This multicenter cross-sectional study was conducted at the First Affiliated Hospital of Kunming Medical University and Nanjing Medical University, China. Glaucoma and myopia patients were recruited between March 2022 and June 2023. The Institutional Review Board of Kunming Medical University and the First Affiliated Hospital of Nanjing Medical University approved the study. Informed consent was obtained from all the participants. And Ethics Committee approval was obtained. This project was conducted in line with the principles of the Helsinki Declaration. A flowchart illustrating the case–control matching process is shown in Figure 1.

All subjects underwent complete ophthalmic examinations, including slit-lamp biomicroscopy, best-corrected visual acuity (BCVA), intraocular pressure (IOP), gonioscopy, axial length (AL) (IOL Master 700; Carl Zeiss Meditec, Dublin, California, USA) and central corneal thickness (CCT; DGH-550; DGH Technology, Exton, Pennsylvania, USA), and standard automated perimetry (Carl Zeiss Meditec, Dublin, CA, USA). Peripapillary RNFL thickness was measured by spectral-domain (SD) OCT (RTVue, XR, Avanti, Optuovue, Inc., Fremont, CA, USA), and peripapillary VD was obtained by OCTA imaging (RTVue, XR, Avanti, Optuovue, Inc., Fremont, CA, USA).

The inclusion criteria for the glaucoma group were: (1) all glaucoma patients had typical glaucoma ocular nerve head damage (vertical cup-to-disc ratio of > 0.6, cup-to-disc asymmetry of > 0.2, the presence of neuroretinal rim thinning or notching, peripapillary hemorrhages) or RNFL defect visible on slit-lamp biomicroscopy and OCT; and (2) VF defects corresponding to the nerve fiber bundle pattern, meeting at least one of the following criteria: nasal steps, paracentral scotomata, arcuate defect, temporal island, and central island. Pattern standard deviation (PSD) outside normal limits (p < 0.05) or glaucoma hemifield test outside normal limits.

For the control group, the inclusion criteria were: (1) spherical equivalent (SE) ≤ −6.00D and −0.50D ≥ SE > −6.00D; (2) no evidence of retinal pathology or glaucoma; (3) a normal Humphrey 30–2 VF; (4) intraocular pressure < 21 mmHg; (5) no long-term ocular or systemic corticosteroid use.

The exclusion criteria for glaucoma and myopia groups were: (1) presence of any diseases other than glaucoma or myopia that could cause VF loss; (2) inability to perform reliable automated VF testing; (3) history of prior ocular surgery or trauma, including refractive surgery, cataract surgery, vitreoretinal surgery, or corneal surgery.

RNFL imaging was performed with the SD-OCT device (RTVue, XR, Avanti, Optuovue, Inc., Fremont, CA, USA) using a scanning laser diode to emit a scan beam with a wavelength of 840 nm to provide images of ocular microstructures (13). Using the 4.5-mm diameter RTVue protocols that calculate the peripapillary images of nerve fiber layer thickness from ILM to NFL layer. The peripapillary RNFL thickness parameters were automatically calculated by the SD-OCT and divided into eight regions. Criteria for determining scan quality included a signal strength of 30 or higher for RTVue, a clear fundus image where the optic disc and scan circle were visible both before and during image acquisition, and the absence of en-face OCT image distortions caused by blinking or eye movements. Additionally, OCT scans with algorithmic failures in retinal layer segmentation that could not be corrected manually were excluded.

The OCT-A images were acquired utilizing a commercially available spectral-domain OCT device (RTVue, XR, Avanti; Optuovue, Inc., Fremont, CA, USA). This system operates at a scan rate of 70,000 A-scans per second, with the scan beam wavelength set at a central value of 840 nm and a bandwidth of 45 nm. The cube scanning protocol involved 73 B-scans, each composed of 768 A-scans, and these scans were captured within a 4.5 × 4.5 mm squared region centered on the optic nerve head (ONH). The capillary densities from the internal limiting membrane (ILM) to the posterior boundary of the retinal nerve fiber layer (RNFL) were visualized using the standard AngioVue software, and the resultant measurement was referred to as the RPC density image. For the quantification of capillary density, the built-in large-vessel masking function was employed. This function eliminates vessels with a diameter of ≥ 3 pixels. In the context of 4.5-mm AngioVue optic disc (AngioDisc) scans (Optuovue 2018.1.1.63 software version), a 3-pixel diameter roughly corresponds to a physical dimension of ≥ 33 μm (14). The peripapillary region was measured at a 1-mm-wide elliptical annulus extending outward from the optic disc boundary (15). The vessel density (VD) was defined as the percentage area occupied by microvasculature, and the central 2 mm diameter circle centered on the optic disc based on the en-face reflectance image was excluded in the 4.5 × 4.5-mm scan area (16). Moreover, the boundary of the optic disc was manually delineated for patients with high myopia. The center of the optic disc was determined by automatically finding the best ellipse fit of the optic disc boundary (Figure 2). The OCTA images with a resolution of 304 × 304 pixels were processed using the Split-Spectrum Amplitude Decorrelation Angiography (SSADA) algorithm to enhance their visualization. Images were rechecked or excluded if they had (1) scan quality < 7; (2) poor clarity; (3) motion artifacts; (4) residual motion artifacts visible as irregular vessel patterns on the en face angiogram; (5) images with fixation error; or (6) segmentation failure.

The peripapillary region of optic nerve head was divided into eight sectors following the regionalization previously described by Garway-Heath (17), and pVD and pRNFL thickness in each sector was calculated. Including: temporal superior (TS), temporal inferior (TI), superior temporal (ST), inferior temporal (IT), superior nasal (SN), inferior nasal (IN), nasal superior (NS), and nasal inferior (NI).

Automated visual field tests were performed with the 30–2 SITA standard program on the Humphrey 750i Visual Field Analyzer (Carl Zeiss Meditec, Dublin, CA, USA) and were grouped into eight sectors that matched the structure (Figure 3) (17). Mean deviation (MD) and pattern standard deviation (PSD) were analyzed. A reliable visual field test was defined: fixation loss rate < 20%, false positive rate < 15%, false negative rate < 15%. The severity of glaucomatous damage was classified as early stage (VF mean deviation (MD) ≥ −6 dB), moderate stage (−12 ≤ MD < −6 dB), and advanced stage (MD < −12 dB).

Based on our previous research method (9), the VFMS at various sectors were defined as the average value of the differential light sensitivity (DLS) obtained at VF test locations corresponding to pVD and pRNFL sectors. The VF used logarithmic units of measurement, decibels (dB), measuring the DLS of the subject, and expressed as non-log-transformed 1/L. When directly analyzing the correlation between the dB value and pVD or pRNFL, a logarithmic curve relationship was obtained (18). According to the formula dB = −10 × log10(L), dB values can be divided by 10 to obtain the log10 values, and then calculate the non-logarithmic to get 1/L (1/L represents the maximum brightness of the instrument/the brightness of the visual object actually detected, L: luminance measured in Lamberts) (19). VFMS presents the average value calculated from each area for l/L, according to the eight sectors of pVD or pRNFL. Then the vasculature-function or structure–function relationships were analyzed by comparing the pVD or pRNFL to the corresponding VFMS.

Propensity scores were estimated using a multivariate logistic regression model with glaucoma as a dependent variable and potential confounders of age and AL as covariates. Using a nearest neighbor matching algorithm with a matching ratio 1:1, the caliper value was set at 0.02. A caliper value of 0.02 was considered to be relatively moderate. It can ensure a certain sample size, effectively balance the covariables, and concurrently reduce the differences between groups (20).

All statistical analyses were performed using SPSS software version 26.0 (IBM Corp., Armonk, NY, USA) and R language version 4.2.3. (1) By K-S single sample test, the population was normal distribution, descriptive statistical analyses [mean ± standard deviation (SD) were used in the evaluation of the data; categorical variables were presented as counts and percentages and were analyzed using the chi-square test]. Differences between groups were compared using Student’s t-test, and differences among more than two groups were compared using ANOVA (One-way analysis of variance). (2) The correlation between pVD, pRNFL, and visual function was analyzed using the Pearson correlation coefficient (r). Given that the data demonstrated a normal distribution, we used the Pearson correlation coefficient to evaluate the linear relationship between two continuous variables. (3) Univariate linear regression analyses were performed to examine the associations between pVD, pRNFL, and VFMS. (4) Receiver operating characteristic (ROC) curves were calculated to indicate the separation between glaucoma and control groups. The area under the receiver operating characteristic curves (AUC) was calculated and compared. The Delong method was used to ascertain the statistical significance of AUC differences. An AUC > 0.7 indicated that the classifier provides clinically meaningful discriminative ability. Multiple comparison corrections for ROC analyses were performed using the Bonferroni method, setting a significance threshold of p < 0.00625. Corrected p-values are available upon request or are presented in the Supplementary material. (5) Statistical significance was set at p < 0.05.

A total of 191 eyes diagnosed with glaucoma and myopia were matched to 580 eyes diagnosed with myopia without glaucoma. Following the matching process (illustrated in Figure 1), 382 eyes were included in the study. These were divided into four groups: 101 eyes with highly myopic glaucoma (HMG, Group 1), 101 eyes with high myopia (HM) without glaucoma (Group 2), 90 eyes with non-highly myopic glaucoma (NHMG, Group 3), and 90 eyes with non-high myopia (NHM) without glaucoma (Group 4). The demographic and clinical characteristics of the four groups were summarized in Table 1. Prior to matching, there was a significant difference in baseline age and axial length (AL) between the glaucoma patients and their corresponding myopia control groups (p < 0.05). The standardized differences of age and AL was greater than 20%. After propensity score matching, no significant differences were observed in age and AL parameters between the two paired groups (p > 0.05). The standardized differences of age and AL were controlled within 20%. Through Propensity Score Matching (PSM), the balance of various covariates has been significantly improved.

The comparisons of average and regional pVD, pRNFL and VFMS are summarized in Table 2 and Figure 4. In the analysis of HMG eyes (Group 1) versus HM without glaucoma eyes (Group 2), the average pVD was significantly lower in Group 1 (37.66 ± 7.61) compared to Group 2 (46.40 ± 5.16, p < 0.001). The pVD in all eight regions was significantly reduced in Group 1 compared to Group 2 (p < 0.05), except for the NI region (p = 0.147). When comparing NHMG eyes (Group 3) with NHM without glaucoma eyes (Group 4), the average pVD was also significantly lower in Group 3 (37.18 ± 8.43) than in Group 4 (52.69 ± 3.60, p < 0.001). The pVD in all eight regions in Group 3 was significantly reduced compared to Group 4 (p < 0.001). Multiple comparison analyses showed no significant differences in average and regional pVD between Group 1 and Group 3 (p > 0.05), except for the TI and NI regions (p < 0.05). Both glaucoma groups (Group 1 and 3) had significantly lower average and regional pVD compared to their respective control groups (Group 2 and 4) (p < 0.05), except for the NI and NS regions (p > 0.05). Comparisons between Group 2 and 4 indicated that high myopia led to decreased pVD across all regions (p < 0.05).

Group 1 showed significantly thinner average pRNFL thickness (71.13 ± 18.66 μm) compared to Group 2 (101.22 ± 16.79 μm, p < 0.001), with reductions in all regions (p < 0.001) except NI (p = 0.906) and NS (p = 0.815). Similar reductions were observed between Groups 3 and 4 (70.14 ± 18.09 μm vs. 113.02 ± 18.83 μm, p < 0.001). No significant differences in pRNFL were found between Groups 1 and 3 (p > 0.05). Both glaucoma groups (Groups 1 and 3) had thinner average and regional pRNFL than controls (Groups 2 and 4) (p < 0.05), except in NI and NS regions (p > 0.05). High myopia was associated with reduced average and regional pRNFL thickness (p < 0.05).

Average VFMS was significantly lower in Group 1 (576.38 ± 412.96) compared to Group 2 (855.11 ± 328.49, p < 0.001), with similar regional reductions (p < 0.001), except in NI and NS regions (p > 0.05). Comparable trends were observed between Groups 3 and 4 (545.01 ± 439.04 vs. 1051.31 ± 347.61, p < 0.001). No significant differences in VFMS were found between Groups 1 and 3 (p > 0.05). Both glaucoma groups (Groups 1 and 3) showed lower VFMS compared to controls (Groups 2 and 4) (p < 0.05), excluding NI and NS regions (p > 0.05).

Table 3 summarizes the correlation analyses between pVD, pRNFL, and VFMS in Group 1 and Group 3. Significant positive correlations were found between average pVD and average VFMS in both groups (r = 0.681 and r = 0.749, respectively, p < 0.001). Similarly, average pRNFL correlated with average VFMS in both groups (r = 0.504 and r = 0.722, respectively, p < 0.001). In both groups, the correlation between average pVD and average VFMS was stronger than that between average pRNFL and average VFMS. as clearly depicted in Figure 5, which presents a scatter plot visualization of these correlations.

Regional analyses revealed significant positive correlations between pVD and VFMS across all regions in both groups (p < 0.05). However, pRNFL and VFMS no significant correlations were observed in NI and NS regions in Group 1 and NI region in Group 3. The strongest correlation between pVD-VFMS and pRNFL-VFMS was in the ST and IT regions for both groups. Notably, the pRNFL-VFMS correlation was stronger than the pVD-VFMS correlation in ST and IT regions (p < 0.001). Figure 6 further complements these findings by illustrating the regional correlation patterns, offering a detailed spatial perspective on the relationships.

Further analysis of pVD, pRNFL, and MD by visual field loss severity is presented in Table 4. In early-stage glaucoma with high myopia, pVD was positively correlated with MD (p < 0.05), while pRNFL had no significant correlation with MD. In moderate-to-advanced glaucoma, both pVD and pRNFL showed significant correlations with MD. In non-highly myopic glaucoma, both pVD and pRNFL were positively correlated with MD in early and moderate-to-advanced stages (p < 0.05).

Linear regression analyses were further conducted between average pVD, pRNFL, and VFMS (Table 5, Figure 7). The results of univariate linear regression analysis indicated that pVD had a significant impact on VFMS in both HMG group (R² = 0.459, p < 0.001) and NHMG group (R² = 0.556, p < 0.001). Similarly, pRNFL also showed a significant effect on VFMS in two groups. In the multivariate linear regression analysis, pVD maintained a significant correlation with VFMS in HMG group (R² = 0.464, p < 0.001) and the NHMG group (R² = 0.566, p = 0.001). However, we observed that pRNFL did not significantly impact VFMS in either group (p = 0.172; p = 0.082, respectively).

Table 6 summarizes the ROC curve analysis assessing the diagnostic capabilities of pVD and pRNFL in HMG and NHMG. Both average and eight regional pVD measurements showed diagnostic significance in both groups, except for the NI region in Group 1. Average pVD had significantly higher diagnostic ability in Group 3 (AUC: 0.945) compared to Group 1 (AUC: 0.822; p < 0.001). In Group 1, the best diagnostic performance for pVD was at the average level, ST and IT region. In Group 3, the most effective diagnostic regions were the average level, ST, IT, and IN regions. Combining the ST and IT regions improved diagnostic accuracy for both groups. Additionally, pVD demonstrated greater diagnostic capability in NHMG than in HMG (p = 0.016).

The ROC analysis also revealed strong diagnostic abilities for average and regional pRNFL measurements in both groups, except for the NI and NS regions in Group 1. Average pRNFL had significantly higher diagnostic efficacy in Group 3 (AUC: 0.948) compared to Group 1 (AUC: 0.881; p = 0.023). In Group 1, the top diagnostic markers for pRNFL were the average level, ST, and IT region. In Group 3, the most robust performance was seen in the average level, ST, IT, and SN region. Combining the ST and IT regions improved diagnostic performance in both groups, with no significant differences between the two groups for this combination (p = 0.750). Figure 8 illustrates these findings.

Table 7 compares the diagnostic performance of pVD and pRNFL for glaucoma in highly myopic and non-highly myopic eyes. In HMG, average pRNFL demonstrated significantly better diagnostic performance (AUC: 0.881) than average pVD (AUC: 0.822, p = 0.026). Similarly, when combining the ST and IT regions, pRNFL showed superior diagnostic ability compared to pVD (AUC: 0.934 vs. 0.905, p = 0.010). In NHMG, no significant differences were observed between the diagnostic capabilities of average pVD and pRNFL (AUC: 0.945 vs. 0.948, p = 0.763) or the combined ST and IT regions (AUC: 0.965 vs. 0.942, p = 0.105).

Evaluate the staging diagnostic abilities of pVD and pRNFL in Table 7. In early-stage HMG, no significant differences were noted between the diagnostic abilities of average pVD and pRNFL or their combined ST and IT regions. For moderate-stage HMG, average diagnostic abilities of pVD and pRNFL were comparable, but pRNFL outperformed pVD in the combined ST and IT regions (p = 0.034). In advanced-stage HMG, pRNFL showed superior average diagnostic performance (p = 0.043), while no significant differences were observed in the combined ST and IT regions.

For NHMG, no significant differences were found in early-, moderate-, or advanced-stage groups between pVD and pRNFL, either at the average level or in the combined ST and IT regions. Across all stages, pVD and pRNFL demonstrated similar diagnostic capabilities in NHMG.

RNFL thinning is a well-established marker of glaucomatous progression, with this study demonstrating progressive reductions from non-high myopic to highly myopic glaucoma. However, structural changes, such as optic disc tilting in high myopia, can alter RNFL distribution, particularly in temporal-superior regions, complicating its use as a reliable biomarker. In contrast, pVD appears to be a more stable and reliable metric in highly myopic patients, as it is less affected by structural changes induced by myopia (19). Previous studies have shown that glaucoma significantly reduces pVD and, with reductions strongly correlating with visual field deterioration, particularly in the TS and TI regions (24). Myopia is also associated with reductions in both pVD and pRNFL (25). In our previous study, pVD reduction was more strongly correlated with high myopia-induced visual function impairment than pRNFL reduction (9). These findings suggest that integrating both structural (pRNFL) and vascular (pVD) metrics may enhance diagnostic precision. Notably, the NI and NS regions showed non-significant differences in pVD and pRNFL between glaucoma and control groups. This may be explained by the fact that nasal sectors are generally less affected in early glaucoma, where damage predominantly occurs in the superior and inferior temporal regions. Furthermore, these nasal regions are more prone to segmentation errors and measurement variability, especially in highly myopic eyes with peripapillary atrophy, potentially reducing the sensitivity of these parameters in those areas (26, 27).

To observe the performance of pVD, pRNFL, and VFMS under varying degrees of myopia in the presence or absence of glaucoma, we conducted ANOVA and post hoc multiple comparisons among the four groups. This allowed us to evaluate whether the damage from glaucoma masked or amplified the damage associated with high myopia. Our analysis confirmed a stepwise reduction in pVD across non-high myopia, high myopia, and myopia with glaucoma (Table 2, Figure 4). High myopia caused diffuse and uniform pVD reduction, whereas glaucoma led to more severe, region-specific pVD damage, particularly in the ST and IT regions. Suwan et al. (25) reported more severe pVD damage in myopic glaucoma compared to non-myopic glaucoma, while our study found no significant differences between these groups except in the NI region. However, their subsequent research demonstrated that the extent of pVD damage was comparable between the groups, which aligns with our findings (28). These discrepancies across studies may be attributed to advancements in research methodology, differences in sample selection, or variability in glaucoma severity among study populations.

We observed a significant decline in VFMS in high myopic eyes compared to non-highly myopic eyes. However, this decline was much less severe than in glaucoma groups. High myopia-induced visual function loss was primarily localized to the nasal regions of the optic disc and visual field (NI, NS, IN, and SN regions), while glaucoma-induced damage predominantly affected the temporal regions (ST, TS, TI, and IT regions). These regional differences provided crucial clues for differentiating high myopia from glaucoma. Notably, no significant differences in VFMS or regional analyses were observed between high myopia with glaucoma and non-high myopia with glaucoma groups. We hypothesize that the damage caused by glaucoma may overshadow myopia-induced damage, masking some differences between the two groups.

Our findings demonstrated a significant correlation between pVD and pRNFL thickness with VFMS in both glaucoma groups. Notably, the relationship between pVD and VFMS was stronger than that between pRNFL and VFMS in highly myopic glaucoma and non-highly myopic glaucoma. Across all regions, pVD exhibited a significant positive correlation with VFMS, whereas pRNFL showed no significant association in the NI and NS sectors. Consequently, pVD emerged as a more robust indicator of VFMS than pRNFL. In the comparison across different stages of glaucoma, it can also be observed that the reduction in pVD is significantly positively correlated with the degree of VF damage, while in the early stage of glaucoma in high myopia, there is no correlation between RNFL and VF damage. Linear regression analysis confirmed that pVD was more consistent than pRNFL in monitoring visual function, particularly in the ST and IT regions, which are critical for glaucoma diagnosis. These results suggest that pVD may be a more reliable marker for tracking glaucomatous visual field damage.

This study evaluated the diagnostic efficacy of pRNFL and pVD across eight peripapillary regions in highly myopic and non-highly myopic glaucoma patients. In highly myopic glaucoma, all regional pRNFL measurements, except for the NI and NS regions, exhibited significant diagnostic capacity, with the ST and IT regions showing the highest performance. For non-highly myopic glaucoma, pRNFL demonstrated strong diagnostic efficiency across all regions, with superior performance in the ST and IT sectors. Our findings revealed that pRNFL exhibited excellent diagnostic performance for glaucoma (AUC: 0.934, 0.942). However, this was slightly lower than the diagnostic ability reported by Fukai et al. (29) (AUC: 0.97), which may be attributed to the inclusion of only myopic glaucoma patients in our study, in contrast to the broader glaucoma population assessed by Fukai et al. (29).

ROC curve analyses of the combined ST and IT regions revealed superior diagnostic accuracy compared to overall or individual regions in both glaucoma groups. Diagnostic evaluation of pVD highlighted its sensitivity for detecting myopic glaucoma. In highly myopic glaucoma, pVD demonstrated strong diagnostic performance in all regions except the NI region, whereas in non-highly myopic glaucoma, pVD performed excellently across all regions, particularly in the ST and IT sectors. Combining the ST and IT regions further enhanced diagnostic efficacy, introducing a novel approach to improving diagnostic accuracy.

Stage-specific analyses revealed that in early and moderate stages of highly myopic glaucoma, pVD and pRNFL exhibited comparable diagnostic abilities. However, in advanced stages, pVD’s diagnostic performance declined significantly compared to pRNFL. For non-highly myopic glaucoma, no significant differences were observed between pVD and pRNFL at any stage, consistent with findings by Yarmohammadi et al. (30) These results suggest that pVD is particularly valuable for early-stage glaucoma diagnosis, but its reliability diminishes in advanced disease stages.