Protocol T finds no OCT biomarker
- Michael Ip presented an ARVO 2026 reanalysis of DRCR Retina Network Protocol T showing no single OCT biomarker consistently predicted diabetic macular edema outcomes. - The strongest signals were not exotic imaging features but starting points: lower baseline vision tracked letter gains, and higher baseline CST tracked thickness reduction. - That matters because DME clinics keep chasing one-image shortcuts, but this result pushes treatment decisions back toward full clinical context.
Optical coherence tomography — OCT — is the retina clinic’s favorite microscope. It gives doctors a layer-by-layer cross-section of the macula, and for years people have hoped one feature on that scan would tell them who with diabetic macular edema will do well on treatment. That would make life simpler. But the new Protocol T reanalysis presented at ARVO 2026 says the simple version probably is not real: there was no single “smoking gun” OCT biomarker that reliably dominated outcome prediction. ### What was actually analyzed? This was not a tiny retrospective chart review. Michael Ip and colleagues went back to the DRCR Retina Network’s Protocol T dataset — the landmark randomized trial that originally compared aflibercept, bevacizumab, and ranibizumab in center-involved diabetic macular edema — and reworked the OCT imaging with mass re-segmentation at the Doheny Image Reading Center. The point was to test a broad menu of scan features inside a well-characterized trial population instead of relying on small, messy datasets. (ophthalmologytimes.com) ### Which OCT features were in the mix? The team looked at hard exudates in the central subfield and grid, disorganization of the retinal inner layers, disorganization of the retinal outer layers, choroidal thickness, and retinal volume. Then they used logistic regression to ask whether those features predicted either functional improvement or anatomical improvement after anti-VEGF therapy. Basically, they were checking whether one scan detail could act like a shortcut for prognosis. (ophthalmologytimes.com) ### What counted as a good outcome? They used two concrete endpoints. Functional success meant gaining 10 or more ETDRS letters in visual acuity. Anatomical success meant at least a 20% reduction in OCT central subfield thickness, or CST. That matters because DME treatment is always juggling two related but not identical goals — seeing better and looking drier on scan. (ophthalmologytimes.com) ### So what predicted vision gain? Not a fancy biomarker. Baseline visual acuity was the dominant driver across all five annual follow-up visits. The odds ratio stayed below 1 — roughly 0.2 to 0.3 — meaning eyes that started with worse vision were more likely to gain 10 or more letters later. That sounds almost obvious, but it is important because it means the strongest predictor of improvement was room to improve, not some special OCT signature. (ophthalmologytimes.com) ### What predicted anatomical improvement? Baseline CST did. Eyes that started thicker were more likely to hit the anatomical endpoint of a 20% or greater CST reduction. Again, the strongest signal was the starting state, not a single structural biomarker layered on top of it. In plain English — swollen retinas are more likely to show a large percentage drop in swelling. (ophthalmologytimes.com) ### Why is this a bigger deal than it sounds? Because retina specialists have been swimming in biomarker papers for years. Earlier studies and abstracts have pointed to things like intraretinal cyst location, fluid volume, DRIL, or outer retinal changes as possible predictors, but the literature has been mixed and often built on smaller or more heterogeneous cohorts. This Protocol T reanalysis does not say OCT features are useless. It says none of them emerged here as the one decisive triage marker clinicians have been hoping for. (ophthalmologytimes.com) ### Does that mean OCT is less useful now? No — just less magical. OCT still tells doctors whether edema is present, how severe it is, and how anatomy changes over time. But this result argues against making big treatment calls from one favored scan feature alone. DME outcomes still seem to depend on the whole picture — baseline vision, baseline thickness, follow-up response, and the clinical context around the patient. (iovs.arvojournals.org) ### What’s the bottom line? The dream was a single OCT clue that would sort DME patients into likely responders and nonresponders. Turns out Protocol T did not find it. The most dependable predictors were the boring ones — where vision and thickness started — which is useful, but not the shortcut the field was looking for. (ophthalmologytimes.com)
