Ophthalmology

Loss to follow-up in anti-VEGF–treated DME causes frequent, often irreversible vision loss

In this national registry cohort of 5660 anti-VEGF–treated DME patients, nearly half returned from >6 months LTFU with worse vision after an average 11.3-month gap. Among 792 eyes with worsened vision and at least 12 months of resumed therapy, about two-thirds had persistent vision loss despite retreatment. Older age, care in less developed regions, better pre-LTFU VA, and LTFU duration beyond 12 months increased risk of post-gap vision decline. Persistent loss was more likely with pre-LTFU VA 20/40 or better and central foveal thickness ≥350 μm at return. 1

HALT, DIMS, and CARE lenses all slow myopia; HALT best controls axial length

This retrospective cohort of 899 myopic children compared three myopia-control spectacle designs with single-vision lenses in real-world practice. All myopia-control lenses significantly slowed axial elongation relative to single-vision lenses, which showed 0.193 mm per year progression. HALT lenses had the lowest adjusted axial progression (0.054 mm/year), outperforming DIMS and CARE for axial-length control. All three lens types reduced myopic refractive progression versus single-vision, but inter-lens differences in spherical equivalent change were not significant. 2

Optic disc OCTA vessel density loss tracks visual field progression in POAG

In 345 POAG patients followed for at least 3 years, visual field progressors showed faster optic disc vessel density reduction than non-progressors across all severities. The annual rate of global vessel density loss and a binary cutoff for marked reduction were both independently associated with visual field progression. Odds of progression rose with faster vessel density decline in early, moderate, and advanced disease in multivariable logistic models. Cox models confirmed higher hazards of progression with greater annual vessel density loss in both early and moderate-to-advanced groups. 3

Tele-ROP vascular-severity risk model safely reduces screening examinations

This diagnostic study used i-ROP and SUNDROP data to build a visit-level TR-ROP risk model from gestational age, postmenstrual age, and vascular severity. Adding AI-derived vascular severity significantly improved discrimination over gestational age alone in development and external validation cohorts. A chosen decision threshold achieved 100% sensitivity for predicting TR-ROP within 2 weeks, with moderate specificity in both datasets. Simulated risk-based scheduling reduced examinations by 28% in i-ROP and 39% in SUNDROP without missing any TR-ROP events. 4