Optical coherence tomography (OCT) has become indispensable in glaucoma care. Its color-coded reports offer an apparently intuitive interpretation of retinal nerve fiber layer (RNFL) thickness: green indicates measurements within normal limits, while yellow and red flag increasingly abnormal values. But these classifications are only as representative as the normative databases behind them.
A new analysis from the Primary Open-Angle African Ancestry Glaucoma Genetics (POAAGG) study suggests that the standard Cirrus HD-OCT database may classify some eyes of African ancestry as more abnormal than an ancestry-specific reference would.
The investigators compared RNFL color codes generated using the Cirrus normative database with codes recalculated using RNFL data from healthy individuals of African ancestry. The analysis included 787 eyes with primary open-angle glaucoma (POAG), spanning the full spectrum of disease severity according to the Enhanced Glaucoma Staging System.
Across the cohort, the ancestry-specific model produced fewer abnormal classifications and more green results. For average RNFL thickness, the Cirrus database classified 27.7 percent of eyes as red, compared with 24.4 percent under the POAAGG reclassification. The proportion classified as green increased from 51.0 to 56.7 percent.
The direction of movement was particularly revealing. Of the average RNFL measurements designated red by Cirrus, 12.8 percent were reassigned to yellow or green. Among those initially labelled yellow, 23.0 percent became green. Although overall agreement between the systems remained high for average RNFL thickness, at 90 percent, differences were statistically significant.
Reclassification varied by retinal quadrant. The temporal quadrant showed the greatest shift: 80.3 percent of Cirrus red codes moved to yellow, while 71.7 percent of yellow codes became green. In the inferior quadrant – a region of particular importance in early glaucoma – 22.1 percent of red classifications shifted to yellow.
The findings may be most consequential in borderline and mild disease, where OCT results can strongly influence diagnosis, referral, and treatment. Among stage 1 eyes, 19 percent of average red codes and 23.1 percent of yellow codes shifted towards greener classifications. Similar changes were observed in stage 0.5 disease, although the smaller number of abnormal results at this stage limits interpretation.
The study highlights a longstanding limitation of normative imaging databases: healthy anatomical variation is not distributed evenly across ancestry groups. Only 18 percent of the Cirrus RNFL reference cohort was African American, despite the disproportionate burden and severity of POAG among individuals of African ancestry. A database that does not adequately capture baseline variation may contribute to “red disease” – apparently abnormal OCT findings in eyes without corresponding pathology.
The study authors appropriately stop short of claiming that the ancestry-specific model improves diagnosis. Sensitivity and specificity were not evaluated, and the POAAGG reference cohort differed from the Cirrus cohort in age and clinical characteristics. The data were also drawn from the greater Philadelphia area, limiting generalizability.
Nevertheless, the study provides a reminder that OCT colors are probability labels, not diagnoses. More representative normative datasets could therefore help clinicians distinguish genuine glaucomatous damage from normal anatomical diversity, and reduce the risk that a red sector is treated as a red alert.
