Amblyopia after ten: what the trials of treatments beyond patching found

Amblyopia is usually explained in two words: "lazy eye". The phrase is convenient and almost entirely wrong. The eye is not lazy — during development the visual cortex receives input of unequal quality from the two eyes and gradually stops relying on the weaker one. That is why acuity in one eye stays lower than the state of the eye itself would explain, even in glasses, and why treatment is addressed to signal processing rather than to the eye.

Pooling 97 population studies covering 4,645,274 children, the estimated prevalence of amblyopia in childhood is 1.36% (95% CI 1.27–1.46%). The authors note significant publication bias, so the figure should be read as an order of magnitude rather than a precise number (meta-analysis, PMID 35601430).

Parents of a child older than seven are usually told two opposite things. First: the critical period has closed, you should have come earlier. Second: there is now a method that changes everything — a headset, a game, a course of exercises. Both statements sound confident, and both need checking. What follows is what the trials found in this specific age group: who took part, what was measured, and what these studies do not say.

"After seven it is too late" — what the largest trial in this age group found

The largest trial comparing treatment with optical correction alone in children older than seven was run by the Pediatric Eye Disease Investigator Group across 49 sites. It enrolled 507 children aged 7 to 17 with amblyopic-eye acuity from 20/40 to 20/400. All were first given optimal optical correction, and only then randomised: one group to patching 2–6 hours a day combined with near activities (children aged 7–12 also received atropine), the other to optical correction alone. A response was defined as an improvement of 10 letters or more (at least two lines) by week 24.

Among children aged 7–12 (n = 404), 53% responded in the treatment group versus 25% with correction alone. Among teenagers aged 13–17 (n = 103) the response rates did not differ overall (25% versus 23%), but among those never previously treated with patching or atropine they were 47% versus 20%. The authors state plainly that most participants, responders included, were left with a residual acuity deficit (PMID 15824215).

Two conclusions follow, and a third does not. First: age seven is not a line beyond which change stops. Second: among teenagers aged 13–17 the response rates diverged specifically in those never treated before (47% versus 20%) — but that is a subgroup result, not proof that prior treatment is what decides the matter: across the whole stratum there was no difference. What the trial does not say is that treatment at this age restores normal vision — "response" here means a measured gain of two lines, not a normal eye.

Does the gain hold after treatment stops

This is a separate question from the size of the gain, and the answers differ by age.

Eighty children aged 7–12 from the same trial whose acuity had improved by at least two lines were followed for a year off treatment, glasses aside. The cumulative probability of losing two lines or more was 7% (95% CI 3–17%), and 82% retained a gain of at least ten letters relative to their pre-treatment level (PMID 17502505).

In children under eight the picture differs. In a cohort of 145 successfully treated with patching or atropine, amblyopia recurred within the first year off treatment in 24% (95% CI 17–32%). One detail from the same observation: among those on 6–8 hours of daily patching, recurrence occurred in 42% when treatment stopped abruptly and in 14% when it was first reduced to two hours — an observational comparison within the cohort, not a randomised one (PMID 15492733).

The step most often skipped: glasses alone

In the trial of children aged 7–17, a quarter of the participants assigned to optical correction alone reached the same two-line response — without patching, atropine or exercises. That is not an argument against the rest of treatment; it is an argument for counting the baseline after a period of full correction rather than from the first number measured at a visit.

A systematic review with meta-analysis of optical treatment included 29 papers, 20 of them with data usable for effect size. The effect was moderate to large in younger and older children alike, and in refractive as well as strabismic amblyopia; meta-regression showed it decreasing with age and increasing with the duration of correction. Heterogeneity was very high (I² = 96.7%), so this speaks to the direction and order of the effect, not to a precise magnitude (PMID 29392811).

The practical consequence is simple: any new method is worth comparing only against the acuity measured after the child has worn correct glasses consistently for some time. Otherwise the improvement produced by glasses is credited to the method.

Patching and atropine in children older than seven

The two classical methods have been compared head to head in this age group: 193 children aged 7–12 with moderate amblyopia (20/40–20/100) received either weekend atropine or two hours of daily patching. Over 17 weeks acuity improved by an average of 7.6 letters with atropine and 8.6 letters with patching, a difference within the pre-specified equivalence limits. Acuity of 20/25 or better was reached by 17% and 24% respectively (PMID 19064841).

A Cochrane review comparing occlusion with pharmacological penalisation reached the same general conclusion: both improve amblyopic-eye acuity, atropine appears as effective as occlusion, and the size of the improvement varies between individual trials (PMID 31461545). One caveat: the review pooled seven trials in which most participants were younger than seven, so it supports the general finding without being separate evidence for ages 7–12.

There is also a study in an even older group, though small and single-centre: 57 patients aged 8–20 with anisometropic amblyopia received either full-time patching or atropine. At six months acuity had improved by 2.38 and 2.34 lines respectively (p = 0.889), with faster recovery under patching — 3.7 versus 4.7 months (PMID 18534880). The sample does not support claims about effect size, but the direction matches the data in 7–12-year-olds.

Binocular treatment: where hope and evidence part company

The idea behind binocular (dichoptic) treatment is elegant and physiologically sensible: instead of switching the stronger eye off, each eye is shown its own part of an image at a different contrast, so the cortex has to assemble the picture from both signals. Trials of that idea give results that depend on age, and the dependence runs opposite to what is usually hoped for.

Ages 5 to under 13. 385 children, a binocular tablet game for one hour a day versus patching two hours a day, 16 weeks. Acuity improved by 1.05 lines in the game group and 1.35 lines in the patching group; the result did not stay within the pre-specified non-inferiority limit. More than 75% of the prescribed game time was completed by 22% of the children, with a median of 46%. In the youngest, previously untreated participants (5–7 years) gains were similar in both groups (PMID 27812703).

Ages 7–12, previously treated. 138 children, a dichoptic game for eight weeks versus continued spectacle wear. There was no difference between groups at four weeks or at eight (PMID 30352226).

Ages 13–16. 100 teenagers, one hour of the game a day versus two hours of patching, 16 weeks. The gain was 3.5 letters in the game group and 6.5 letters with patching; after adjustment for baseline acuity the difference was −2.7 letters in favour of patching. More than 75% of the prescribed game time was completed by 13% of participants. The authors state directly that these data cannot separate a weak treatment effect from weak adherence (PMID 29196184).

Age 7 and above, including teenagers and adults. The double-masked BRAVO trial: 115 participants aged 7 and older (mean age 21.5 years; 77% with prior occlusion) played at home for an hour a day for six weeks, either a dichoptic game with different contrast for each eye or a placebo game in which both eyes saw the same image. Acuity improved by 0.06 logMAR in the active group and 0.07 logMAR in the placebo group; the difference adjusted for baseline acuity and age stratum was −0.02 logMAR (95% CI −0.06 to 0.02, p = 0.25). There were no differences in secondary outcomes either, including stereoacuity and suppression (PMID 29302694). Unlike the trials above, the comparison here was not against patching but against a dummy — and no difference was found. It is the largest double-masked multicentre comparison of a dichoptic game against a placebo game; smaller trials of the same design exist too.

Review. The Cochrane review of binocular treatment in children aged 3–8 found a single eligible randomised trial, contributing 68 children to the analysis. The acuity gain at 16 weeks was clinically comparable to patching; the authors rated certainty as moderate and emphasised that there are no data on whether the result holds at one year (PMID 35129211).

The professional society's summary. The American Academy of Ophthalmology assessment of binocular treatment included 20 studies. Improvement over standard treatment was shown by two level I–II studies enrolling 147 patients in total; the five studies that showed no such improvement were larger and more rigorously designed, enrolling 813 patients in total. Its conclusion: there is no level I evidence supporting binocular treatment as a substitute for patching and optical treatment, and it cannot be recommended as a replacement for standard therapy (PMID 31619356; literature search updated April 2019). The date matters: the review closed before the regulator authorised the two products discussed below, so it says nothing about them. A later document from the same academy — its amblyopia practice pattern, updated in February 2024 — does list the authorised digital therapeutics among the available options. That does not overturn the earlier conclusion; it refines it, because "does not replace standard treatment" and "may be used" are different statements.

Adherence in these trials is not a technical footnote but possibly the main finding: in two of them only a minority completed the prescribed game time, and the authors name this as a possible explanation of the weak result. No general rule that older participants comply less follows from these numbers: in the Dig Rush trial in 7–12-year-olds adherence was markedly higher than in the other two.

What the regulator authorised, and on exactly what data

A regulatory authorisation is permission to market a device for a stated indication, not proof of benefit outside that indication. So two things are worth reading: which age is stated, and what data the decision rests on.

Luminopia One. De Novo authorisation DEN210005, 20 October 2021, indicated for children aged 4–7. The basis was a randomised controlled trial: 105 children aged 4–7 at 21 sites, 12 weeks; amblyopic-eye acuity improved by 1.8 lines in the treatment group and 0.8 lines in the group that continued glasses alone (PMID 34534556).

The age extension. Submission K243819, cleared 9 April 2025, extended the indication to "aged 4 to <13". The support was not a new randomised trial but a prospectively designed real-world registry: 334 patients, of whom 290 were aged 4 to under 13 with amblyopia associated with anisometropia and/or strabismus. Mean acuity gain was 1.1 lines (95% CI 0.92–1.3) across the whole group, 1.2 lines (95% CI 1.0–1.4; n = 186) in ages 4–7 and 0.95 lines (95% CI 0.66–1.3; n = 104) in ages 8–12. The registry had no control group. Before starting, participants had spent on average 2.8 years in refractive correction and 1.8 years on patching and/or atropine, and the course itself lasted about 8.2 months on average (FDA documents K243819, DEN210005).

CureSight. Clearance K221375, 29 September 2022, indicated for children aged 4 to under 9. The basis was a multicentre randomised trial with masked assessment: 103 children aged 4 to under 9, 16 weeks, a pre-specified non-inferiority margin of one line, which the result did not exceed (PMID 36306974).

All of this concerns one specific regulatory category: "digital therapy device for amblyopia" (21 CFR 886.5500). As of September 2026 it contains two products, and both indications stop short of thirteen — Luminopia at "under 13" and CureSight at "under 9". So an eight-year-old is covered by both, the full 8–12 range by Luminopia alone, and the data behind that extension are an uncontrolled observation rather than a comparison against an alternative. For thirteen and above the category holds no indication.

That does not mean nothing is on the market for a teenager. Devices cleared under older rules and in other regulatory categories are sold for those ages too: the AA-1 system, for instance, was cleared by the FDA back in August 2001 under the haploscope category — two decades before the digital therapy device category for amblyopia existed at all. Such products cannot be compared with the two above on the word "authorised": an authorisation is read together with its year, its category and the data attached to it.

The drugs that were meant to reopen the window

Levodopa was studied as an add-on to patching precisely in older children. In a placebo-controlled randomised trial, 139 children aged 7–12 with residual amblyopia after patching received levodopa or placebo three times daily alongside two hours of daily patching for 16 weeks. By week 18 acuity had improved by 5.2 letters with levodopa and 3.8 letters with placebo; the difference adjusted for baseline acuity was 1.4 letters (two-sided 95% CI −0.4 to 3.3). The authors concluded there was no clinically or statistically meaningful advantage (PMID 25676904).

This is a case where a negative result is more useful than a positive one: it closes a direction that would otherwise absorb years and money.

Perceptual learning and vision therapy

Here the evidence is thinnest and the marketing loudest. An illustration of the level of data available: a randomised trial in which 52 children aged 4–12 were assigned to three groups — two hours of patching, monocular perceptual learning, and patching combined with vision therapy. Acuity and stereoacuity improved in all three groups; between-group differences in stereoacuity were not statistically significant, while on acuity the authors reported an ordering — patching with vision therapy did best, then perceptual learning, then patching alone (PMID 39396111). But with 17–18 children per group that ordering is an observation awaiting confirmation, not a basis for choosing a protocol.

Some methods deserve a separate note because they are sold far more actively than they have been studied. Devices using rotating gratings (CAM stimulators) were tested in a double-masked trial as far back as 1981: 19 people viewed rotating gratings for 15 minutes a week over ten weeks, and another 19 viewed an identical rotating but blank disc. No significant change in acuity was found within or between groups (PMID 7032305). Note the comparison: the gratings were tested against a dummy rather than against patching, so the trial answers whether the device itself adds anything.

Drugs acting on serotonin transmission have been studied as a possible way to support plasticity in adults. A crossover trial of citalopram failed to reach its planned sample size, enrolled seven adults and found no significant difference from placebo (PMID 31281343). A placebo-controlled trial of fluoxetine enrolled 55 adults (aged 18–54, mean 27.2); alongside patching, acuity improved by 0.20 logMAR on the drug and 0.08 logMAR on placebo, a significant between-group difference (p = 0.04, PMID 39304030). That is a reason to keep studying, not to prescribe.

A general rule for reading this market: absence of evidence is not evidence of absence, but it is also not a reason to pay for a method as though it were established. The question that works best here is what exactly the method was compared against, and how many people took part.

What this changes in the conversation with your doctor

None of the above is a prescription: choosing treatment for amblyopia is the ophthalmologist's work, and the decision follows the individual picture rather than a review. But the questions worth asking follow directly from these studies:

1. How long has the child worn full correction, and what acuity was measured after that period — that is the baseline against which any next step is judged.

2. Was acuity measured separately for each eye and separately at near, not only at distance? A near deficit does not show up in a distance number.

3. Was refraction measured under cycloplegia? Measurements with and without drops are not comparable, and a series of non-comparable points shows no trend.

4. Was stereoacuity tested? It is a separate property and does not follow automatically from acuity.

5. If a digital method is offered: at what ages was it studied, what was it compared against, and on what data was it authorised? The age range in an authorisation is a checkable fact.

6. How is treatment meant to end? The data on stopping differ between younger and older children, and the question is worth asking before starting rather than after.

This is also where Lonevi's usefulness begins and ends: the record shows which numbers have been measured and how they changed, including those brought back from eye appointments. We do not prescribe treatment and do not judge it; we help keep a series of measurements from falling apart, because the trend lives in the series. There is a separate piece on sharing your record with a doctor, and another on what the summary does not do.

This material is informational and does not replace a consultation with a physician. Diagnosis and treatment of amblyopia are determined by an ophthalmologist.