Investigation of choroidal vascular alterations in eyes with myopia using ultrawidefield optical coherence tomography angiography

Introduction

High myopia (HM) is a major threat to public eye health. Epidemiological surveys in 2020 showed that there were about 193 million patients with HM worldwide,1 and it is estimated to reach 938 million by 2050.2 The prevalence of HM is higher in Asian populations, such as in China, Singapore and Japan,3 and it is showing an upward trend.4 In adulthood, HM can cause a series of ocular complications, including glaucoma, cataracts, macular degeneration and retinal detachment,5 leading to progressive loss of visual quality and, in severe cases, blindness, which imposes a heavy economic burden on individuals and society.

The pathological mechanism underlying HM remains poorly understood but is widely believed to be a combination of environmental and genetic factors.6 The eye is an oxygen-demanding organ, and maintaining the integrity and function of ocular structures relies heavily on a sufficient blood supply. Recently, several studies have revealed that hypoxia may play an important role in the development of HM and proposed a novel hypothesis.7–10 These studies postulate that long-term near-distance use of the eye may cause insufficient choroidal blood perfusion; subsequent hypoxia leads to scleral remodelling and causes elongation of the axial length (AL), and the change in scleral tensile strength causes morphological and haemodynamic changes in the vascular network of the fundus, which causes further progression of myopia, and they have found hypoxia evidence based on a cellular study conducted in a wet laboratory11 12; however, there is still insufficient clinical evidence. Oxygen supply to the retina is primarily provided by the retinal and choroidal vessels; the retinal vessels have been extensively reported to change in myopia.10 13–16 In contrast, because of masking by the retinal pigment epithelium layer, observation of choroidal vessels requires invasive examination using indocyanine green angiography, which can only be used on a few patients and is not suitable for large population screening. With the advancement of optical coherence tomography angiography (OCTA) technology, non-invasive quantitative and in-depth analyses of choroidal blood flow are possible, which may provide clinical evidence for the hypoxia hypothesis.

Several studies have analysed eyes with myopia on 3×3 mm or up to 12×12 mm OCTA scans. The results showed that as AL increased, the perfusion area and vascular density in the macular region significantly decreased.17–22 Similar changes were observed in the peripapillary capillaries and deep parafoveal regions of the retina.23 However, restricted by the scanning area of the machine, information regarding the peripheral area is insufficient. With the introduction of ultrawidefield swept-source OCTA (SS-OCTA), which offers a faster scanning speed and clearer 24×20 mm images, two studies have reported a negative correlation between choroidal vascular density (ChVD) and thickness (ChVT) and AL in most peripheral grids of the fundus,24 and choroidal thinning was most evident in the macular region.25 Although these two studies provided some peripheral information on eyes with myopia, they had some prominent limitations. First, though both studies used the latest SS-OCTA, the built-in software algorithms were not the latest versions, resulting in inaccurate measurements. Additionally, the sample size was small, which did not represent all myopia groups and may have affected the statistical significance of the results. Therefore, this study aimed to compare the influence of myopia on ChVD and ChVT in a cohort study with a larger sample size using SS-OCTA with an updated version of the built-in software, which may provide a more accurate measurement of choroidal blood vessel parameters. This approach contributes to a broader understanding of the role of choroidal vascular changes in myopia progression.

Materials and methods

Image analysis

Images of the choroidal vessel layer were generated automatically using built-in software (figure 1A). In some cases, the error in automatic segmentation was manually corrected for the entire scan volume. We chose 3×3 grids (comprising nine rectangles: temposuperior, superior, nasal-superior, tempo, macular, optic disc, tempo-inferior, inferior and nasal-inferior) with a total area of 17 mm × 17 mm to analyse the parameters of the choroid (figure 1B). To correct the eye movement-incurred imaging errors, all images were manually evaluated to confirm proper placement of the 3×3 grids, with the macular fovea placed centrally (figure 1B). The choroidal density (ChVT figure 1C) and choroidal thickness (ChVD figure 1D) in the nine grids were obtained using the built-in software algorithm. The ChVT extends from 29 µm below the Bruch’s membrane to the choroid–sclera interface, and the ChVD is calculated as Choroidal Vessel volume Index (CVI), which is defined as the ratio of choroid vessel volume to choroid volume within a designated three-dimensional region, equivalent to a 3D vessel density. In the unit of percentage, the larger the value, the denser the vessels; CVI is a default format of choroid vessel density. These vessel metrics were corrected by inputting the actual AL values into the software to adjust for ocular magnifications.

Request permission

Figure 1

(A) Representative B-scan showing choroidal vascular thickness (ChVT). The green line shows the boundaries of the choroidal thickness extending from 29 µm below the Bruch’s membrane to the choroid–sclera interface. (B) Optical coherence tomography angiography (OCTA) en-face image showing 3×3 grids (comprising nine grids: tempo-superior, superior, nasal-superior, tempo, macular, optic disc, tempo-inferior, inferior and nasal-inferior) centred on the fovea with a total area of 17×17 mm. The grids were chosen to analyse the parameters of the choroid. (C–D) Representative OCTA en-face image showing choroidal vascular density (%; image C) and ChVT using a heat map (μM; image D).

Results

Correlation between ChVT and ChVD and AL

Pearson’s correlation analysis indicated that the ChVD and a thinner ChVT were negatively correlated with AL in all grids (p=0.000, p≤0.001) (online supplemental table 1, figures 2 and 3), respectively. Regression lines showed choroidal thinning with increasing AL, which was most evident in the macular grid (β = −34.20, p=0.000); the grids were ranked in decreasing order of regression coefficients as follows: macular, superior, nasal-superior, optic disc, inferior, tempo, tempo-inferior, tempo-superior and nasal-inferior (figure 2). The decrease in ChVD with increasing AL was most evident in the optic disc grid (β = −2.19, p=0.000); the grids were ranked in decreasing order of regression coefficients as follows: optic disc, macular, tempo-inferior, nasal-inferior, nasal-superior, inferior, tempo, superior and tempo-superior (figure 3).

Request permission

Figure 2

Scatterplots of choroidal vascular thickness and axial length (AL) in 724 eyes with myopia. Pearson’s correlation indicated that choroidal thinning was correlated with longer AL in all grids (p=0.000). Regression lines showed choroidal thinning with increasing AL, which was most evident in the macular grid (β = −34.20, p=0.000). Varied warm colour levels represent the regression coefficient values.

Request permission

Figure 3

Scatterplots of choroidal vascular density (ChVD) and axial length (AL) in 724 eyes with myopia. ChVD was negatively correlated with AL in all grids (p≤0.001), and the decrease in ChVD with increasing AL was most evident in the optic disc grid (β = −2.19, p=0.000). Varied warm colour levels represent the regression coefficient values.

Discussion

Myopia progression is accompanied by haemodynamic changes in the central and peripheral regions of the retina and choroid. The emergence of OCTA has made it possible to conveniently explore changes in the blood vessels of the fundus in myopia cases. Most on-duty OCTA machines can only capture 6×6 mm or up to 12×12 mm images. Herein, a newly developed SS-OCTA system was used. Using vertical-cavity surface-emitting laser with 100-nm bandwidth, a set of large-diameter lenses, 128-Hz eye tracking rate, and 15-s average acquisition time in the ultrawidefield OCTA made it possible to obtain a 24×20 mm rectangular scan while maintaining a lateral resolution of 10 µm, thereby enabling the rapid screening of a large population of eyes with myopia. In addition, the 1060-nm light source empowered a 24×20 mm imaging range, which was particularly suitable for observing the vessels of the fundus in eyes with myopia with abnormally prolonged AL. In the present study, eyes with a lens power of −17.00 D and AL of 32.31 mm could be imaged clearly. Our research was based on a large cohort of 3000 people, mostly composed of highly educated computer programmers with good compliance. The results of ultrawidefield SS-OCTA showed high reproducibility in assessing ChVT and ChVD metrics, and all examinations were performed by a well-trained photographer.

Many studies have shown a significant correlation between AL and ChVT. The main aim of this study was to identify spatiotemporal choroidal vascular changes in eyes with myopia using ultrawidefield SS-OCTA based on a larger cohort. We observed that there were non-negligible individual differences in the choroidal vascular structure, such as those related to age and sex (data not provided), which may affect the results. Therefore, a larger sample size may reduce the impact of individual differences on the results. In this study, we enrolled 362 participants (724 eyes) aged between 21 and 54 years, which was three–five times the average sample size of previous similar studies. Additionally, we included a group of 88 eyes (SER=−11.13 ± 2.48, AL=28.05 ± 1.20) with SM, which has not been reported in other similar studies; the corresponding data under extreme conditions grant better understanding of the long-term effects of myopia on retinal blood vessels. Furthermore, we updated the built-in analysis software to the latest version (V.1.3.3.1); compared with the old version, the newly upgraded software has made significant improvements in identifying and measuring vascular structures with an improved algorithm. Our results showed that ChVT and ChVD were negatively correlated with AL in all nine grids (p≤0.001), which to the best of our knowledge was not reported previously.

In the ChVT group analysis, significant choroidal thinning appeared in two grids of the LM group, indicating that myopia affected ChVT in the early stage, with the number increasing quickly to four grids in the MM group and eight–nine grids in the HM and SM groups, indicating that myopia has a profound and rapid impact on ChVT. Consistent with previous studies, we determined that choroidal thinning was most evident in the macular grid; this is possibly related to the CHVT being thickest in the macular grid and thus most sensitive to the pressure caused by the elongation of the eyeball. But unlike most previous studies, our results showed that myopia also caused obvious choroidal vascular thinning in the optic disc grid in early stage; only one study by Harb et al supports this finding,26 and this may be related to the fact that posterior staphyloma often occurs in this area.27

ChVD has also been reported to be significantly reduced in eyes with myopia.24 25 In our ChVD group analysis, no significant change in ChVD was observed in the LM group; however, with myopia progression, four–five of nine grids in the MM and HM groups and seven grids in the SM group showed a significant decrease in ChVD, indicating that myopia affected ChVD in the later stage. We believe that the most likely reason is that blood vessels have better elasticity and adaptability. To express this hypothesis, we created a hypothetical model and used mathematical calculations to explain (online supplemental figure 1): in the early stages of myopia, as the AL elongates, the sclera exerts slight pressure on the choroid, causing it to thin. At this stage, CHVT decreases, although the vessels become slightly flattened, the CHVD does not change (online supplemental figure 1B). As myopia progresses and the AL becomes excessively long, the blood vessels are not only flattened but also stretched, becoming thinner. At this stage, both CHVT and CHVD decrease (online supplemental figure1C). Therefore, through this hypothetical model, we can easily conclude that the flattening of blood vessels has minimal impact on CHVD; only when blood vessels become thinner does it affect CHVD.

Unlike all previous studies, the most evident decrease in ChVD appeared in the optic disc grid, followed by the macular grid. As we mentioned above, CHVD is primarily affected by stretch forces, which lead to the thinning of blood vessels. We speculate that the stretch force is greatest in the optic disc area, with the appearance of myopic crescents in the optic disc area being the best evidence of this. Combined with our ChVT analysis, our results indicate that myopia may have a significant impact on the blood supply of the optic disc, which has not been reported in previous studies. The clinical significance of this finding is unclear, and whether it is related to the myopia-induced thinning of the optic disc nerve fibre layer remains unknown, with further research needed to confirm this theory.

The macular region requires the highest amount of oxygen. Our results indicated that myopia has the most evident effect on ChVT, followed by ChVD in the macular grid, indicating that myopia has a significant impact on the blood supply to the macular area and may cause hypoxia, which may be the precipitating factor of neovascularisation in eyes with HM. Determining to what extent the decrease in ChVT and ChVD would induce choroidal neovascularisation (CNV) and establishing new risk factors for the prediction and prevention of CNV is crucial.

This study had some limitations. First, compared with the normal group, the proportion of men gradually increased as the degree of myopia increased. This may be related to the fact that our research population comprised mainly male computer programmers, and the noticeable differences in the sex ratio in the MM, HM and SM groups may have impacted the results. Second, although examinations using SS-OCTA have shown excellent replicability in other studies with large-scale population screening, our examinations were performed by one well-trained examiner, which may limit the replicability of our data.