Highlight
- Hyperreflective abnormal areas (HAAs) in the middle retinal layers measured by en face OCT independently identify eyes with referable diabetic retinopathy (DR) beyond established clinical factors.
- In a cohort of 275 diabetic eyes without center-involving diabetic macular edema (DME), HAAs improved diagnostic discrimination with a significantly higher AUC for referable DR.
- The association between HAAs and vision-threatening DR was directionally positive but did not significantly enhance predictive accuracy beyond clinical variables.
- HAAs may reflect lesion-agnostic microvascular damage and represent a novel, non-invasive imaging biomarker to stratify diabetic eyes at high risk needing timely referral.
Study Background
Diabetic retinopathy (DR) is a leading cause of vision impairment globally, escalating with the duration and management of diabetes. Early identification of referable and vision-threatening DR is critical for timely intervention to prevent progression and vision loss. Traditionally, clinical assessments incorporating systemic factors such as glycemic control, hypertension, and duration of diabetes guide referral decisions, supplemented by fundus photography or OCT imaging to detect macular edema or proliferative changes. However, subtle microvascular changes within the retinal architecture, particularly in middle retinal layers, often precede clinical manifestations detectable by conventional methods.
Recent advancements in optical coherence tomography (OCT) technology, specifically en face imaging, enable detailed visualization and quantification of retinal microstructures. This study investigates the diagnostic utility of hyperreflective abnormal areas (HAAs) seen on en face OCT within the middle retinal layers, an area encompassing 10 μm above the inner plexiform layer to 10 μm below the outer plexiform layer, hypothesized to capture early microvascular injury signatures in diabetic eyes without evident center-involving DME.
Study Design
This accuracy analysis included 275 eyes from 185 diabetic patients spanning a broad range of DR severity but excluding those with center-involving diabetic macular edema (DME). Manual quantification of HAAs was performed on averaged en face OCT images from the middle retinal slab. HAAs were expressed as a percentage of the analyzable retinal area.
The study employed generalized estimating equation logistic regression models to distinguish referable and vision-threatening DR status. A base clinical model factored in age, sex, diabetes duration, hemoglobin A1c, and hypertension. A full model incorporated HAAs alongside these parameters. Model performances were compared using the area under the receiver operating characteristic curve (AUC), internal cross-validation, and calibration analysis.
Key Findings
For referable DR discrimination, the inclusion of HAAs significantly improved the model’s predictive performance. The full model achieved an AUC of 0.969 compared to 0.865 in the base clinical model alone, reflecting a statistically significant improvement (ΔAUC 0.104, p=0.002). HAAs showed a strong independent association with referable DR, with an odds ratio (OR) of 3.42 per 0.1% increment in HAA area (95% CI, 2.37–7.97). These results were robust on internal cross-validation, reinforcing the reliability of HAAs as an imaging biomarker for referable DR.
For vision-threatening DR, the full model also demonstrated higher discrimination (AUC 0.865 vs. 0.778 for base model), though this improvement did not reach statistical significance (p=0.126). The odds ratio per 0.1% increase of HAA was 1.23, suggesting a positive but weaker association with more advanced DR stages.
These findings highlight HAAs as a potential structural biomarker aligning with microvascular injury distinct from lesion-specific markers visible on traditional fundus imaging or OCT central retinal thickness measurements.
Expert Commentary
The detection of hyperreflective abnormal areas in middle retinal layers offers a novel perspective on DR risk stratification. Unlike conventional markers like hemorrhages, microaneurysms, or DME detectable by fundus photography or thickness maps, HAAs may represent diffuse microvascular alterations or inflammatory changes that elude standard imaging but portend disease progression.
The methodology of manual HAA quantification on en face OCT is labor-intensive but provides granularity and spatial mapping that may guide future automation efforts. Future work should focus on validation in diverse populations, reproducibility across imaging platforms, and integration with automated artificial intelligence algorithms for scalable screening applications.
Limitations include the cross-sectional design limiting causal inference, exclusion of eyes with center-involving DME which may alter biomarker prevalence, and a moderate sample size. Moreover, the nonsignificant improvement for vision-threatening DR suggests HAAs may be more sensitive for earlier risk detection than advanced disease.
Conclusion
This study demonstrates that hyperreflective abnormal areas in middle retinal layers on en face OCT significantly improve the identification of eyes with referable diabetic retinopathy beyond established clinical risk factors. The biomarker shows promise for enhanced diagnostic discrimination and could facilitate earlier referral decisions to prevent vision-threatening complications.
While the association with vision-threatening DR was less conclusive, HAAs capture lesion-independent microvascular structural changes, positioning them as a complementary biomarker for risk stratification in diabetic eye disease. Continued research is warranted to refine quantification methods, validate findings across broader cohorts, and explore clinical integration for screening and monitoring diabetic retinopathy.