The evidence for dichoptic treatment: where we stand today

This piece summarises the evidence for dichoptic approaches in amblyopia for a clinical reader. The picture has sharpened considerably in recent years: dichoptic treatment today is not a conceptual proposal but an approach tested in pivotal randomised trials.

The aim is not to defend a product but to report the state of the field as it is: what has been shown, where the difference appears, and what remains open.

Alongside the loss of acuity, amblyopia involves disordered binocular function: suppression, difficulty with fusion and reduced stereopsis. Monocular treatments target acuity; binocular function is not exercised directly.

The dichoptic proposition is to reduce suppression and target binocular function directly by giving each eye a separate but complementary stimulus. Contrast balancing is the standard adjustment used to let the weaker eye's signal find a place in perception, consistent with the contrast dependence described in the binocular rivalry literature.

In the Luminopia pivotal trial, 105 children aged 4 to 7 were assigned either to dichoptically modified video content plus full-time glasses or to full-time glasses alone. At 12 weeks the gain in the amblyopic eye was 1.8 lines in the treatment group and 0.8 lines in the comparison group; the one-line difference was significant and the trial was stopped early for success at the planned interim analysis (Xiao et al., Ophthalmology, 2022). It should be stated that the comparison arm was glasses alone, not patching.

The CureSight pivotal trial compared directly against patching. In 103 children aged 4 to under 9, an eye-tracking-based dichoptic video treatment was non-inferior to patching: the gain in the amblyopic eye was 0.28 logMAR against 0.23 logMAR (Wygnanski-Jaffe et al., Ophthalmology, 2023).

The same trial recorded a 0.40 log arcsecond improvement in stereoacuity in both groups. Binocular gain here is a measured result rather than a theoretical target.

Response to binocular training has also been reported in adults with amblyopia (Li et al., Current Biology, 2013), which is among the planks of the argument about plasticity beyond the critical period.

The efficacy of patching is not in dispute; what is in dispute is how much of the prescription is delivered. In a randomised trial using occlusion dose monitors, the group prescribed 6 hours a day actually received a mean of 4.2 hours, and the group prescribed 12 hours received 6.2 hours. In both groups the delivered dose was markedly below what was prescribed (Stewart et al., BMJ, 2007).

On the dichoptic side the picture differs. In the CureSight trial, adherence measured 91% in the dichoptic group against 83% for patching, and the difference was statistically significant.

One further detail sharpens the contrast: across 16 weeks the dichoptic group was prescribed 120 hours in total and the patching group 224 hours. The equivalent visual gain was achieved on roughly half the prescribed time.

Adherence here is not a comfort heading but an efficacy variable directly: however sound its biological rationale, treatment that is not carried out does not change the outcome.

The mixed results in the field can be read from the same place. In a multicentre trial of a single binocular iPad game, across 385 children aged 5 to under 13, the gain in the amblyopic eye was 1.05 lines in the game group and 1.35 lines with patching. The primary non-inferiority analysis was indeterminate and a post hoc analysis suggested that this particular game was not as good as patching (Holmes et al., JAMA Ophthalmology, 2016).

The adherence data from the same trial supply the context: of the children with usage logs available, only 22.2% completed more than three quarters of the prescribed treatment, and median adherence was 46%.

Set the two pictures side by side and the conclusion is clear. With the dichoptic principle held constant, median adherence was 46% for a single repetitive game and 91% for rich video content the child chose. The determining variable is not the biological rationale of the method but whether the content holds a child across months.

The design implication follows: dichoptic presentation should be attached not to one game but to a pool of content that does not run out and that the child can choose from. The video-based trials solved this one way; a broad and varied game library aims to solve the same problem on the game side. In both cases the variable being targeted is identical, a dose that can be sustained for months.

Optimal dose and duration: no standard protocol has settled for session length, weekly frequency or total treatment duration.

Patient selection: which type and depth of amblyopia responds best is not yet clear.

Durability: long-term follow-up data are needed on holding the gains and on maintenance protocols.

Outcome measures: standardised reporting of stereopsis and suppression alongside acuity would make trials easier to compare.

Finally, scope: the data above belong to specific products and show the results of those implementations rather than the general validity of a method. No published trial exists for our own program.

Optical correction is the precondition of every protocol and is not in question. Dichoptic treatment positions alongside patching and atropine as the option that targets binocular function directly. Where adherence is the problem, its case is twofold: comparable efficacy and measured high adherence.

Follow-up should track stereopsis and suppression alongside acuity. In digital programs, recording the delivered dose is the only way to separate non-response from insufficient treatment at a plateau.