SD-OCT allows the reproducible and accurate measurement of several thickness parameters, including circumpapillary RNFL (cpRNFL), and macular thickness parameters, one of which is the ganglion cell complex (GCC; RNFL + ganglion cell layer [GCL] + inner plexiform layer [IPL]) thickness, and another is the GCIPL (Carl Zeiss Meditec)/GCL + (Topcon, Tokyo, Japan) (GCL + IPL) thickness. To determine if the thickness parameters in each eye are abnormally diminished, they are statistically compared, by using confidence intervals (CIs), with the parameters of a normative database of the corresponding age range and gender. Abnormal thinning (red color) is defined as falling below the lower 99th percentile of the CI limit of the normative database, and borderline thinning (colored with yellow) as between the lower 95th and 99th percentile CI limits (Fig. 3.6). Such a method, based on comparison of measurements, has the following three limitations that lower the ability to identify glaucoma:
The first and second limitations addressed above hold true regardless of the presence, absence, or severity of myopia. In particular, a large overlap between normal and early diseased eyes due to the wide distribution of normal thickness values clearly indicates the difficulty in perfectly discriminating normal from early glaucomatous eyes. Underestimation resulting from area averaging further decreases the ability to discriminate for all the thickness parameters, as does the third limitation mentioned, which is specific to myopic eyes. Thus, we need to interpret the results of statistical comparisons with CIs from normative databases with these limitations in mind.

Myopic deformation of peripapillary structures leads to failures in cpRNFL thickness measurement in some highly myopic eyes (7.1 %) [28]. Specifically, two types of deformation are responsible for cpRNFL segmentation (software generation of boundary lines) failure: steep peripapillary scleral slope and large PPA. The cpRNFL varies in height in eyes with high myopia, being most commonly highest in the nasal quadrant and lowest in the temporal quadrant, giving the structure a dome-like appearance in cpRNFL B-scan (Fig. 3.8) [28]. The temporal-most portion of the cpRNFL (the bottom of the dome-like appearance) appears in the inferior half of the image acquisition frame and generally shows low signal intensity [28]. This is due to the fact that signal intensity decays with deeper axial distance, a disadvantage of SD-OCT technology [73]. In eyes with an inferiorly tilted disc, the inferior peripapillary portion is much lower than superior one, which results in a sigmoid appearance. Consequently, the signal intensity of the inferior cpRNFL is rendered very low. Such low signal intensity can result in segmentation algorithms being unable to recognize the RNFL boundaries, straying instead into the deeper retina (Fig. 3.8) [28]. Another common cause of segmentation failure in highly myopic eyes is inclusion of PPA on the circumpapillary circle scan. In this case, the segmentation algorithm appears to be pulled away from the cpRNFL boundaries by the high scleral reflectivity within the PPA (Fig. 3.8) [28].

We now understand that highly myopic eyes often have structural limitations, such as abnormal profiles of the cpRNFL and peripapillary steep slope. These limitations may be responsible for inaccuracies in cpRNFL measurements to detect glaucoma. A receiver operating characteristic (ROC) regression analysis [74, 75] was conducted in the following model considering the modeling covariates age, sex, visual field mean deviation (MD), axial length, and signal strength index:
where Φ = probit function, MD = visual field mean deviation, AL = axial length, and SSI = signal strength index.

This model revealed that the effects of longer axial length on glaucoma diagnostic performance differ depending on specific thickness parameters used. Using cpRNFL thickness, a longer axial length was likely to be associated with a poorer diagnostic performance, whereas this was not the case when macular GCC thickness was used [28].

A deviation map is a display of abnormally and borderline-thinned areas on the OCT projection fundus image (Fig. 3.9). Thickness measurements are performed on each A-scan. When the measurement of each A-scan, or the average measurement of several neighboring A-points (the “superpixel”), is below the lower 99th percentile of the CI limit of the normative database, and between the lower 95th and 99th percentile CI limits, it is exhibited in red (abnormal) and yellow (borderline), respectively. In the deviation map, glaucomatous RNFL damage often appears as a pattern of RNFL defects (Figs. 3.9 and 3.10) [76, 78]. These defects occur along the retinal nerve fibers that arise from the optic disc and extend radially or in an arcuate shape. Because false-positive A-scans or superpixels are rarely arranged in this unique pattern, an RNFL defect-like abnormal area in the deviation map is an excellent indicator of glaucomatous damage and its progression (Figs. 3.9 and 3.10) [76, 78]. However, we need to note that this pattern also often gives rise to false-positives in high myopia.

Compared to non-highly myopic eyes, in eyes with high myopia, the superotemporal and inferotemporal arcuate regions with the thickest RNFL (the double hump) are shifted temporally. The RNFL double-hump angle decreases by approximately 3.38° for every 1-mm increase in axial length [79]. In contrast, normative databases in commercially available SD-OCT devices often primarily contain data from non-highly myopic eyes. For this reason, superotemporal and inferotemporal regions where double humps (thickest RNFL) are located in non-highly myopic eyes are classified as abnormal thinning in highly myopic eyes (Fig. 3.11) when compared with a normative database. The abnormally thinned area has a typical RNFL pattern. However, this finding is a false-positive resulting from the application of a normative database without refractive matching. Thus, we need to exercise care in interpreting deviation maps from eyes with high myopia in OCT devices where a normative database for the condition is not available. Currently, to the best of our knowledge, such a database is only available in the RS-3000 Advance (Nidek).

It is not only in SD-OCT that average measurements have been compared with normative databases, it is also a standard practice for images from the Heidelberg Retina Tomograph (HRT) confocal laser scanning ophthalmoscope (Heidelberg Engineering, Heidelberg, Germany), the GDx-VCC scanning laser polarimeter (Carl Zeiss Meditec), and the Stratus time-domain OCT (Carl Zeiss Meditec). SD-OCT has higher axial resolution than these conventional imaging devices and much higher imaging speed than time-domain OCT. These advantages of SD-OCT are useful for increasing measurement accuracy and reproducibility [80], but not in resolving the three limitations mentioned above. Speckle noise-free B-scans, which are currently standard B-scans in all SD-OCT instruments, allow observation of single retinal layers, including the GCL, and peripapillary structures, involving optic disc margin anatomy. Because resolution is thus drastically improved in SD-OCT technologies, we should develop further diagnostic strategies based on direct observation of GON-related structural changes, rather than statistical inter-person comparison alone.

Individual macular retinal layers, including the RNFL and GCL, are highly symmetrical between superior and inferior hemispheres [27, 81]. Although each of the macular retinal layers decays with aging, the symmetrical shape remains with aging (Fig. 3.12) [81]. Macular GCL thickness increases rapidly to a peak around 1 mm superior and inferior to the foveal center (Figs. 3.13 and 3.14) [27] and gradually decreases with greater distance from the fovea (Figs. 3.13 and 3.14). Furthermore, the macular GCL is slightly thinner on the temporal than on the nasal side (Fig. 3.14) [27]. As a result, the GCL thickness map generated from three-dimensional macular raster scans shows a Landolt ring-like shape (Fig. 3.12) [81]. Importantly, this characteristic GCL structure is almost identical to the RGC density topography that has been obtained from enucleated human eyes (Fig. 3.14) [82].

In glaucoma, the symmetric GCL structure is apparently diminished in the locations of the corresponding VF defects and optic disc neuroretinal rim thinning/RNFL defects (Fig. 3.15) [28]. Thinning of the GCL is even seen in a large part of eyes with preperimetric glaucoma (Fig. 3.16) [27]. At this early stage, GCL thinning appears to be localized and abrupt.

The photographic appearance of the optic disc, including disc size, tilting, cyclotorsion, and shape, was highly variable in highly myopic eyes, regardless of VF defect severity (Fig. 3.17) [19–22, 28]. The area of PPA also varied in size and circumference (Figs. 3.2, 3.3, and 3.4). In non-glaucomatous highly myopic eyes, macular layer structure is uniform regardless of the highly variable myopia-related optic disc deformation, which includes tilting, cyclotorsion, and PPA (Fig. 3.17) [28]. Importantly, the uniform layer structure has highly symmetrical shapes in vertical scans, even in highly myopic eyes [28]. The macula lies centrally along the globe’s optical axis, whereas the optic disc is nasally shifted from the optical axis. The fact that the macula is less affected by myopic globe elongation than both the optic disc and the peripapillary sclera is probably due to this difference in position. Excessive myopic globe elongation leads to abnormal scleral extension in the macula, which causes posterior staphyloma. Posterior staphyloma in these eyes mainly shows Curtin classification type I [83], in which the macular layer symmetry remains [28]. In rare cases of type V (inferior staphyloma), type VI/VII (combined staphyloma), and type IX (septal staphyloma), the macular structure may be highly affected. Eyes with such asymmetrical deformation of posterior eye globe often have an extremely long axial length (~30 mm and sometimes more). Such cases may be better classified as pathological myopia. In the glaucoma service, highly myopic glaucomatous eyes usually have an axial length of 28 mm and less and often have symmetrical macular layer structures.

Subjective evaluation of structural optic disc abnormalities is not favored in the field of glaucoma diagnosis research. This is because interobserver reproducibility for evaluation of the optic disc appearance is poor. However, in clinical practice, optic disc evaluation has been a gold standard criterion for glaucoma diagnosis. The advent of SD-OCT has allowed clear observation of each macular layer, including the RNFL and GCL. This has reintroduced the possibility that subjective evaluation of cross-sectional RNFL and GCL images may be a useful method for glaucoma diagnosis [27]. As discussed above, variable deformation of the optic disc makes it difficult to detect early glaucomatous changes in the optic disc, particularly in high myopia. Because the macula is highly uniform in size and shape regardless of the highly variable myopic changes of the optic disc, evaluation of the cross-sectional macular retinal layers may now be reproducible and accurate. To test this hypothesis, we asked well-trained glaucoma specialists to evaluate optic disc photos and serial vertical SD-OCT B-scans and found that in non-highly myopic eyes, interobserver agreement was excellent for both optic disc photos and SD-OCT B-scans, whereas in highly myopic eyes, interobserver agreement was poor for optic disc photos but excellent for SD-OCT B-scans [28]. In addition, in non-highly myopic eyes, classification of glaucomatous and non-glaucomatous eyes was almost perfect for both optic disc photos and SD-OCT B-scans, while in highly myopic eyes, photographic classification was less accurate than SD-OCT classification. Thus, subjective assessment of SD-OCT macular serial vertical scans may be useful for supporting glaucoma diagnoses made using photographic disc assessment and software analysis in highly myopic eyes.

Glaucoma patients who have paracentral scotomas are at high risk of visual acuity loss [84, 85]. Moreover, paracentral VF defects can occur even in the early stages of the disease [18, 86–88]. Oversight or delay in detection of such defects causes physicians to underestimate the severity of glaucoma and progression of central vision loss. Risk factors for developing paracentral scotomas are normal-tension glaucoma (NTG) [89, 90], low maximum untreated intraocular pressures (IOPs) [91], frequent disc hemorrhage [91], systemic factors [91], and high myopia [16, 17, 42, 92, 93]. In advanced glaucoma, high myopia has been associated with significantly higher frequencies of cecocentral scotomas located just temporal and inferior to the fixation point [16, 17]. Atypical RNFL defects, including papillomacular bundle defects, are found in highly myopic eyes with primarily moderate-to-severe VF defects [42]. In this report, longer axial length, larger optic disc, and normal-tension glaucoma are risk factors for papillomacular bundle defects. Importantly, high myopia is a significant risk factor for papillomacular bundle defects and inferotemporal paracentral VF defects, even in early glaucoma [18]. Thus, it is of paramount importance to detect paracentral scotomas as early as possible, particularly in highly myopic eyes. That said, evaluation of optic disc appearance to predict paracentral VF defects is difficult, even in non-highly myopic eyes. This is much more the case in highly myopic eyes.