Findings from a recent study published in Journal of VitreoRetinal Diseases reviewed the optical coherence tomography (OCT) biomarkers reported across diabetic macular edema (DME) research and sorted out which ones actually track with visual acuity (VA) and response to intravitreal injections (IVI).
A handful held up across studies. Others are still short on the data needed to use them with confidence.
Give me some background first.
Some context: DME is one of the most common causes of vision impairment in patients with diabetes. One systematic review estimated the global prevalence of OCT-diagnosed DME at 5.47%.
Treatment options are well established. Predicting who will benefit from them is not.
In practice: Two patients with similar-looking edema can respond very differently to the same antivascular endothelial growth factor (anti-VEGF) regimen, and there is no dependable way to sort them at baseline. That gap is what pushed attention toward what the OCT scan shows before the first injection.
Now, talk about the study.
Researchers ran a literature review pulling together the OCT biomarkers reported in DME and mapping each one against two outcomes: final VA and response to intravitreal anti-VEGF and corticosteroid therapy.
The question: Which markers on a baseline scan carry real prognostic weight, and which ones are still short on evidence?
What made it into the review?
Investigators searched Ovid MEDLINE, Embase, the Cochrane Library, and Web of Science from inception through August 2024. The synthesis was narrative rather than quantitative, with effect estimates pulled directly from individual studies where they were reported.
The biomarkers covered spanned six groups:
- Structural: disorganization of the retinal inner layers (DRIL), epiretinal membrane (ERM), ellipsoid zone (EZ) integrity, external limiting membrane (ELM) continuity, foveal eversion, and Müller cell cone integrity
- Thickness-based: central retinal thickness (CRT), central subfield thickness (CST), and foveal and parafoveal retinal thickness
- Fluid-based: subretinal fluid (SRF) and intraretinal cystoid spaces
- Hyperreflective features: hyperreflective dots and hyperreflective walls of foveal cystoid spaces
- Photoreceptor and pigment epithelium: cone outer segment tip, photoreceptor outer segment length, and retinal pigment epithelium (RPE) integrity
- Choroidal: choroidal vascularity index, choroidal thickness, and subfoveal choroidal thickness
Findings?
Across most studies, absence of baseline DRIL or ERM, along with preserved EZ and ELM continuity, was associated with better final best-corrected visual acuity (BCVA) and a stronger response to IVI.
DRIL emerged as the highest-risk predictor of final BCVA in one analysis, with an odds ratio (OR) of 6.71 (P = .001) against comparators including SRF, choroidal vascularity index, and EZ discontinuity. Its absence was linked to a BCVA gain of more than 5 letters.
Worth noting: Extensive DRIL above 500 µm predicted both functional and morphologic resistance to anti-VEGF in patients receiving 3 monthly doses of aflibercept or ranibizumab.
Photoreceptor markers tracked closely with outcomes. Good responders to anti-VEGF, defined as a gain of ≥10 Early Treatment Diabetic Retinopathy Study (ETDRS) letters, had 2.3 and 5.4 times less area of EZ and ELM disruption at baseline than poor responders who gained fewer than 5 letters.
That said: In a separate analysis, EZ disruption carried a slightly greater risk of poor functional response to anti-VEGF (OR, 10.96; P < .001) than DRIL did (OR, 7.05; P = .034). On receiver operating characteristic analysis, EZ was the stronger of the two (area under the curve 0.71 vs 0.64 for ELM).
Tell me more.
CRT had a mixed record: One study found CRT negatively predicted the ability to gain more than 10 letters of final BCVA (P < .001). Another reported only weak correlation with baseline BCVA (r = −0.344), which weakened further over 36 months of follow-up (r = −0.259). Because CRT dropped reliably with both anti-VEGF and corticosteroid IVI, it does not appear to obstruct treatment response.
Greater baseline CRT did predict the need for more than 6 anti-VEGF injections. And a lower CST below 405 µm was associated with nearly 90% of patients gaining at least 1 ETDRS line by 12 months of anti-VEGF therapy (P < .001).
On thickness metrics generally: Only outer retinal thickness predicted BCVA gains. Foveal and parafoveal outer retinal thickness were significant predictors of a greater increase in BCVA, while inner retinal thicknesses were not, pointing to outer retinal damage as the bigger driver in DME.
Hyperreflective dots split the literature: Thought to represent inflammatory cells such as microglia, their baseline presence predicted better VA improvement with anti-VEGF in one study (P = .010) and 2.21 times more visual recovery after 12 months of intravitreal therapy in another. A different study found the opposite: their absence predicted improved outcomes after dexamethasone.
SRF was equally inconsistent: Patients with SRF were more likely to gain more than 5 letters on anti-VEGF (P = .008), and more than 15 letters on corticosteroids (P = .036). Persistent SRF, though, has been tied to RPE and photoreceptor damage. Baseline SRF also drove the need for additional corticosteroid therapy (P = .005), which the authors read as a marker of more chronic, severe disease.
One more to watch: Choroidal vascularity index. Patients gaining more than 5 ETDRS letters had higher baseline values (0.69 vs 0.66), and higher values significantly predicted BCVA gains with anti-VEGF (P = .006), while lower values correlated with better corticosteroid response.
Age and intraocular pressure affect it less than they affect subfoveal choroidal thickness, which makes it the more reliable of the two.
ERM, by contrast, showed little prognostic value:: Several studies found its absence was not correlated with BCVA improvement after treatment, and good and poor anti-VEGF responders showed similar ERM frequencies.
Limitations?
This was a narrative review with no quantitative pooling. Without pooled data, the authors could not formally assess statistical heterogeneity using the I² statistic or compare effect sizes across studies.
Other considerations:
- Included studies varied in design, population characteristics, and how individual biomarkers were defined
- Data on choroidal vascularity index, RPE continuity, CRT, and hyperreflective dots are still thin
- EZ and ELM are difficult to assess in eyes with severe edema, which cuts into their baseline reliability at exactly the point clinicians want to use them
Expert opinion?
No outside expert commentary was included in the study.
Their position: DRIL, hyperreflective dots, and CRT respond more to global cytokine suppression with corticosteroids than to anti-VEGF agents. The authors pointed to one study in which patients switched early to a long-term fluocinolone acetonide implant showed DRIL in 25% of eyes, versus 90% in the late-switch group, along with greater BCVA gains.
Anything else?
Two newer markers are worth tracking: Eyes without foveal eversion improved on either anti-VEGF or corticosteroids while everted eyes stayed relatively static, and everted eyes needed significantly more corticosteroid retreatment. Separately, patients with a good anatomic response to anti-VEGF at 1 month (CST ≤250 µm) were more likely to have an intact Müller cell cone (P < .001), though the association with BCVA did not hold at 12 months.
Also of interest: The authors raised an open question about who or what is reading the scan. Prognostic value may not translate the same way across manually annotated OCT scans, reading center grading, and algorithmic interpretation. Deep learning models already show high concordance with expert manual assessment for fluid segmentation in DME, and the authors called for research into whether large language models can identify and quantify OCT biomarkers as well.
Take home.
Baseline DRIL and EZ/ELM integrity are among the first OCT features to assess when counseling patients with DME. Absence of DRIL and preserved EZ/ELM continuity are associated with better final visual acuity and a stronger response to intravitreal injections.
Eyes with extensive DRIL or hyperreflective dots may respond better to corticosteroids than to anti-VEGF therapy.
As such: CRT is still useful for tracking anatomy, but it should not be carrying prognostic weight on its own.