In this study of a large treated clinical population of subjects with or suspected of glaucoma, we reported significant differences in adjusted rates of change of pRNFL and mGCIPL parameters between various racial-ethnic groups. To our knowledge, this study represents one of the largest analyses of longitudinal SD-OCT data describing both pRNFL and mGCIPL rates among a diverse population of glaucoma patients cared for in routine clinical practice. Of note, Hispanic eyes comprised almost half of the study population, a unique feature in this study. Our study identified faster rates of pRNFL and mGCIPL loss among NHB eyes across most of the spectrum of glaucomatous disease. The lack of any difference in rates among eyes with severe glaucoma may have been due to both smaller sample size and more intense treatment among these subjects. While there was a statistically significant difference in age between groups (Table 1), the estimated rates were adjusted for age as well as other key clinical parameters, such as CCT, IOP, and baseline pRNFL and mGCIPL thickness values. After accounting for these variables, NHB eyes were still noted to have significantly faster rates of pRNFL and mGCIPL loss. Hispanic eyes also tended to have faster rates among certain parameters as well, although most of these comparisons did not reach statistical significance.

Significant differences in mGCIPL rates topographically mirrored areas of significant differences in pRNFL rates in most cases. Among glaucoma suspects, we observed faster thinning in the inferior pRNFL quadrant and inferior/inferotemporal/inferonasal mGCIPL sectors in NHB eyes. Conversely, Hispanics exhibited faster rates in the inferior pRNFL quadrant and inferonasal mGCIPL sector among mild glaucoma eyes, although the former was not statistically significant. Among moderate glaucoma eyes, the superior pRNFL quadrant and supertemporal mGCIPL sector both featured faster rates among NHB eyes. These topographically correlated findings strengthen the likelihood of key differences between racial-ethnic groups.

We observed a strong association between global pRNFL and global mGCIPL rates stratified by glaucoma severity. The correlation coefficient (R2) between pRNFL and mGCIPL was consistent across disease stages, with slightly lower values for earlier disease (Supplementary Fig. 2). The lower correlation coefficient observed among glaucoma suspect eyes could be attributed to early changes occurring in more peripheral regions of the posterior pole, which may not have been captured by the 6x6mm GCIPL cube centred on the fovea. While counterintuitive due to proximity to the RNFL “floor”, the stronger correlation among severe eyes could be explained by our exclusion criteria; we excluded all eyes with a baseline pRNFL thickness <57 µm from the analysis, likely preventing poor correlation due to the “floor effect”. Glaucomatous loss among severe disease is likely to occur closer to the fovea, which is more likely to be captured by the Cirrus GCIPL scan.

Rates of pRNFL loss were generally greater in magnitude than those of mGCIPL, which is likely due to the differences in dynamic range of these tissues. Our finding is similar to previous studies that reported the rate of global pRNFL loss to be significantly faster than global mGCIPL loss [14, 27]. Axons from the inferotemporal, inferior, and inferonasal pGCIPL sectors all enter the optic nerve head primarily in the inferior pRNFL quadrant. Thus, pRNFL loss in this region can be thought of a composite “sum” of the respective mGCIPL losses, likely leading to a faster rate of loss when compared to just one mGCIPL sector. In addition, the Cirrus mGCIPL scan provides data only from a foveal cube, possibly missing macular damage outside of this area. However, those damaged axonal fibres would still be detected on the pRNFL scan. Prior studies have reported similar patterns of sectoral pRNFL and mGCIPL loss but in smaller cohorts of patients [13, 23, 27].

After adjusting for age, sex, CCT, IOP, and baseline thickness, we were impressed that NHB eyes still had significantly faster rates of pRNFL and mGCIPL loss. Numerous studies have demonstrated more progressive glaucomatous disease among Black patients [6, 16, 28, 29]. The Baltimore Eye Survey demonstrated that glaucoma prevalence among individuals of African descent is six times higher than that of the European American population, with a six-fold higher likelihood of glaucoma-related blindness [16]. The younger age of NHB subjects in our study (p < 0.001; Table 1) was also consistent with the findings of the Baltimore Eye Survey, which suggested that glaucoma tends to be diagnosed significantly earlier in the Black population. In a prospective cohort study spanning approximately 3 years and including a total of 528 healthy, glaucoma suspect, and glaucomatous eyes, Bowd et al. reported race-related differences in the rate of change of Bruch Membrane Opening-Minimum Rim Width, but only within the glaucoma suspect group [30]. No significant variation in the rate of change of pRNFL was observed. Notably, the authors did not stratify eyes with overt glaucoma according to the severity of the condition. Additionally, the study only analysed two groups (i.e., European descent and African descent), with a notable difference in the number of eyes included compared to our report. Recently, the investigators of the Primary Open-Angle African American Glaucoma Genetics (POAAGG) study described a median rate of pRNFL loss of −1.60 µm/year among Black subjects, nearly three times higher than a previously reported normal age-related rate of change of −0.52 µm/year [31]. These results are in contrast to the slower rates of pRNFL loss noted in the Duke Glaucoma Registry (−0.76 μm/year) and in the Diagnostic Innovations in Glaucoma Study (−0.98 μm/year), both which were predominantly comprised of White subjects (67.8% and 67%, respectively) [8, 14]. It is important to note that the POAAGG study had a non-comparative design, focusing solely on African American, African descent, or African Caribbean patients. Direct comparisons with other racial and ethnic groups were not made in the study, and the authors did not analyse mGCIPL data, sectoral pRNFL data, or stratify rates based on glaucomatous severity [31]. With respect to rates of OCT parameters among Hispanic patients, we believe that our study represents the first report documenting pRNFL and mGCIPL rates in a large clinical cohort.

Aetiologies for the faster rates of loss among NHB subjects could be related to different factors. A major strength of studying a clinically treated population is analysing the potential impact of non-clinical parameters on testing results, such as poor medication adherence, which are often non-issues in prospective studies due to exclusion criteria. Socioeconomic disparities and differences in access to healthcare have been described to contribute to differences in glaucoma outcomes among racial and ethnic groups [26, 32]. Black patients are more likely to face barriers to care such as lack of insurance, transportation difficulties, limited access to medications, as well as lower adherence to follow-up and treatment regimens, which can negatively impact health outcomes [26, 32, 33]. These disparities, coupled with worse baseline disease at diagnosis and at a younger age, could contribute to faster rates of pRNFL and mGCIPL loss, especially in earlier disease as observed in this study [26, 32]. In contrast, treatment adherence or surgical intervention might be greater in severe glaucoma given the perceptible impact on vision, leading to a non-significant difference in rates (Supplementary Table 3). Addressing these socioeconomic challenges may potentially ameliorate the disparities observed in this study, although further analyses are required as we were unable to distinguish between potential social influences. In addition, genetic differences could predispose NHB eyes to rapid disease progression; for example, the POAAGG study identified potential genetic factors that could contribute to worse disease among Black patients, specifically the TC genotype in a single nuclear polymorphism near the TRIM66 gene [31, 34]. In our study, CCT was significantly lower among NHB eyes (Table 2), potentially suggesting a difference in sclera and lamina cribrosa biomechanics, which may dictate how elevated IOP or IOP fluctuations impact pRNFL and mGCIPL loss. These findings highlight the complex interplay of genetic, ocular, and socioenvironmental factors that may influence the progression of glaucoma and contribute to the observed disparities between racial-ethnic groups [35].

Strengths of this analysis include the study cohort, which consists of a large, clinically treated population with considerable follow-up. Of note, our study population was unique given the substantial representation of Hispanic subjects, accounting for nearly 50%. In addition, the statistical modelling approach of longitudinal joint mixed modelling in this study minimises the degree of measurement error by allowing two associated outcomes to be modelled together. This study was also unique in its modelling of topographically related sections of pRNFL quadrants and mGCIPL sectors. Furthermore, excluding tests conducted after any surgical procedure carried out during the follow-up period enabled a more conservative analysis of the rate of pRNFL or mGCIPL loss, reducing the influence of iatrogenic factors.

Use of EHR data has its inherent limitations, including incorrect ICD or CPT coding. The diagnosis of glaucoma or glaucoma suspect was made based on the clinician’s judgement. In addition, any glaucoma surgeries that occurred outside of BPEI would not have been accounted for when excluding post-intervention tests. Furthermore, certain trends were observed in the data as previously noted, but these differences did not reach statistical significance in our current sample. These trends may represent true differences that would reach statistical significance with a larger sample size. Finally, our clinical population did not have sufficiently large cohorts of other racial-ethnic groups (e.g., Asian or Native American), limiting our conclusions regarding these individuals.

In summary, our analysis of a treated clinical glaucomatous population highlights significant differences in pRNFL and mGCIPL rates across different racial-ethnic groups after stratifying by disease severity. Rates of mGCIPL loss mirrored those of pRNFL, with non-Hispanic Black eyes having faster rates of pRNFL and mGCIPL loss among glaucoma suspect, mild glaucoma, and moderate glaucoma eyes. These findings reflect the importance of considering various aetiologies, such as socioeconomic disparities, that may contribute to these differences in rates of OCT loss among glaucoma patients.

Supplemental material is available at Eye’s website.