Diversity in optical coherence tomography normative databases: moving beyond race

Abstract

Normative databases of optical coherence tomography (OCT) metrics, such as retinal nerve fiber layer (RNFL) and macular thickness, are critical to clinical use of OCT imaging. In order to accurately represent the range of normal variation in patient populations, these normative databases must themselves be adequately diverse. Thus far, diversity in OCT normative databases has largely been defined as racial diversity. However, this has largely been based on self-reported “race,” which is inconsistent and generally not scientifically rigorous as a form of categorization. Moreover, there is a great deal of variation even within any single racial group, suggesting that other drivers of variation, such as geography or socioeconomic status, may be more important metrics for diversity. Finally, race itself is a proxy for the biological variation that must be represented in such samples, and as such racial diversity does not itself inherently equate to adequate biologic diversity. As clinical use of OCT continues to grow, including to international settings, it is increasingly important that normative databases built into OCT systems accurately represent the populations to which they are applied. Race is not an ideal sole or even primary means of assessing sample diversity in this context. In future normative OCT database construction, other forms of diversity should be considered.