Discussion
The tissue thickness of the cpRNFL and mGCL++ demonstrated the highest relative loss in early glaucoma, followed by micro-VD, specific perimetry and SAP. Previous studies have reported tissue thinning between the RNFL and IPL at the macular region,7 25 26 with cpRNFL26–28 appearing before the decreasing retinal sensitivity measured with specific perimetry, including FDT,29 30 SWAP,31 Flicker2 or Pulsar3 11 perimetry, and before its detection by SAP. The difference between normal and suspected glaucoma or early glaucoma has been compared using measurement values or assessed using AUC values. Although measurements such as the tissue thickness, micro-VD and visual field sensitivity have similar sensitivity to detect an abnormality and measurement variability, the degree of differences among the tissue thickness, micro-VD and visual field sensitivity cannot be directly compared because of their different units and dynamic ranges. Therefore, for an accurate and direct comparison, they should be converted to relative loss values.
Moghimi et al reported a floor value where no further structural change could be detected, neither in tissue thickness nor in micro-VD.32 The dynamic range adjusted with the floor value in their study was 43.9 µm for the cpRNFL, 22.8 µm for the GCC (corresponding to GCL++ in this study), 16.7% for the perifoveal VD and 18.8% for the cpVD. The dynamic range in our study was 59.0 µm for cpRNFL, 77.2 µm for GCL++, 10.5% for mVD and 25.1% for cpVD, thus differing from Moghimi et al’s report,32 potentially owing to differences in the measurement device, analysis area and method. However, the dynamic range of the micro-VD was narrower than the tissue thickness, suggesting that adjusting for the dynamic range is required for an accurate and direct comparison of the degree of loss among the tissue thickness, micro-VD and visual field sensitivity. Although the data in our study were obtained with a cross-sectional design, the degree of relative tissue thickness loss adjusted by the dynamic range for the GCL++ (−24.7%) and cpRNFL (−25.8%) revealed significantly more relative loss than the visual field, including Pulsar (−10.1%) and HFA (−5.9%), and the Pulsar also demonstrated more relative loss compared with HFA.
According to the vascular theory of glaucoma, optic nerve damage is caused by reduced ocular blood flow, potentially leading to axonal ischaemia.33 Supporting this assumption, Shiga et al,34 using laser flow speckle flowgraphy, reported a significantly reduced ocular blood flow (mean blur rate) in patients with preperimetric glaucoma (11.8±2.4, arbitrary unit) compared with normal subjects (13.5±2.6, arbitrary unit); cpRNFL was also significantly reduced in patients with preperimetric glaucoma (80.00±8.80 µm) compared with normal subjects (92.56±6.19 µm). Although it was not calculated in the above paper,34 when the effect size, representing the extent of differences, was calculated based on the above results,34 the cpRNFL (effect size d was 1.7) was larger than the ocular blood flow (effect size d was 0.68). Thus, the destruction of neural tissue in glaucoma may be greater than vascular dysfunction only on the retinal surface or optic nerve head.
Micro-VD loss in the macular or circumpapillary regions is a sign of an early glaucoma change.7 25 26 35–37 Hou et al7 detected both mGCC thinning and micro-VD loss in the whole macular region in early glaucoma, although the relative loss of mGCC (9.8%–9.9%) was greater than that of micro-VD (6.6%–6.9%) in early glaucoma. Wang et al25 reported a greater relative loss of GCC (22.0%) than mVD (15.4%) using a similar protocol. This study also demonstrated a higher relative loss in GCL++ (−24.7%) than in mVD (−17.3%). The degree of thinning of the macular tissue was larger than that of micro-VD in early glaucoma. However, previous studies did not compare the cpVD and cpRNFL regions.7 25 Here, the relative loss value of the cpRNFL (−25.8%) was also higher than that of cpVD (−17.0%). These results clearly showed that the degree of tissue thickness loss in both macular and circumpapillary regions was larger than that of micro-VD in early glaucoma. However, only morphological changes in the surface layer of the retina were measured both in previous studies7 25 and in this study. In this study, morphological changes in the deeper layers at the very early stage of glaucoma were not measured.
A previous in vivo study in non-human primates with experimental glaucoma reported a substantial loss of anterior orbital optic nerve axons (~10%–15%) before the appearance of RNFL thinning with OCT.38 Additionally, cpRNFL retardance, represented by birefringent changes in the ordered structural array of cytoskeletal proteins within its axons, measured using scanning laser polarimetry, occurred before cpRNFL thinning in both cross-sectional39 and longitudinal40 studies. However, morphological changes occurring at a very early stage of glaucoma could not be demonstrated with our study design.
Another novelty of our study was the evaluation of the degree of loss in specific perimetry. Pulsar perimetry, following a similar principle as FDT,34 has been used to propose that, in its early stages, glaucoma may predominantly damage the magnocellular retinal ganglion cells projecting into the magnocellular layers of the lateral geniculate nucleus.4 11 41 Accordingly, we hypothesised that the degree of loss in the Pulsar would be more similar to the tissue thickness than to the micro-VD; however, we observed that the degree of loss in the tissue thickness (−25.8% to −21.2%) and micro-VD (−17.3% to −17.0%) was larger than that in the Pulsar (−10.1%). Regarding detectability, Pulsar (AUC=0.78) was comparable with micro-VDs (AUC=0.78) and standard perimetry (AUC=0.78). A previous study also reported no significant difference in the diagnostic performance between SAP and matrix FDT.42 This may be attributed to the fact that only the magnocellular pathway is not significantly attenuated at the very early stages of glaucoma or that the Pulsar stimulus may reflect the magnocellular component of the ganglion cells, which has a minor redundancy in the human retina, as well as the parvocellular component of the ganglion cells, which is highly redundant in the human retina.
Previous studies reported that cpVD (AUC=0.86–0.93) demonstrated better discriminative ability than mVD (AUC=0.71–0.87) for differentiating between early glaucoma and normal eyes,43–45 whereas one study reported more prominent mVD damage in the peripheral area.26 In our study, the AUC was 0.78 for both mVD and cpVD. The discrimination ability between the previous studies43–45 and our study may be attributed to the difference in analysis areas. Regarding the circumpapillary region, in previous reports, the region of interest consisting of two circles was either wide (inner circle: 2 mm diameter; outer circle: 6 mm diameter)43 or small (inner circle: BM opening; outer circle: 0.75 mm outwards from the inner circle)26 44 45 compared with the current study (inner circle: 3.2 mm, outer circle: 3.6 mm). In the macular region, the region of interest in a previous study was smaller (a 1.5 mm-wide circular annulus centred on the macula captured with 3×3 mm)44 45 than that in this study (a 6×6 mm square centred on the macula). However, the AUC value within the same analysis region was comparable between a previous study (AUC=0.77)26 and our study (AUC=0.78), suggesting the need for standardisation of the analysis area in micro-VD measurements.
There were limitations in the current study. First, of the 90 eyes in patients with glaucoma, 74 used at least prostaglandin eye drops. A previous study reported that VD increased due to IOP lowering with prostaglandin eye drops,46 suggesting that this may have influenced the small relative loss in micro-VDs compared with the tissue thickness of cpRNFL and mGCL++. Accurate comparisons would be needed in eyes without eye drops. Second, the inclusion criterion for patients with glaucoma was the presence of typical glaucomatous changes in the optic nerve head. Therefore, the degree of relative loss in structural changes may have been greater than that in functional changes. Third, it was not possible to match the analysis area of OCT and OCT-A imaging by correcting the ocular magnification because the scan size was different between OCT and OCT-A imaging.
In conclusion, although the data were obtained with a cross-sectional design, the dynamic range-adjusted loss in early glaucoma was higher in the tissue thickness change, followed by micro-VD and retinal sensitivity changes with specific perimetry and later with SAP. Discrimination ability between glaucomatous and healthy eyes was also the highest in tissue thickness and comparable between the micro-VD and visual fields.