Introduction

The effect of ocular magnification and axial length (AL) on optical coherence tomography (OCT) parameters such as circumpapillary retinal nerve fiber layer thickness (cpRNFLT) in highly myopic (HM) eyes is well known from past literature1,2. The true size of an obtained image is dependent on magnification of the camera and eye in non-telecentric optical systems3. Thickness values are mainly comprised of vertical components and the impact of magnification error on the thickness itself is expected to be minimal. However, the area upon which OCT parameters are measured is enlarged in eyes with longer AL due to magnification error3. Since the distribution of cpRNFLT is not uniform and known to become thinner with the distance from the optic nerve head (ONH)4, this enlargement of measurement circle causes underestimation of cpRNFLT in HM eyes5. However, later studies using magnification corrected measurement areas have found no correlation6 or even slightly positive correlations7 between AL and cpRNFLT, emphasizing the importance of correcting for magnification error on measurement area when interpreting cpRNFLT of myopic eyes.

Circumpapillary capillary density (cpCD) measured from OCT angiography (OCTA) is another promising parameter in detecting early glaucomatous damage8 in HM eyes9. According to a systematic review by Llanas et al., only 8% of the reports using OCTA performed magnification correction on its measurement area10. Although the density nature of cpCD makes it less susceptible to magnification error, Sampson DM et al. reported that magnification error causes a − 20% to + 10% change of vessel density in foveal area and a − 3% to + 2% change in parafoveal area11. A similar impact is expected to occur in the assessment of peripapillary microvasculature due to magnification effect induced measurement area changes. However, little is known about the effect of magnification error on circumpapillary OCTA parameters.

Furthermore, the effect of AL on cpCD has been controversial in past reports. CpCD is expected to be to be less influenced by myopic anatomical changes because large vessels, which shifts temporally with AL elongation, are excluded from analysis9. However, to the best of our knowledge, there has been no report to date that examines the effect of AL and magnification error of cpCD over a wide scan area excluding large vessels, which are the main advantages of assessing HM eyes with OCTA.

We hypothesized that magnification error and AL elongation in HM eyes affect cpCD although its impact is expected to be smaller than that of cpRNFLT. It is important to ascertain these effects by remeasuring cpCD and cpRNFLT using the magnification corrected measurement area instead of simply estimating them from the uncorrected values. Considering that HM is increasing dramatically12 and known as the greatest risk factor of developing open angle glaucoma13, our evaluation is crucial for better understanding and interpreting OCTA results in HM eyes. The purpose of this study is to evaluate the effect of magnification error and AL on global and sectoral cpCD and cpRNFLT in healthy eyes.

Results

One hundred fifty-four eyes of 77 normal subjects met the inclusion criteria. We included 72 eyes of 72 subjects in the current study after excluding 29 eyes of 24 subjects (10 eyes for history of retinal disease, 6 eyes for suspected pathological myopia, 13 eyes for insufficient quality and/or segmentation errors of OCTA scans) and randomly selecting one eye from subjects whose both eyes were eligible. Table 1 presents background data of the study subjects. Intra-visit reproducibility of the angiography scans calculated by intra-class correlation was 0.886 (95% confidence interval was 0.758–0.956) for global cpCD, 0.863 (0.714–0.947) to 0.955 (0.898–0.983) for sectoral cpCD, 0.943 (0.874–0.979) for global cpRNFLT and 0.904 (0.792–0.963) to 0.995 (0.988–0.998) for sectoral cpRNFLT.

Table 1 Demographic and clinical characteristics of subjects.
Full size table

Average uncorrected and corrected values of cpCD, cpRNFLT and percentage difference between corrected and uncorrected values (ΔcpCD, ΔcpRNFLT) are presented in Table 2. ΔcpCD-AL and ΔcpRNFLT-AL relationships corrected for age and signal strength indices (SSI) are shown in Table 3. ΔcpCD-AL association was positively significant in global (p < 0.001) and in the inferior nasal and the superior nasal sectors (both p < 0.001). ΔcpRNFLT-AL association was positively significant in global and all sectors (all p < 0.001). ΔcpRNFLT-AL correlations were significantly stronger than ΔcpCD-AL correlations in global and all sectors (all p < 0.001, Fig. 1).

Table 2 Uncorrected and corrected values of cpCD and cpRNFLT and the difference between values before and after magnification correction.
Full size table
Table 3 Relationship of ΔcpCD and ΔcpRNFLT with axial length.
Full size table
Figure 1
figure 1

Scatter plots of global and Garway-Heath sectoral Δcircumpapillary capillary density (cpCD), Δcircumpapillary retinal nerve fiber layer thickness (cpRNFLT) and axial length (AL). R and p values are the results of Pearson’s correlation. ΔcpRNFLT demonstrated significantly stronger correlation with AL in comparison with ΔcpCD in global and all sectors (all p < 0.001).

Full size image

Relationship of corrected cpCD and cpRNFLT with AL adjusted for age and SSI are shown in Table 4. Corrected cpCD did not significantly associate with AL in global and all sectors, while corrected cpRNFLT showed significant increase with AL in global and temporal sector (p = 0.005 and p < 0.001, respectively). Scatter plots of corrected global cpCD and cpRNFLT with AL are shown in Fig. 2. Scatter plots of corrected sectoral cpCD and cpRNFLT with AL are presented in Supplementary Figs. S1 and S2, respectively.

Table 4 Relationship of corrected cpCD and cpRNFLT with axial length.
Full size table
Figure 2
figure 2

Scatter plots of uncorrected and corrected global circumpapillary vessel density (cpCD), circumpapillary retinal nerve fiber layer thickness (cpRNFLT) and axial length (AL). R and p values are the results of Pearson’s correlation.

Full size image

Discussion

Our results indicate that uncorrected cpCD values were underestimated in eyes with AL longer than the nominal AL (24.46 mm for Cirrus 6000) with sectoral differences in the ΔcpCD-AL relationships, but to a much lesser degree in comparison to cpRNFLT. Furthermore, magnification corrected global cpCD did not decrease with longer AL while magnification corrected cpRNFLT thickened with longer AL especially in the temporal sector.

The significant positive association between global ΔcpCD and AL indicated that an enlargement of measurement area due to magnification error in eyes with longer AL resulted in an underestimation of cpCD. Since cpCD is calculated as a density parameter, an enlargement of measurement area should not result in underestimation of cpCD if the distribution of capillaries is uniform among the circumpapillary area. The fact that cpCD is underestimated as the measurement area shifts to the periphery suggests that distribution of capillaries becomes sparser with distance from the ONH center. Our results are supported by histological studies finding radial peripapillary capillaries to be most prominent around the ONH and less dense towards the macula14. Furthermore, since age15 and SSI16 are also known to affect cpCD, we were able to confirm this capillary distribution even after correcting for these effects. Considering that OCTA has the ability to visualize superficial capillaries in vivo that are not always observable by other methods of ocular vascular evaluation such as histological studies of donor eyes or fluorescein angiography17, our study may have been able to quantitatively capture the distribution gradient of peripapillary capillaries.

A similar positive association between global ΔcpRNFLT and AL was observed, suggesting decay of cpRNFLT with distance from the ONH in agreement with past literature4. However, the effect of AL was significantly smaller on ΔcpCD than that on ΔcpRNFLT in global and all sectors (Fig. 1). For example, if we apply the multiple regression analysis equation obtained from our data on an eye with AL of 28 mm, the expected underestimation of global uncorrected cpCD would be 1.0% while that of cpRNFLT would be 13.6%, showing that using uncorrected cpRNFLT can lead to substantial misevaluation in HM eyes. The advantage of cpCD in magnification error can be mainly attributed to cpCD being calculated as density values in comparison to cpRNFLT being thickness values measured upon a fixed diameter circle. Although this result was as expected considering the density nature of cpCD, we were able to reveal that cpCD is not free from impact of magnification error. However, the effect was much smaller in comparison to cpRNFLT, and its clinical impact is expected to be small. Our current results corroborate the clinical potential of cpCD as a stabler parameter for management of glaucomatous eyes with HM9.

Sectoral analysis revealed that ΔcpCD- AL associations were significant in the superior nasal and inferior nasal sectors but not in the other sectors. This suggests that within the 2.0–4.5 mm annulus area, cpCD decrease with distance from the ONH center is observed mainly in the nasal sectors. This finding is consistent with reports that the decay slope of cpRNFLT is steepest in nasal and inferior quadrants4. Gabriele ML et al. reported that cpRNFLT was related inversely with distance from ONH only in the nasal quadrant while other quadrants showed an initial increase and gradual decrease18. Analogous variation of gradient and distribution among sectors may also exist for cpCD.

Magnification corrected global and sectoral cpCD did not significantly correlate with AL, indicating that cpCD did not decrease in eyes with longer AL even after adjustment for age and SSI. Several previous studies have reported decreased peripapillary vessel density in non-pathological HM eyes19,20,21, attributing this to AL elongation induced thinning of the retina which reduces oxygen demand and vascular endothelial growth factor production22. However, these prior studies either did not correct OCTA measurement area for magnification error or exclude HM eyes with suspected glaucoma. On the other hand, Hirasawa K et al. reported no significant correlation between peripapillary vessel density and AL in both normal and glaucomatous eyes after magnification corrections in accordance with our results23. However, since their analysis was based on a rather narrow measurement area of a 3.2–3.6 mm ONH centered annulus and did not exclude large blood vessels that have been reported to enlarge in caliber with AL elongation24, the detection of AL-related change of microvasculature may not have been fully represented in their data. Our study analyzed cpCD over a wider area of a 2.0–4.5 mm annulus excluding large vessels to confirm that cpCD did not decrease with AL elongation in healthy eyes.

Magnification corrected temporal cpRNFLT demonstrated a positive association with AL in agreement with previous reports25, suggesting that the temporal shift of RNFL bundles accompanied with AL elongation has an effect on the cpRNFLT distribution and that cpRNFLT in HM eyes need to be handled with caution especially when compared with normative databases that do not include HM eyes.

There are some limitations to our study. First, ocular magnification is influenced by several optical properties other than AL, such as corneal curvature and ocular refraction. There are methods of magnification correction taking these other parameters into consideration26. However, the effect of AL on magnification is least prone to error and our method is currently widely used in OCT studies3,27. Secondly, the corrected measurement area extended beyond the obtained OCTA image in eyes with AL shorter than nominal AL leading to data loss on the periphery of the measurement area of these eyes. However, this data loss only occurred in eyes with AL shorter than nominal AL. The percentage of area where the loss occurred in those eyes was 3.6 ± 4.0% of the corrected measurement area of cpCD and the impact on our results is expected to be minimal. Thirdly, our study was conducted only on Japanese subjects. Considering that cpRNFLT is reported to differ among races28, cpCD may present different results in other races.

In conclusion, magnification error causes underestimation of cpCD in eyes with longer AL, especially in the nasal sectors. However, underestimation of cpCD due to magnification effect was significantly smaller in comparison to cpRNFLT. Furthermore, no decrease in age and SSI adjusted cpCD was observed with AL elongation. Our results reveal the effect of magnification error and AL on cpCD, providing essential knowledge in interpreting cpCD of HM eyes.

Methods

Participants

Protocols for this retrospective cross-sectional study were approved by the Research Ethics Committee of the Graduate School of Medicine and Faculty of Medicine at The University of Tokyo (Identifier: 2217) and adhered to the tenets of the Declaration of Helsinki. Patients gave informed consent for their information to be stored in the hospital database and used for retrospective research at their first visit. Study participants were notified of the protocol posted at the outpatient clinic and were provided with the opportunity to opt-out of the study.

Participants of this study included healthy subjects who consulted for a routine eye examination or refractive error from the University of Tokyo Hospital (Tokyo, Japan), Yotsuya Shirato Eye Clinic (Tokyo, Japan) and Tajimi Iwase Eye Clinic (Gifu, Japan) between June 2020 and February 2023.

All participants underwent the following ocular examinations: refraction and corneal curvature radius measurements (TONOREF II, NIDEK, Aichi, Japan), best corrected visual activity (BCVA), AL measurements (OA-2000, TOMEY, Aichi, Japan), central corneal thickness (CCT) measurements (CEM-530, NIDEK, Aichi, Japan), slit-lamp examination, Goldmann applanation tonometry, gonioscopy, fundus examination including ONH examination, optic disc stereophotography (nonmyd WX, Kowa, Tokyo, Japan) and OCT/OCTA imaging (Cirrus HD-6000 with AngioPlex OCTA, Carl Zeiss Meditec, Dublin, CA, USA).

Diagnosis of each eye was conducted independently by two glaucoma specialists (KA, HS) and disagreements were resolved by a third adjudicator (MA). Inclusion criteria for this study were age > 20, BCVA of 20/25 or better, AL < 30 mm, attainment of good quality OCTA scanning with SSI ≥ 9, normal optic disc appearances without glaucomatous ONH changes (i.e. neuro-retinal rim narrowing, notching and the presence of retinal nerve fiber layer defects) on fundus examination and fundus stereo photographs, intraocular pressure (IOP) < 21 mmHg and no abnormal findings on slit-lamp examination and fundus examination.

Eyes with corneal opacity, clinically significant cataract, retinal disease, non-glaucomatous optic neuropathy and a history of glaucoma, corneal, vitreous and refractive surgery were excluded from this study. Eyes with pathological myopia or its suspects (diffuse chorioretinal atrophy, patchy chorioretinal atrophy, lacquer cracks, myopic choroidal neovascularization and macular atrophy) were excluded by checking fundus photographs and macular OCT scans29. When both eyes of one subject were eligible for the study, one eye was randomly selected.

Optical coherence tomography angiography imaging

OCTA imaging was performed using the Cirrus HD-6000 with AngioPlex OCTA (Carl Zeiss Meditec, Dublin, CA, USA) with a scan speed of 100,000 A-scans per second and eye tracking technology. OCTA images were obtained with an angular sweep corresponding to 4.5 mm × 4.5 mm in an eye with nominal AL of 24.46 mm after pupillary dilation.

CpCD was calculated using the Zeiss AngioPlex software (version 11.5, https://www.zeiss.com/meditec/en/products/optical-coherence-tomography-devices/cirrus-6000-performance-oct.html) as the density of the superficial capillary plexus between the inner limiting membrane and RNFL inside a 4.5 mm diameter Bruch’s membrane opening (BMO)-centered circle excluding the 2 mm diameter BMO-centered optic disc area. Radial peripapillary capillary (RPC) plexus slab images were generated by the device and cpCD values were obtained as the total area of perfused vasculature per unit area in the region of measurement. Large vessels of more than 30 μm in width were excluded from the analysis.

CpRNFLT parameters were measured upon a 3.46 mm diameter BMO-centered circle from ONH angiography volume scans and not from separate ONH cube scans to ensure exact alignment of measurement areas between the two parameters.

All OCTA scans were checked by two experienced examiners (KA, HS) and cases with poor quality scans due to (1) vitreous opacity, (2) motion artifacts and blinks, (3) defocusing, (4) segmentation errors (erroneous identification of the borders of RNFL or incomplete segmentation by the automated algorithm and PPA associated artifacts) in measurement areas30, (5) missing data due to excessive disc tilt and (6) displacement of the measurement center were excluded16. Three consecutive ONH angiography scans were taken on the same visit from 15 eyes to assess intra-visit reproducibility.

Magnification correction

Measurement areas for cpCD and measurement circles for cpRNFLT were adjusted for magnification error using the modified Littmann’s formula27. According to this formula, the relationship between true size, t, and the measured size, s, can be expressed as: t = p × q × s, where p is the magnification factor related each imaging system, and q is the magnification factor related to each eye, q = 0.01306 x (AL − 1.82). The factor p can be calculated from the modified Littmann’s formula27 when the nominal AL for the imaging system is known, and Cirrus 6000 assumes a nominal AL of 24.46 mm, p = 1/q = 1/ [0.01306(24.46—1.82)] = 3.382. The RPC slab images and cpRFNLT maps were generated with custom Zeiss software (version 11.5, https://www.zeiss.com/meditec/en/products/optical-coherence-tomography-devices/cirrus-6000-performance-oct.html) and exported from the device. CpCD and cpRNFLT values were calculated from the same images with uncorrected measurement circle/areas and magnification corrected measurement circle/areas using GNU Octave (software version 9.2.0, https://www.gnu.org/software/octave/doc/v9.2.0/) (Fig. 3). Garway-Heath sectoral31 cpCD and cpRNFLT values were also obtained.

Figure 3
figure 3

Illustration of magnification uncorrected and corrected measurement area of circumpapillary capillary density (cpCD) and measurement circle of circumpapillary retinal nerve fiber layer thickness (cpRNFLT) in an eye with 27.76 mm axial length. Global and sectoral cpCD and cpRNFLT values are presented. IN: Inferior nasal; IT: Inferior temporal; SN: Superior nasal; ST: Superior temporal; N: nasal; T: temporal. (a) The red area represents the magnification uncorrected measurement area for cpCD. Global and sectoral uncorrected cpCD are shown in the red square. (b) The yellow area represents the magnification corrected measurement area for cpCD. Global and sectoral corrected cpCD are shown in the yellow square. (c) The red circle represents the magnification uncorrected measurement circle for cpRNFLT. Global and sectoral uncorrected cpRNFLT are shown in the red square. (d) The yellow circle represents the magnification corrected measurement circle for cpRNFLT. Global and sectoral corrected cpRNFLT are shown in the yellow square.

Full size image

ΔcpCD and ΔcpRNFLT were defined as the difference between values before magnification correction (uncorrected cpCD, cpRNFLT) and values after magnification correction (corrected cpCD, cpRNFLT).

$$ \Delta {\text{cpCD}}(\% ) = \frac{{{\text{corrected}}\;{\text{cpCD}}(\% ) - {\text{uncorrected}}\;{\text{cpCD}}(\% )}}{{{\text{uncorrected}}\;{\text{cpCD}}(\% )}} $$
$$ \Delta {\text{cpRNFLT}}(\% ) = \frac{{{\text{corrected}}\;{\text{cpRNFLT}}\;(\upmu {\text{m}}) - {\text{uncorrected}}\;{\text{cpRNFLT}}\;(\upmu {\text{m}})}}{{{\text{uncorrected}}\;{\text{cpRNFLT}}\;(\upmu {\text{m}})}} $$

Statistical analysis

All data are reported as the mean ± standard deviation unless otherwise specified. Intra-visit reproducibility was evaluated by intra-class correlation coefficients. Multiple linear regression analyses were used to explore the effect of AL on ΔcpCD and ΔcpRNFLT (ΔcpCD-AL, ΔcpRNFLT-AL, respectively) as well as on corrected cpCD and cpRNFLT, after adjusting for age and SSI. Bonferroni’s correction for alpha-error were conducted to account for multiple comparisons. Steiger’s test was used to evaluate differences in Pearson’s correlation coefficients between ΔcpCD-AL and ΔcpRNFLT-AL relationships. Statistical analyses were performed with commercially available software (SPSS version 27.0; SPSS, Inc., Chicago, IL, USA).