Alternative therapies versus patching for childhood amblyopia: a modality-stratified meta-analysis

Patching remains the standard treatment for childhood amblyopia but is constrained by adherence and psychosocial burden, motivating pharmacologic, optical, and binocular/digital alternatives that differ fundamentally in mechanism. Rather than asking whether alternatives are, on average, equivalent to patching - a question prior syntheses have answered by pooling mechanistically dissimilar interventions - we asked a more clinically useful question: does efficacy depend on the type and generation of therapy?

A PROSPERO-registered systematic review and meta-analysis (PRISMA 2020) searched PubMed/MEDLINE, EMBASE, and Cochrane CENTRAL (2000-23 July 2026) for randomized controlled trials (RCTs) comparing an alternative therapy with patching as the active comparator in children predominantly aged ≤9 years. The primary outcome was change in amblyopic-eye best-corrected visual acuity (BCVA, logMAR). Mean differences (MD = Δintervention - Δpatching; negative favors the experimental therapy) were pooled with random-effects models (REML, Hartung-Knapp). A modality subgroup analysis was prespecified; analyses by device generation and by treatment duration were post hoc exploratory analyses prompted by clinical heterogeneity.

Eleven RCTs (1582 children randomized; 1467 contributing to the primary-outcome analysis) were included (nine of the eleven tested binocular/digital therapies; atropine and Bangerter filters were each represented by a single trial). Overall, alternative therapies did not differ from patching (MD + 0.005 logMAR; 95% CI - 0.022 to +0.032; p = .68), but heterogeneity was substantial (I² = 62%). This pooled null concealed a modality-dependent pattern: atropine (MD + 0.034), Bangerter filters (+0.040) and first-generation binocular games favored patching, whereas contemporary dichoptic/eye-tracking therapies did not (between-subgroup p = .046). Within the binocular/digital subgroup, the contrast between first-generation games (+0.041) and contemporary dichoptic systems (-0.016) was the most robust signal in the dataset (between-subgroup p = .002).

In children predominantly aged ≤9 years, these exploratory analyses suggest that efficacy may depend on which binocular therapy is used rather than on binocular therapy per se: contemporary dichoptic/eye-tracking systems showed no statistically detectable difference from patching, whereas first-generation games, penalization and Bangerter filters did not exceed it. The pooled null should not be read as uniform equivalence, and the absence of a detectable difference does not establish non-inferiority to patching. These generation-stratified findings identify potentially important heterogeneity within binocular therapies and generate a hypothesis requiring confirmation in adequately powered, modality-specific trials.