Myopia that begins in the early school years tends to progress until the late teens, and the earlier it starts the higher it ends. The concern is less the spectacles than the eye itself. Axial elongation stretches the retina and choroid, and the lifetime risks of myopic maculopathy, retinal detachment and glaucoma rise with the degree of myopia, without a threshold below which the eye is safe (Flitcroft, 2012). Projections by Holden et al. (2016) that about half of the world's population may be myopic by the middle of the century gave the subject urgency, and the argument that each dioptre of progression avoided reduces later risk has made modest treatment effects seem worth having (Bullimore and Brennan, 2019).

Atropine is the oldest pharmacological treatment for progressive myopia and the best studied. At full strength it is effective but leaves children with dilated pupils and no near focus, which limited its use to a few health systems. The finding that much weaker solutions retained part of the effect with few symptoms changed practice in East Asia first and then more widely (Wu et al., 2019). Low concentration atropine is now prescribed in many countries, in some under a licence and in others as an unlicensed compounded preparation.

This review addresses the three issues named in its title. They are linked. A concentration that slows myopia strongly but is followed by a loss of the gain on stopping may be no better over a childhood than a weaker one, and a treatment that may run for many years needs a safety record of matching length.

The article is a narrative review of the published literature. It relies on the randomised trials that define the field, their extension reports, published meta-analyses and the consensus and assessment documents of professional bodies. It was not conducted as a systematic review, and it does not pool results. Low concentration is used here in the sense common in the literature, meaning 0.05% and below, with 0.01% as the concentration most often studied. Effect sizes are described in words. Readers who need the figures for a particular trial should consult the original reports, since trials differ in age at entry, baseline progression, ethnicity and the outcome on which they were powered.

Atropine is a non-selective antagonist at muscarinic acetylcholine receptors. The early rationale for its use was that it paralyses accommodation, which was thought to drive myopia. That explanation has been set aside. Atropine inhibits experimental myopia in the chick, whose ciliary muscle is striated and is not controlled by muscarinic receptors, so the effect cannot depend on accommodation. McBrien et al. (2013) review the alternatives: action at muscarinic receptors in the retina, choroid or sclera, effects on retinal dopamine release, and non-muscarinic mechanisms at the high tissue concentrations that topical dosing may produce. None has been established. The practical consequence is that concentration cannot be chosen from first principles and has had to be found by trial.

Figure 1 places the main trials in sequence. An early randomised comparison of several concentrations in Taiwan suggested that weaker solutions retained useful activity (Shih et al., 1999). The first Atropine for the Treatment of Myopia study (ATOM1) then showed in a placebo-controlled design that 1% atropine in one eye markedly slowed both refractive progression and axial elongation over two years (Chua et al., 2006).

Rebound means progression after cessation that is faster than would be expected in an untreated child of the same age. It was first documented after ATOM1. In the year after 1% atropine was stopped, treated eyes progressed faster than placebo eyes, and part of the earlier gain was lost (Tong et al., 2009). Some of the apparent refractive shift reflects the return of accommodative tone when cycloplegia wears off, but axial length data showed that true acceleration also occurred.

The washout phase of ATOM2 showed that rebound depended on concentration. Children who had used 0.5% or 0.1% progressed quickly after stopping, while those who had used 0.01% progressed least (Chia et al., 2014). When progressing children were then restarted on 0.01%, the group that had received 0.01% throughout ended five years with the least overall myopic progression, although axial length differences between groups were small (Chia et al., 2016). This result did more than any other to establish 0.01% in practice. It carries the limitation already noted, the lack of a contemporaneous untreated group across the five years.

LAMP examined the same issue with random assignment at the start of the third year to continue treatment or stop (Yam et al., 2022). Continued treatment was better than stopping at every concentration. Rebound after stopping was related to concentration but was small in absolute terms, and it was greater in younger children. The investigators concluded that stopping at a young age is inadvisable and that 0.05% remained the most effective regimen over three years even after allowing for rebound. How best to withdraw treatment is unresolved. Tapering the concentration or the frequency before stopping is widely practised and is reasonable on pharmacological grounds, but no randomised comparison of tapering against abrupt cessation is available to support a particular schedule.

Safety has to be considered on two time scales, the months and few years covered by trials and the decades that follow a childhood of treatment.

Four gaps stand out. The first is the response in children of European and African ancestry, where placebo-controlled results conflict; a large multicentre trial in the United Kingdom was designed to address this (Azuara-Blanco et al., 2020). The second is the choice between 0.01% and higher low concentrations outside East Asia, where 0.05% has hardly been tested against placebo. The third is the management of stopping: when, at what age, and whether to taper. The fourth is prevention. The LAMP2 trial found that 0.05% reduced the incidence of myopia in children who were not yet myopic (Yam et al., 2023), which raises the prospect of treating children before the condition is present and makes the absence of long-term safety data more pressing still.

Judging response in an individual child is harder than the trial reports suggest. Progression slows with age whether or not treatment is given, so a child whose refraction changes little during a year of drops may or may not have benefited, and one who continues to progress may still be doing better than without them. Axial length measured by optical biometry is less affected by cycloplegia and more repeatable than refraction, which is why its routine use in follow-up is widely advocated. Agreed criteria for an inadequate response, and evidence on what to do when one is found, are lacking.

Future trials would be more useful if they reported axial length as a primary outcome alongside cycloplegic refraction, stated the formulation in full, followed children after cessation, and continued observation into adulthood through registries. Combination with optical interventions is being studied and falls outside the scope of this article.

Low concentration atropine slows myopia progression in many children, with an effect that increases with concentration within the low range and that is clearer for refraction than for axial length at 0.01%. The evidence is strongest in East Asian populations and inconsistent elsewhere. Rebound after stopping is real, grows with concentration and with younger age, and is small at the low concentrations studied, so that continued treatment through the years of fastest progression is better supported than early cessation. Short-term tolerability is good. Long-term safety has not been shown to be a problem and has not been shown to be assured, because the necessary follow-up does not yet exist. Prescribers and families are therefore making a decision with a known near-term benefit and an unquantified distant risk, and the literature supports saying so plainly.

To be declared by the author(s) before publication.

To be declared by the author(s) before publication.