In this study, we evaluated multiple determinants of macular GCIPL thickness in normal Chinese adults and demonstrated that the thinning of GCIPL thickness was associated with older age, thinner pRNFL, and weaker OCT scanning signal strength. In general, gender, laterality, refractive status (when the refractive error is between + 1D and − 5D), IOP, CCT, and AL had no significant impacts on macular GCIPL thickness.
Age is one of the most significant impact factors in determining macular GCIPL thickness. In this study, we found that the overall trend of GCIPL thickness changing with age was as follows: GCIPL thickness increased slowly with age in younger adults; after reaching the peak at 40–49 years of age, it decreased rapidly with age, which was consistent with the findings of Mwanza et al.(17) However, our previous findings in normal Chinese subjects indicated that pRNFL thickness was comparatively thicker in teenagers and reached its peak at 20–29 years of age, then gradually became thinner with age.(25) Similar results were found in studies in other Asian populations.(26, 27) These findings suggested that GCIPL and RNFL thickness changes may not necessarily be synchronized.
We found that for each additional year over 40 years of age, the average, minimum, superotemporal, superior, superonasal, inferonasal, inferior, and inferotemporal GCIPL thickness decreased by 0.229, 0.200, 0.127, 0.249, 0.273, 0.295, 0.273, and 0.173 µm, respectively. Animal experiments demonstrated that the age-related change of RGC predominately manifested as axon loss, while the RGC cell count is relatively constant.(28, 29) As the retina expands with age while the total cell counts retain, the RGC density decreases. The phenomenon of retina expansion with age was also found in human eyes, but the main difference with the animal eyes was that the number of RGC soma also declined with age in human.(30) Therefore, it is explainable that the GCIPL gets thinner with age in an OCT-based thickness evaluation, as shown in this study. It has been proven by multiple studies that the GCC or GCIPL thickness decreased with age, even though the age-related RGC loss varies in extent: Balazsi et al,(31) Repka et al,(32) Mikelberg et al,(33) Jonas et al,(34) Blanks et al,(35) Harman et al,(36) and Kerrigan-baumrind et al(37) respectively reported the annual RGC loss was between 0.07–0.61%, in studies with the sample size between 12 eyes to 72 eyes. In this study, we further approved that the age-dependent GCIPL thickness change was nonlinear with age. However, this age-related variability of the OCT measurements may not be completely attributed to inter-subject variability in retinal neurology, which is considerably significant even in normal human eyes. When ganglion cells reduce with age, the migrant amacrine cells and other non-neuronal components may partially compensate the space which is previously predominated by ganglion cells. As such, the actual cell loss may be masqueraded and the age-RGC loss correlation may become more unpredictable. Moreover, the RGC layered in the macular region, making it more complicated to evaluate the defined pattern of region- and eccentricity-associated, age-dependent RGC loss.
When evaluating the potential causative impacts of axial length on GCIPL thickness, just as the investigations concerning its impacts on RNFL thickness, contradictory conclusions were drawn. Some studies proposed that the GCIPL and RNFL thickness were negatively correlated to axial length.(38–41) Studies with a larger sample size and/or a wider range of refractive status, generally indicated that only less than 0.5% GCIPL thickness change was attributed to per millimeter axial length change.(16–18) Such minor changes could hardly reflect any practical clinical significance. Another reason that axial length may have some impacts on the GCIPL thickness measurement, but not necessarily the actual anatomic cell counts may be ascribed to the optical effects. Since the Cirrus OCT model eye adopts a calibrated value of 24.46 mm as the default axial length setting with a fixed measuring angular distance of approximately 12°, the actual scanning area would be larger than the “standard” retinal area due to the optical magnification effect in eyes longer than 24.46 mm. As the macular ganglion cell counts drop dramatically as the eccentricity increases outward from the foveal center, average ganglion cell estimates or GCIPL thickness in these eyes may therefore be underestimated for this reason. On the contrary, in eyes shorter than the set value, the actual scanning area is smaller than the preset area where the ganglion cells are more crowded and thus a thicker GCIPL measurement may be generated falsely.
Other studies declined the direct impacts of axial length on RGC growth or apoptosis and claimed that axial length had no significant correlation with macular GCIPL thickness,(42, 43) to which our findings was consistent with. The relatively small sample size could be one possible cause. Also, our inclusive criteria for spherical equivalent refractive error were − 5.00 D and + 1.00 D. The exclusion of highly myopic eyes restricted the variability of axial length, thus minimizing the interference of magnification effects on location of the retinal area being scanned that could potentially influence the GCIPL thickness measurements. Similarly, we didn’t find significant association between refractive error and GCIPL thickness. The difference of average, minimum, and most sectoral GCIPL thickness between the two refractive groups were not significantly different, except that in inferonasal and inferior sectors. Histological studies of both human and animals have found that the RGC were denser nasally than temporally, and superiorly than inferiorly, with distinct inter-subject variability, which might have indicated the discrepant anatomic distribution pattern in emmetropic and myopic eyes.(44) The thickest GCIPL was detected in the superonasal sector, in which no significant difference in emmetropic and myopic groups was found, suggesting that the density in this sector may have partially offset the difference caused by refractive error and/or axial length. This sector may have poorer performance in diagnosing glaucoma due to its least glaucomatous susceptibility.
The finding that pRNFL thickness had significant positive correlation with macular GCIPL thickness was not surprising and was consistent with previous studies.(17, 45) As the axon and soma of the ganglion cells, these two cellular components are closely related and both can be remarkably affected by glaucoma. Thus, these two parameters are both important and sensitive for early detection of glaucoma. The regression analysis showed that the mean RNFL thickness decreased by 0.901 µm for every 1 µm decrease in average GCIPL thickness. Overk et al(46) found that lesions in the axon may occur earlier than that in the soma in some neurodegenerative diseases such as Alzheimer's disease and Parkinson's disease, indicating a ‘reverse’ pathogenesis pathway for the primary causative factor of glaucoma. In the pathophysiological development process of glaucoma, whether GCIPL is affected primarily and causes changes in pRNFL, or vice versa, still needs further investigations.
The stronger the OCT signal strength, the deeper retinal tissue the light achieved.(47) As the reflection of the boundaries got enhanced, the segmentation of each layer was more accurate. The signal strength of Cirrus OCT was found to be positively correlated with RNFL thickness.(48) When the signal strength reaches 7 or more, the RNFL is less likely to measure thicker as the signal strength increases. In this study, although there was significant difference between the signal strength of the Macula Cube 512 × 128 scanning protocol and that of the 200 × 200 scanning protocol, all signal strength, as our inclusion criteria indicated, was ≥ 6. No significant difference was found in GCIPL thickness measured by the two scanning protocols, which may suggest that the GCIPL thickness measurement was stable and accurate when the signal strength is strong enough. The correlation between the signal strength of the 512 × 128 protocol and the GCIPL thickness was not as strong as that between the signal strength of the 200 × 200 protocol and the GCIPL thickness. This could be explained by the denser scan, which is more likely to generate better image quality, and the stronger signal strength of the former protocol.
There were several limitations of this study. First, it was a cross-sectional retrospective study with comparatively small sample size. Second, only ocular predictors were evaluated. Systemic predictors such as history of diabetes, cigarette smoking history, blood pressure, serum lipid levels should be taken into account. Third, there was a lack of a multivariable model synthesizing the effects of age and RNFL thickness, which were the two potential determinants found using univariable regression model. More comprehensive investigations are expected in future studies.