This week in Ophthalmology

This AudioScholar briefing is generated by artificial intelligence from PubMed abstracts for healthcare professionals and is not medical advice.

This week: steroids tested against placebo for ethambutol optic neuropathy, four-year faricimab data in diabetic macular edema, and a sharper genetic risk score for glaucoma.

Welcome to This Week in Ophthalmology. We're covering 10 papers this week: 3 in depth, 4 in brief, 3 quick signals.

Our first paper, in Ophthalmology, is a double-blind randomized placebo-controlled trial from Takkar and colleagues at a tertiary centre in India [1]. It enrolled 100 patients who developed visual loss while taking ethambutol and met predefined criteria for ethambutol-induced optic neuropathy, a condition with no established treatment beyond stopping the drug. Patients received intravenous methylprednisolone, 1 gram a day for 5 days, or placebo, on top of standard care. The primary endpoint was an improvement of two or more Snellen lines at six months. That threshold was reached by 72 percent of patients on methylprednisolone and 46 percent on placebo, a statistically significant difference that held after adjustment and across subgroups. Visual field mean deviation and visual field index were also better in the steroid arm, and no serious treatment-related adverse events occurred. The main limitation sits in the data itself: mean final visual acuity did not differ significantly between the groups, so the responder benefit did not carry through to average acuity. It is also a single-centre trial of modest size, and the authors call the findings hypothesis-generating. Even so, it is the first randomized trial of corticosteroids in this condition, and the first controlled signal for any therapy.

Also in Ophthalmology, Sheth and colleagues report four-year results from RHONE-X, the open-label, nonrandomized two-year extension of the YOSEMITE and RHINE phase 3 trials of faricimab in diabetic macular edema [2]. In all, 1474 patients entered the extension, and every patient moved to faricimab treat-and-extend at intervals up to 16 weeks, whether previously on faricimab or aflibercept. The primary endpoints were safety. Discontinuation due to adverse events occurred in 1.5 percent of patients and intraocular inflammation in 1.3 percent. Visual gains from the parent-trial baseline held at roughly ten letters across all three groups, and more than 90 percent of patients had no macular edema by the central subfield thickness criterion at study end. Patients needed a median of seven or eight injections over two years, and roughly 80 percent of patients reached intervals of 12 weeks or longer. The limitations follow from the design: efficacy endpoints were exploratory, there was no comparator arm, and close to one in five participants did not complete the extension, which tilts results toward patients who did well. What it adds is long-term durability and safety data at a large scale.

The third paper, again in Ophthalmology, comes from MacGregor and colleagues, who built a new polygenic risk score for open-angle glaucoma from genome-wide data on more than six million people across several ancestry groups, folding in related traits such as intraocular pressure and optic disc measures [3]. In Australians of European ancestry, people in the top tenth of the score had about ten times the odds of glaucoma compared with everyone else, more than double the performance of the group's earlier score. Results were similar in United States European ancestry data and lower in UK Biobank, at about five-fold. People in the highest decile were diagnosed roughly 25 years earlier than those in the lowest. The score also predicted treatment escalation, glaucoma surgery and risk in first-degree relatives. The key limitation is equity: performance fell in South Asian, East Asian and African ancestry groups, to between about two- and four-fold for the top tenth, although absolute risk still rose with the score in every group. This is a prediction study, not a trial of screening, so whether score-guided care improves outcomes is untested. The authors report the score is available through an accredited genotyping workflow.

Now, four in brief.

Staying with neuro-ophthalmology alongside the ethambutol trial, the ONSET study in Ophthalmology, from Mori and colleagues, is a retrospective cohort of 601 consecutive patients with MRI-confirmed first-episode optic neuritis at 11 Japanese referral centres, followed for a median of two years [4]. Just over a quarter of patients had a further central nervous system inflammatory event, most often recurrent optic neuritis, followed by encephalitis and myelitis. By last follow-up, only about a third of patients remained classified as single isolated optic neuritis, and the share diagnosed with multiple sclerosis rose from about 3 percent to about 13 percent. Most later events occurred outside maintenance therapy, often after it was stopped. The design is retrospective with short median follow-up, and a Japanese referral cohort may not reflect other populations.

Returning to diabetic macular edema, Farrell and colleagues in Retina reviewed 118 eyes treated with the 0.19 milligram fluocinolone acetonide implant, followed for about 33 months on average [5]. Ocular hypertension above 25 millimetres of mercury developed in about 9 percent of eyes, and only one eye in the whole cohort needed incisional glaucoma surgery, with the rest managed by drops or observation. Onset was variable, with over a quarter of cases appearing after two years, and higher baseline pressure raised the odds of a rise. This is a single retrospective series in which most eyes had prior corticosteroid injections, so rates may differ in eyes without that exposure.

Back to glaucoma, and to a question of how far its evidence base can be trusted, a JAMA Ophthalmology audit by Endo and colleagues examined three public glaucoma datasets from one laboratory that are widely used as independent validation cohorts for artificial intelligence [6]. Linking records and then comparing images directly, they confirmed that at least 31 percent of eyes in Harvard-GF reappear as identical OCT volumes in Harvard-FairVision under different identifiers. No release disclosed the overlap, and none of the citing studies they could read reported it. A classifier tested across the overlapping sets scored an area under the curve of 0.90, against 0.78 on unseen eyes, with inflation in every racial subgroup. The evidence is firm for the two datasets with OCT volumes and weaker for the third.

Finally in this group, Orprayoon and colleagues in Retina report a retrospective series of 256 eyes that underwent vitrectomy for symptomatic floaters [7]. Rhegmatogenous retinal detachment developed in 14 eyes, about 5.5 percent, at a median of 19 days after surgery, and most detachments were macula-on. All 14 eyes were reattached without long-term silicone oil. Three of those eyes lost three or more lines, about one percent of the whole cohort. As a single series without a comparison group, it gives a local estimate of risk rather than a firm population rate.

Three quick signals.

In Survey of Ophthalmology, Sahu and Balakrishnan offer a narrative review of pegcetacoplan for geographic atrophy, arguing that slower lesion growth has not translated into visual acuity gains, while neovascular conversion and rare retinal vasculitis shift the risk balance [8]. They propose biomarker-based patient selection. It is an argument built on existing data, with no new results.

In Progress in Retinal and Eye Research, Ramtohul and colleagues reassessed 207 published cases of acute zonal occult outer retinopathy and assigned about nine in ten cases a more specific contemporary diagnosis, most often multiple evanescent white dot syndrome [9]. They propose a pattern-based approach led by OCT and autofluorescence. The authors state this does not show the original diagnoses were wrong.

And in the American Journal of Ophthalmology, Lin and colleagues gave high-dose intravitreal topotecan, 100 micrograms every four weeks, to 15 patients with refractory or recurrent retinoblastoma [10]. The globe was saved in 14 of 15 eyes, seeds were controlled in every eye, and electroretinography did not change. A small uncontrolled series cannot settle the authors' first-line claim.

Stepping back across this week's papers, the firmest evidence comes from a single randomized trial at one centre, and even there the responder endpoint and average acuity told different stories. Much of the remaining work is retrospective, uncontrolled, or extended without a comparator. The dataset audit adds a further layer, showing that validation cohorts can be less independent than reported, so apparent performance in several areas may rest on narrower foundations than the published figures suggest.

If you only have time for one paper this week, make it the ethambutol optic neuropathy trial [1]. It is the first randomized evidence for any treatment in a drug toxicity long managed by withdrawal alone, and it reopens the question of whether early steroids change visual recovery.

That's your roundup for This Week in Ophthalmology. The full transcript and references are available on the episode page. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And a quick rating on Apple Podcasts or Spotify helps other physicians discover the show.

This is an automated summary generated by artificial intelligence, which can make mistakes. Use it to choose which papers to read in full.