Rebound: what happens to myopia when treatment stops?

Most myopia-control trials answer one question: how much slower does myopia progress on treatment? Far fewer answer the question parents in Hamilton ask us almost every week: what happens when treatment stops? A new systematic review and Bayesian network meta-analysis in Ophthalmology and Therapy, published online on 7 September 2026, tries to put on-treatment efficacy and post-treatment "rebound" side by side across the whole evidence base. This article is based on the published abstract; full-text details such as effect sizes, confidence intervals, risk-of-bias ratings, funding statements and the definition of "clinically meaningful" rebound were not available to us.

What the review did

The authors searched for randomised controlled trials of interventions for paediatric myopia and identified 117 RCTs covering 24 active interventions. Twenty-two RCTs and one study arm were excluded from the quantitative synthesis because outcomes were missing or because the trials violated the "transitivity" assumption that network meta-analysis relies on (more on that below). Outcomes were change from baseline in spherical equivalent refraction (SER) and axial length (AL) after one or two years of treatment, and again 12 months after treatment stopped.

Two design choices are worth noting. First, the analysis was Bayesian and hierarchical, with a "dose-response ordering constraint": the model assumes that a higher atropine concentration cannot be less effective than a lower one, which improves precision but also builds a dose-response shape into the results. Second, the review defined low-dose atropine (LDA) as any concentration below 0.1%, which spans the 0.01%, 0.025% and 0.05% arms we discussed in our earlier post on the LAMP low-dose atropine trial.

What the abstract reports

• LDA concentrations of 0.01% and above significantly slowed myopia progression compared with an inactive control over two years, in a dose-dependent manner.
• LDA showed numerically larger but statistically non-significant effects compared with optical interventions (the optical options are not itemised in the abstract).
• Moderate-to-high dose atropine and repeated low-intensity red-light (RLRL), alone or combined with orthokeratology, showed greater short-term efficacy. However, 12 months after cessation these interventions were associated with clinically meaningful rebound exceeding the progression seen in untreated controls.
• LDA was not associated with clinically meaningful rebound.
• Combining 0.01% atropine with orthokeratology improved efficacy relative to LDA monotherapy over two years, but post-treatment rebound data were unavailable for the combination.
• Meta-regressions indicated baseline SER, baseline progression rate, geography and race as effect modifiers, albeit with borderline significance.

The authors conclude that, within the current evidence base, atropine concentrations of 0.01% to 0.05% showed sustained on-treatment benefit with minimal post-treatment rebound.

A note on the language. "Statistically significant" means a result would be unlikely to arise by chance if there were truly no difference; "non-significant" means the data cannot rule out no difference. It does not mean the treatments are equivalent. Without the underlying estimates and their 95% credible intervals (the Bayesian equivalent of a confidence interval, the range within which the true effect most plausibly sits), we cannot say how wide that uncertainty is.

Why rebound changes the calculus

The clinically useful message is the trade-off. A treatment that slows progression strongly for two years but is followed by a year of accelerated progression may leave a child's eye no better off than a treatment with a gentler on-treatment effect that holds after stopping. The abstract puts higher-dose atropine and red-light therapy in the first camp and low-dose atropine in the second. That is consistent with what we flagged in our review of red-light therapy and its safety caveats: strong short-term axial-length results, with the durability and safety questions still open.

A relative percentage reduction over one or two years is a snapshot; what matters is the cumulative difference once the child is out of the progression window. Rebound data are the missing half of that calculation.

Who wrote it

The abstract itself contains no funding or conflict-of-interest statement, so we cannot confirm who paid for the analysis. What is visible in the PubMed author record is that five of the seven authors are affiliated with IQVIA, a contract research and consulting organisation, and two are affiliated with Santen, a pharmaceutical company that markets a low-dose atropine product for myopia in several markets. This is, in other words, an industry-authored review whose headline conclusion favours low-dose atropine. That does not make the analysis wrong, but readers should apply the same scrutiny we ask for in every sponsored trial: who chose the outcome definitions, who set the "clinically meaningful" threshold, and who decided which trials were excluded.

Limitations of this evidence

• Abstract only. No effect sizes, credible intervals, total number of children, or risk-of-bias assessments were available to us. "Significant" and "clinically meaningful" are used without the thresholds behind them.
• Rebound data come from a minority of trials. Most myopia RCTs have no washout phase. The 12-month post-cessation comparisons will rest on a small subset of studies, and the abstract does not say how many.
• Industry authorship. Authors are affiliated with IQVIA and Santen; funding could not be verified from the abstract.
• Transitivity. Network meta-analysis assumes trials are similar enough to be compared indirectly. The authors excluded 22 trials for missing outcomes or transitivity violations, and flagged geography, race, baseline refraction and baseline progression as modifiers, which is a reminder that most atropine data come from East Asian cohorts and may not transfer directly to Hamilton children.
• Modelling assumptions. The dose-response ordering constraint improves precision at the cost of building in the assumption it then appears to confirm.
• Safety. The abstract mentions safety in the search but reports no adverse-event results.
• Proxies. Refraction and axial length are surrogates; no trial has yet followed children to the pathology that matters.

What we do with this in practice

For the optometrist team at Rose Optometry, the review does not change the menu but it does sharpen the conversation. When we discuss myopia-control options at Rose Optometry with families, we now talk explicitly about what is known about stopping each option, not only about starting it. Where rebound is documented, that means planning the exit before the first bottle or lens is dispensed, and continuing axial-length measurement after cessation rather than assuming the benefit is banked. Families who want to talk this through can book a free Myopia Chat with Jagrut.

Jagrut Lallu — BOptom (Hons), MSc Specialty Lenses (Hons), FIAOMC. Therapeutic Optometrist & Contact Lens Specialist. Founder, New Zealand Eye Research Centre; Partner, Rose Optometry; WCO Asia-Pacific Myopia Ambassador; International Myopia Institute NZ Ambassador; Clinical Senior Lecturer, Deakin School of Optometry; Honorary Teaching Fellow, University of Auckland.

Reference

• Jovanović J et al. Low-Dose Atropine and Other Interventions for Pediatric Myopia: A Systematic Review and Bayesian Network Meta-analysis of Efficacy and Post-treatment Outcomes. Ophthalmol Ther. 2026 (published online 7 September 2026). PMID 42704413. https://doi.org/10.1007/s40123-026-01478-y