Beyond the Traffic Light: Navigating “Red-Green Disease” in OCT Glaucoma Diagnostics
Optical coherence tomography (OCT) has become a vital component of glaucoma care, with its color-coded peripapillary retinal nerve fiber layer (RNFL) maps offering high intuitive appeal to clinicians. The familiar “traffic light” system—green for within normal limits, yellow for borderline, and red for outside normal limits—is ubiquitous and significantly shapes diagnostic confidence. However, a recent commentary warns against “red-green disease,” or the overreliance on these automated outputs without understanding the limitations of the normative data behind them.
The Statistical Foundation: Smaller Than You Think
The “normal” thresholds we rely on are defined by proprietary reference datasets that are often surprisingly small. For example, the Cirrus HD-OCT normative database was constructed from just 284 eyes, representing a narrow demographic range of White, Chinese, African-descent, and Hispanic individuals. Other major manufacturers, such as Spectralis, Avanti, and Triton, utilize similarly modest sample sizes, often drawn from restricted geographic regions or single ancestry groups. These limited cohorts form the statistical backbone that determines whether a patient’s results appear as green, yellow, or red on the screen.
The Assumption of Uniformity
The current color-coding process assumes that a small reference population can accurately reflect the vast biological diversity of the general patient population. In reality, mean RNFL thickness and optic disc morphology vary significantly based on ancestry, axial length, and disc size. This leads to two primary clinical risks:
- False Positives: Patients with high myopia or unusually large optic discs may show red or yellow sectors simply because their baseline anatomy falls outside the narrow range of the reference population.
- False Negatives: Conversely, eyes with naturally thicker nerve fiber layers or small discs may remain “green” despite the presence of early glaucomatous damage.
Because these color-coding algorithms are proprietary and opaque, it is difficult for clinicians to assess how they perform across different demographic or biometric profiles.
Clinical Strategies for the Modern Practice
To avoid the pitfalls of “red-green disease,” clinicians should interpret RNFL color maps as statistical summaries rather than categorical diagnoses. While these maps carry significant psychological weight, especially for trainees, they must be contextualized within a broader diagnostic framework.
Experts recommend the following approach:
- Correlate with the B-scan: Always check the B-scan to ensure segmentation accuracy.
- Multimodal Correlation: Correlate OCT findings with functional testing (visual fields) and structural confirmation (clinical optic disc appearance).
- Exercise Caution with Specific Cohorts: Be particularly wary when interpreting results for highly myopic patients or those from underrepresented ancestries.
- Look for Symmetry and Trends: In cases where normative data is questionable, inter-eye symmetry and longitudinal assessment provide a much more reliable picture than a single cross-sectional scan.
A Call for Industry Transparency
The future of OCT interpretation should move away from static categorization and toward individualized risk detection. Manufacturers could assist this transition by displaying key metadata directly on OCT reports, such as the sample size, mean age, and racial composition of the normative database used. Long-term solutions include the creation of a global, multi-ethnic normative consortium and the implementation of “algorithmic fairness audits” to publish false-positive and false-negative rates across different demographics.
Ultimately, the RNFL color map is an elegant tool, but its clarity depends entirely on the integrity of the reference populations defining “normal”. By promoting transparency and building inclusive frameworks, we can ensure that the colors we trust truly reflect the diversity of the patients we serve.
Eye; https://doi.org/10.1038/s41433-026-04600-3
