0.12; p<0.001).

That aggregate result sounds decisive until the year-by-year results enter the room. In year 1, axial growth was 0.03 mm with red light and 0.28 mm with orthokeratology (p<0.001). In year 2, it was 0.14 versus 0.21 mm, and the difference did not reach statistical significance (p=0.06). The study cannot tell us whether red-light benefit waned, adherence declined, baseline differences between self-selected treatment groups distorted the early result, or chance produced the pattern.

The cohort points toward a red-light advantage over orthokeratology. It does not establish one.

Because the study was retrospective and non-randomized, treatment selection could have tracked factors that also affect myopia progression. Red-light session compliance was not verifiable, whereas lens wear is more readily documented. The analysis also lacked a formal sensitivity analysis for unmeasured confounding. The two-year average should never travel without those qualifications.

Choroidal thickening is a marker, not a completed mechanism

Researchers have proposed that 650 nm light changes retinal energy metabolism, dopamine signaling, and choroidal tone, which could resist axial elongation. The molecular chain remains a hypothesis from preclinical photobiology. Human studies in this evidence base do not provide retinal biomarker confirmation of mitochondrial activation or dopamine release.

There is stronger evidence for a downstream structural change. In a prospective observational study by Ruilin Xiong et al., children receiving repeated low-level red light showed increasing macular choroidal thickness over 12 months, and an early increase predicted later treatment efficacy. A 2026 retrospective machine-learning analysis by Hu et al. likewise identified axial and choroidal changes as key predictors of treatment response.

These findings make choroidal OCT a plausible monitoring tool, especially if clinicians use red light. They do not prove that choroidal thickening causes the treatment effect, much less validate the proposed molecular pathway. A predictive biomarker can be useful without being the engine under the hood.

Safety and scope set the practical boundary

Meta-analyses have reported no serious adverse events during trial follow-up of up to two years, including the safety-focused reviews by Youssef et al. and Deng et al. That is reassuring within the window studied. It does not answer whether repeated retinal exposure over five to ten years of childhood has cumulative effects, or whether trials used standardized imaging surveillance capable of detecting subclinical photoreceptor stress.

Protocol variation adds another problem. Trials used irradiance of roughly 1 to 3 mW/cm², differed on weekday-only versus daily schedules, varied in compliance monitoring, and used different minimum ages. Most randomized studies used a single commercial EYERISING/Sightful system. The evidence supports that family of protocols more directly than it supports every red-light device sold with a reassuring shade of crimson.

Generalizability remains uncertain because all constituent randomized trials in the two network meta-analyses were conducted in Chinese pediatric populations. Ambient light exposure, baseline myopia risk, and population differences could affect real-world results elsewhere.

Finally, this is myopia control evidence, not prevention evidence. The red-light trials enrolled children with established myopia. A 2025 prevention network meta-analysis by Lee et al. addresses at-risk children before myopia develops, a fundamentally different population. No randomized evidence supports using repeated low-level red light to prevent myopia onset in pre-myopic children.

What families can reasonably conclude

For a child with established myopia, repeated low-level 650 nm red-light therapy is a legitimate option with randomized evidence of slower axial elongation and refractive progression. The best calibrated estimate from the wavelength network meta-analysis is about 0.28 mm less axial elongation per year, not the larger 0.38 mm estimate that includes non-randomized studies.

It may rank ahead of atropine and orthokeratology in indirect comparisons, but direct randomized evidence against orthokeratology is still absent. The one direct cohort favored red light overall but found no statistically significant axial-length difference in year 2. Treatment should therefore involve an eye-care clinician, careful adherence tracking, and a discussion of monitoring that may include choroidal OCT. The evidence supports slowing progression in already-myopic children over the short term; it does not yet support prevention claims, decade-long safety claims, or confidence that one red wavelength has solved myopia control.