No 'smoking gun' OCT biomarker found for DME outcomes: Insights from Protocol T

The study examined whether granular OCT features could predict visual and anatomical outcomes following anti-VEGF therapy. Ophthalmology Times spoke with Dr. Ip about the key takeaways and what they mean for clinical practice.

Can you describe the dataset and study design you used for this analysis?

Our research group leveraged the DRCR Retina Network Protocol T dataset, which was originally a landmark comparative effectiveness clinical trial evaluating the relative efficacy and safety of 3 anti-VEGF therapies. However, that was not the focus of our work. We went back and re-analyzed the OCT data from the 5-year follow-up period, performing mass re-segmentation at the Doheny Image Reading Center.

We examined a range of OCT biomarkers—including hard exudates in the central subfield and grid, disorganization of the retinal inner layers (DRIL), disorganization of the retinal outer layers (DROL), choroidal thickness, and retinal volume. We then used logistic regression analysis to relate these biomarkers to both functional and anatomical outcomes.

Our functional outcome was defined as an improvement in visual acuity of 10 or more letters, and our anatomical outcome was a reduction in OCT central subfield thickness (CST) of 20% or more. A lot of prior work in this space has come from small groups using retrospective studies with heterogeneous datasets, so we felt this well-characterized dataset offered a meaningful opportunity to address some of the existing controversy in the literature.

What did the analysis reveal about predictors of functional outcomes?

When we looked at functional outcomes—specifically, the likelihood of gaining 10 or more ETDRS letters—baseline visual acuity emerged as the dominant driver. Across all five annual follow-up visits, baseline visual acuity had an odds ratio of less than 1, with a strong P value of less than 0.05. The odds ratio effect was consistently in the range of 0.2 to 0.3.

In practical terms, this means that the lower a patient's baseline visual acuity, the more likely they are to achieve a clinically meaningful gain in vision. This is perhaps intuitive—patients starting with worse vision have more room to improve—but it was consistently borne out across all five years of follow-up in our multivariate models.

What about anatomical outcomes—did OCT CST prove to be a predictor there as well?

Yes, and this was one of the more compelling findings. For anatomical outcomes—defined as a 20% or greater reduction in OCT central subfield thickness—baseline CST was strongly and consistently associated with achieving that endpoint. Across all five annual visits, the odds ratio for baseline CST was greater than 1, with a statistically significant P value at or below 0.05. The odds ratios ranged from approximately 2 to 6, indicating a very robust association.

The summary finding is straightforward: the thicker the retina at baseline, the more likely a patient is to achieve a meaningful reduction in CST over time. Again, there is some intuitive logic here, since a very thickened retina has more potential for measurable anatomical improvement.

Were any of the more granular OCT biomarkers—such as DRIL, DROL, or hard exudates—predictive of outcomes?

This is really the crux of what we were hoping to find, and the honest answer is no. None of the more granular OCT biomarkers we evaluated were able to reliably predict visual or anatomical outcomes at future follow-up visits. DRIL, DROL, hard exudates in the central subfield or grid, choroidal thickness, retinal volume—none of these rose to the level of an independent, consistent predictor after multivariate analysis.

This is consistent with the broader literature, which has been quite mixed. Different studies find different things, often because datasets are small and heterogeneous and biomarker definitions are inconsistent. We undertook what was essentially a mass analysis of a very high-quality dataset, and to the best of our knowledge, there is no smoking gun OCT biomarker beyond central subfield thickness for anatomy and baseline visual acuity for function.

What are the clinical implications of these findings for practicing ophthalmologists?

The immediate practical takeaway is this: when you see a patient with DME at baseline, you cannot look at the OCT and say, 'Because there is DRIL or DROL, or because there are hard exudates in the central subfield, this patient will or will not respond to anti-VEGF therapy.' On a population level, and I think we can translate this to the individual patient level as well, those more granular features simply do not carry the predictive weight many of us had hoped they might.

What you can do is counsel patients based on two key parameters: baseline visual acuity and baseline OCT central subfield thickness. A patient who presents with a very thick retina and significantly reduced visual acuity is actually a candidate for a more optimistic prognosis—you can tell them there is a good chance they will see meaningful improvement in both vision and retinal thickness with anti-VEGF treatment. That is genuinely useful information to share during the initial encounter.

What do these findings mean for future research directions in this space?

The search certainly goes on. I think we need to continue looking for specific OCT biomarkers that might carry predictive value, but we need to do so more rigorously. One of the critical steps forward is standardizing the definitions of OCT biomarkers across studies. DRIL, for example, is defined differently in different papers, and that inconsistency makes it nearly impossible to draw firm conclusions from the aggregate literature.

Equally important is the methodological approach: biomarkers need to be assessed independently using proper multivariate logistic regression analysis, rather than examined in isolation or with unadjusted statistics. I am optimistic that with better harmonized definitions and more rigorous analytical frameworks, future research may yet identify biomarkers with genuine predictive utility. But at this stage, the data do not support the existence of a single OCT feature that clinicians can reliably use to predict outcomes beyond what baseline vision and CST already tell us.

This interview has been edited for length and clarity.