Myopia control: strong evidence for high-dose atropine, weak evidence for low-dose

Several myopia-control treatments do slow how fast a child's short-sightedness worsens, but the certainty of the evidence runs almost exactly opposite to how widely each is used. Cochrane's living network meta-analysis, updated in February 2025 with 104 randomised trials covering 17,509 children, rates high-dose atropine and orthokeratology at moderate certainty — its highest grade anywhere in this field — while rating the low-concentration atropine that dominates real-world prescribing at very low certainty [s1]. Two large trials outside East Asia reached opposite conclusions about whether 0.01% atropine works at all [s2] [s3].

What the network meta-analysis found

The review included 104 studies randomising 17,509 children aged 4 to 18, with searches to 19 February 2024 [s1]. Two-thirds (66.3%) were conducted in China or other Asian countries and 14.4% in North America; 84 studies (80.8%) compared an intervention against an inactive control, and durations ran from 12 to 48 months [s1]. An important structural caveat sits underneath all of it: most of the networks were poorly connected, so the estimates rest on direct head-to-head pairwise comparisons rather than on the indirect inference a network meta-analysis is meant to provide [s1].

Untreated children in the control arms progressed by a median of −0.65 dioptres over one year and their eyes lengthened by a median of 0.33 mm [s1]. Against that baseline, the one-year mean differences in refraction were: high-dose atropine (0.5% or more) 0.90 D (95% CI 0.62 to 1.18), moderate certainty; repeated low-intensity red light 0.80 D (95% CI 0.71 to 0.89), very low certainty; medium-dose atropine (0.1% to under 0.5%) 0.55 D (95% CI 0.17 to 0.93), low certainty; low-dose atropine (under 0.1%) 0.25 D (95% CI 0.16 to 0.35), very low certainty; and multifocal spectacle lenses 0.14 D (95% CI 0.08 to 0.21), low certainty [s1].

On axial length — the measure that tracks the sight-threatening complications of high myopia rather than the strength of the prescription — orthokeratology reduced elongation by 0.18 mm (95% CI −0.21 to −0.14) at moderate certainty, one of only two moderate-certainty results in the review alongside high-dose atropine's 0.33 mm [s1]. Combining orthokeratology with low-dose atropine probably outperformed orthokeratology alone by a further 0.12 mm (95% CI −0.15 to −0.09), also moderate certainty [s1].

Two absences in that review deserve as much attention as the effect sizes. Adverse events and treatment adherence were not consistently reported across the trials, and evidence on rebound — whether progression accelerates after treatment stops — was limited [s1]. A treatment that slows progression for two years and then gives the ground back is a different proposition from one that banks the gain, and the trials mostly have not been designed to tell those apart.

The low-dose atropine problem

Low-dose atropine is the most commonly discussed pharmacological option, and it is where the Western trial evidence turns awkward.

CHAMP, a double-masked placebo-controlled phase 3 trial of a preservative-free formulation, ran from November 2017 to August 2022 at 26 sites in North America and five European countries [s2]. It randomised 576 participants aged 3 to 16 to placebo, 0.01% or 0.02% atropine in a 2:2:3 ratio; 573 (99.5%, mean age 8.9 years, SD 2.0) received treatment, and 489 (84.9%) aged 6 to 10 at randomisation formed the modified intention-to-treat set [s2]. Its prespecified primary endpoint was the proportion of eyes responding to 0.02% atropine — less than 0.50 D of progression at 36 months. That endpoint failed: odds ratio 1.77 (95% CI 0.50 to 6.26; P = .37), with no significant effect on mean refraction either (0.10 D; 95% CI −0.02 to 0.22; P = .10) [s2]. The 0.01% concentration, tested as a secondary endpoint, did hit all three of its marks: responder OR 4.54 (95% CI 1.15 to 17.97; P = .03), refraction 0.24 D (95% CI 0.11 to 0.37; P < .001), and axial elongation −0.13 mm (95% CI −0.19 to −0.07; P < .001) [s2]. A trial whose primary endpoint fails while a lower dose succeeds on secondary endpoints is a result to hold loosely, not a demonstration.

The Pediatric Eye Disease Investigator Group tested the same 0.01% concentration head-on. Its randomised, placebo-controlled, double-masked trial enrolled 187 US children aged 5 to 12 with −1.00 D to −6.00 D of myopia from 12 practices between June 2018 and September 2022, assigning them 2:1 to nightly atropine or placebo for 24 months [s3]. At 24 months the adjusted mean change in refraction was −0.82 D on atropine and −0.80 D on placebo, an adjusted difference of −0.02 D (95% CI −0.19 to +0.15; P = .83) [s3]. Axial elongation was 0.44 mm versus 0.45 mm, a difference of −0.002 mm (95% CI −0.106 to 0.102) [s3]. The authors' conclusion was unambiguous: these results do not support the use of 0.01% atropine eye drops to slow myopia progression or axial elongation in US children [s3].

Where that leaves the question

Two well-conducted trials in overlapping Western populations, using the same concentration, disagree. Cochrane's very-low-certainty rating for low-dose atropine is not fence-sitting; it is the accurate description of a literature that includes a clean null [s1] [s3]. The interventions with the firmest evidence are high-dose atropine, whose side-effect profile is the reason it is not in routine use, and orthokeratology, an overnight rigid contact lens that carries the infection risks of any contact lens wear and whose adverse events the trials did not report consistently [s1].

Red-light therapy ranks near the top of the effect-size table and near the bottom of the certainty table simultaneously — a 0.80 D difference at one year, rated very low certainty [s1]. That combination is exactly the profile of a treatment that could turn out to be genuinely effective or could turn out to be an artefact of small, short, mostly single-region trials, and the review does not resolve it.

This article describes published trial results and is not medical advice. Concentrations, dosing and suitability for any individual child are clinical decisions for an ophthalmologist or optometrist, and several of the interventions above are unlicensed or investigational in some countries.