This Week in Ophthalmology — May 14, 2026

Welcome to This Week in Ophthalmology. This week we're covering 10 notable papers spanning major advances in glaucoma surgery, the expanding role of artificial intelligence in diagnostics, and new guidance on several pressing clinical challenges. Let's dive in.

Our first theme this week is glaucoma surgery, where two key papers provide a view of both current challenges and a potential new solution.

A large, retrospective study from the IRIS Registry, or Intelligent Research in Sight, published in the *American Journal of Ophthalmology*, gives a sobering look at real-world outcomes of glaucoma surgery in young patients—those aged 18 to 40—with secondary glaucoma [1].

The Study Investigators analyzed nearly 2,500 eyes that underwent either minimally invasive glaucoma surgery, known as MIGS, trabeculectomy, or tube shunt surgery between 2013 and 2024.

Results The two-year failure rates were strikingly high across the board: 50.6% for trabeculectomy, 52.1% for tube shunts, and similar rates for MIGS procedures like goniotomy and canaloplasty, which were around 53% and 49% respectively. This means that about half of all procedures failed within two years, based on a composite definition that included insufficient intraocular pressure reduction, hypotony, or the need for additional glaucoma procedures. Outcomes also varied significantly depending on the cause of the secondary glaucoma. For instance, in uveitic glaucoma, canaloplasty had the lowest failure rate at about 36%, while in steroid-induced glaucoma, trabeculectomy had the lowest failure rate at about 38%, though it came with a 10% risk of hypotony.

Conclusions The key clinical message is that for young patients with secondary glaucoma, surgical success is far from guaranteed, and the choice of procedure may need to be tailored to the specific glaucoma subtype. While the authors caution that baseline disease severity may influence these comparisons, the data underscore the complexity of managing this patient population.

While those real-world failure rates are a challenge, a study in *Ophthalmology* offers a specific strategy to improve outcomes for trabeculectomy [4].

The Study This was the three-year, long-term follow-up from the 'Bevacizumab in Trabeculectomy Study,' a single-center, double-blinded randomized controlled trial. Patients undergoing trabeculectomy with mitomycin-C, or MMC, were randomized to receive a single intraoperative intravitreal injection of bevacizumab or a placebo.

Results At 36 months, the results were significant. Trabeculectomy survival was substantially higher in the group that received bevacizumab. The hazard ratio for failure was dramatically lower in the bevacizumab group for both complete success, which was defined as achieving target pressure without medication, and qualified success, where medications were allowed. Patients in the bevacizumab group were also significantly less likely to require topical glaucoma medications at the three-year mark.

Conclusions The conclusion is direct and actionable: a single intravitreal dose of bevacizumab given during trabeculectomy with MMC resulted in improved surgical success at three years. This provides an evidence-based adjunct that could boost success rates for this cornerstone glaucoma procedure.

Next, we turn to a major theme this week: the application of artificial intelligence and advanced imaging for risk stratification and diagnosis across the specialty.

Starting with diabetic retinopathy, a study in the *British Journal of Ophthalmology* used ultra-widefield, or UWF, imaging to identify new predictors of progression [10].

The Study This was a two-year prospective study of over 500 eyes in an Asian cohort with diabetes. The investigators used an automated algorithm to quantify retinal lesions on UWF photos.

Results The key finding was that increased peripheral retinal hemorrhage density was an independent predictor for progression of diabetic retinopathy, or DR, nearly tripling the odds. This was even after adjusting for other risk factors. The presence of predominantly peripheral lesions, or PPLs, was also a strong predictor, increasing the risk fourfold. The overall two-year progression rate was just under 10%.

Conclusions This study reinforces the importance of the retinal periphery in DR and suggests that automated, quantitative biomarkers like hemorrhage density on UWF imaging could be powerful tools for risk stratification, helping to identify high-risk patients who may need closer monitoring or earlier intervention.

Similar machine learning techniques are also being applied to high myopia. Another paper in the *British Journal of Ophthalmology* explored using radiomic features from choroidal optical coherence tomography, or OCT, images to identify pathological myopia [2]. Investigators found that four specific choroidal intensity and shape features were strongly correlated with axial length and visual acuity. A classification model built using these AI-derived features was significantly better at identifying pathological myopia than a model using standard clinical parameters like axial length and spherical equivalent, with an area under the curve, or AUC, of 0.97 versus 0.94 in the internal test set. These quantitative biomarkers could serve as a more reliable way to monitor high myopia progression.

Broadening the scope, *Survey of Ophthalmology* published a comprehensive narrative review on the role of artificial intelligence, or AI, in managing retinopathy of prematurity, or ROP [3]. The review summarizes how AI applied to retinal imaging, clinical text data, and even smartphone images can transform ROP diagnosis, risk prediction, and treatment suggestions. While highlighting the immense potential to improve screening, especially in regions with limited resources, the authors also point to persistent challenges in data heterogeneity and model generalizability that need to be overcome for real-world clinical deployment.

And finally in this theme, a paper in *JAMA Ophthalmology* titled "Generative Artificial Intelligence-Driven Voice Assistance for Patient Education in Ophthalmology" [6] points toward future applications. While the article is a commentary without study data, its inclusion this week highlights the growing interest in using AI not just for diagnostics, but also for enhancing patient communication and education.

Our final section covers new guidance on clinical conundrums and a look at future directions in diagnostics and clinical trials.

First, a highly anticipated joint consensus statement from the North American Neuro-Ophthalmology Society and the American Academy of Ophthalmology, published in *Ophthalmology*, addresses the potential link between glucagon-like peptide-1 receptor agonists, or GLP-1 RAs, and non-arteritic anterior ischemic optic neuropathy, or NAION [8].

Guidance The statement concludes that while some retrospective observational studies suggest a small possible increased risk of NAION in patients taking these drugs, other studies report no correlation. The overall magnitude of risk, if any, remains low. The expert panel emphasizes that current evidence is limited and subject to bias.

Clinical Bottom Line The recommendation is not to routinely discontinue GLP-1 RAs due to ocular concerns. Instead, the focus should be on shared decision-making between each patient and their care team, weighing the well-established systemic benefits of these medications against a low and uncertain ocular risk.

Next, for clinicians managing retinal vein occlusion, or RVO, a common question is when treatment can be safely stopped. A study in the *British Journal of Ophthalmology* provides some answers from the Fight Retinal Blindness! registry [9].

The Study Researchers analyzed eyes with RVO that had suspended anti-VEGF therapy for at least six months. They looked for predictors that determined which eyes could successfully stay off treatment for a full 24 months.

Results Success was possible, but only in about half the cases. 51% of eyes that attempted a treatment holiday remained treatment-free at two years. The successful attempts were characterized by starting the suspension earlier, having achieved greater vision improvement from baseline, and having received more intensive treatment in the first six months. Eyes that successfully stopped treatment even continued to show slight improvement in vision and anatomy during the first six months off therapy, while unsuccessful attempts were already stable or deteriorating.

Conclusions The clinical implication is that the best candidates for a treatment holiday are those who respond well and early to intensive initial anti-VEGF therapy. These findings can help guide conversations with patients about the likelihood of successfully stopping injections.

Shifting to a diagnostic puzzle, a case report in *JAMA Ophthalmology* highlights the power of new genetic sequencing techniques [7]. Researchers investigated a 16-year-old male with classic aniridia whose condition remained undiagnosed despite extensive testing, including sequencing of the PAX6 gene and even short-read whole-genome sequencing. Using advanced methods called optical genome mapping and long-read whole-genome sequencing, they identified a tiny, balanced intrachromosomal translocation. A 55-kilobase segment containing the PAX6 coding sequence was deleted and re-inserted elsewhere on the same chromosome. This physically separated the gene from its essential regulatory region, explaining the aniridia. This case demonstrates that for the 5 to 10 percent of aniridia patients who lack a molecular diagnosis, these advanced sequencing methods may be necessary to find tiny structural variants missed by current standard-of-care tests.

Finally, and looking toward the future of clinical research, a paper in *JAMA Ophthalmology* proposes a way to make clinical trials for geographic atrophy, or GA, much more efficient [5]. The study suggests that incorporating a 'run-in' phase—where a patient's natural GA growth rate is measured before they are randomized to a treatment—can drastically reduce the required sample size. For example, a simulated two-year randomized trial needing 156 patients for 90% power could achieve the same power with just 40 patients if it included a six-month run-in phase. An even more efficient 'single-arm' design, where each patient serves as their own control, could require as few as 14 patients. This innovative trial design could accelerate the development and approval of new treatments for GA by making trials smaller, faster, and less expensive.

If you only have time for one paper this week, make it the long-term follow-up of the Bevacizumab in Trabeculectomy Study from *Ophthalmology* [4]. It provides strong, three-year, randomized evidence that a single intraoperative dose of bevacizumab improves trabeculectomy survival, a simple intervention that could enhance outcomes for our glaucoma surgery patients.

Here are the key takeaways from this week in Ophthalmology: First: Consider a single intraoperative dose of bevacizumab to improve three-year success rates for trabeculectomy, which may help counter the high baseline failure rates seen in complex glaucoma surgery.

Second: When assessing patients with diabetic retinopathy or high myopia, the retinal periphery is key. Automated analysis of ultra-widefield imaging shows that biomarkers like peripheral hemorrhage density are powerful, independent predictors of disease progression.

Third: For patients on GLP-1 agonists like semaglutide who are concerned about NAION, the current consensus from NANOS and the Academy is that the risk is low and not definitively causal. This supports shared decision-making rather than routine discontinuation of these important systemic medications.

Fourth: When considering a treatment holiday for RVO patients, the best candidates are those who responded well and early to intensive initial anti-VEGF therapy. About half of such patients may be able to remain off treatment for up to two years.

Fifth: The future of ophthalmology will be increasingly shaped by AI-driven diagnostics and smarter clinical trial designs. New methods are already showing promise for quantifying risk in retinal disease and accelerating research for conditions like geographic atrophy.

That's your roundup for This Week in Ophthalmology. The full transcript and references are available on the episode page in your AudioScholar library. This is an AI-curated summary — for clinical decisions, always consult primary sources and current guidelines. See you next week.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.