Choroidal Circulation in 8- to 30-Year-Old Chinese, Measured by SS-OCT/OCTA: Relations to Age, Axial Length, and Choroidal Thickness

Abstract

Purpose

The purpose of this study was to evaluate and explore the determinants of choroidal vascularity and choriocapillaris perfusion in a Chinese population aged 8 to 30 years old.

Methods

Three hundred eighty eyes from 380 subjects aged 8 to 30 years were included in this cross-sectional study. Submacular choroidal thickness (ChT), total choroidal area (TCA), luminal area (LA), stromal area (SA), choroidal vascularity index (CVI), and choriocapillaris flow deficit (CcFD) were estimated using images obtained from optical coherence tomography (OCT).

Results

In this population, the mean ChT was 260.4 ± 63.3 µm, TCA was 1.56 ± 0.38 mm², LA was 0.94 ± 0.25 mm², and SA was 0.62 ± 0.15 mm². The mean CVI was 60.25 ± 3.21% and CcFD was 11.95 ± 1.98%. Multivariable analyses showed that higher CVI and LA was associated with older age, thicker ChT, and shorter AL; and lower CcFD was associated with shorter AL. However, the associations were not uniformly rectilinear between CcFD and age. Specifically, CcFD was positively associated with age in subjects ≤19 years old and negatively associated with age in subjects >19 years old.

Conclusions

Development of the choroidal medium- and large-sized vascular layers and choriocapillaris was different across patients aged 8 to 30 years old. Greater axial length was associated with attenuated choroidal circulation. Choroidal thickness correlated well with choroidal vascularity, but not with choriocapillaris perfusion. Further comprehensive and longitudinal assessment of choroidal vasculature and choriocapillaris perfusion will help greatly to understand the physiological and pathological mechanisms responsible for myopia.

The choroid is a highly vascularized structure containing blood vessels and extravascular stroma. Previous studies have suggested that the choroid is implicated not only in the pathogenesis of many chorioretinal diseases, but also in the regulation of eye growth and myopia development.¹ For this reason, thoroughly characterizing the choroid in human subjects would be of value for understanding the pathophysiology of myopia.

The mammalian choroid is composed of three vascular sublayers: the innermost layer containing choriocapillaris, the middle layer containing medium-sized vessels (Sattler's layer), and the outermost layer containing large vessels (Haller's layer).¹ It mainly supplies nutrients and enables metabolic exchange between the blood and outer retinal layers, which have one of the highest metabolic rates among all tissues in the body. The choroidal circulation is complicated and multifactorial, but the development of optical coherence tomography (OCT) has enhanced the ability to obtain detailed and high-resolution images and to perform qualitative and quantitative analyses of choroidal characteristics in vivo. Structural OCT provides some quantitative information about choroidal thickness (ChT) and choroidal vascular dimensions (e.g. choroidal vascularity index [CVI]),^2–4 whereas OCT angiography (OCTA) measures choroidal blood flow signals, of which flow deficits in the choriocapillaris (choriocapillaris flow deficits [CcFDs]) are of special interest.^5,6 In normal middle-aged and elderly adults, various factors – such as age, axial length (AL), and ChT – have been reported to be associated with quantities of choroidal vascularity and choriocapillaris perfusion.^4,7–9 In children and young adults, however, the major concern is with developmental changes in choroidal properties, as the eye continues to grow rapidly during these ages and is therefore at high risk of developing refractive errors, such as myopia. Given the importance of choroidal circulation, it is critical to characterize in detail the developmental changes in choroidal vascularity and choriocapillaris perfusion in human subjects.

To this end, we conducted a study on a Chinese population of people aged 8 to 30 years old, assessing the relationships of age, AL, and ChT, with choroidal vascularity and choriocapillaris perfusion.

Methods

Study Design

This cross-sectional study was approved by the ethics committee of the Eye Hospital of Wenzhou Medical University. A total of 570 subjects were recruited from the public, optometry outpatients at the Eye Hospital, and students of Wenzhou Medical University from May 2020 to October 2022. All participants were treated in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants and the parents of those <18 years old. Of the 570 recruited subjects, 380 were eligible for this study, by the criteria given below (Fig. 1).

Figure 1.

Flow diagram of participants. A total of 570 subjects were recruited, and 380 of them met the inclusion criteria.

The participants were healthy children and young adults between 8 and 30 years of age, with no history of smoking or significant ocular or systemic diseases. Subjects with good choroidal images in at least one eye were included. Detailed demographic data are provided further below. The exclusion criteria were any one of the following: anisometropia of at least 1.00 diopter (D), ametropia of spherical equivalent refraction (SER) more than +1.00 D or less than −6.00 D of SER, astigmatism of more than 1.50 D, best-corrected Snellen visual acuity of less than 20/25 in each eye, AL greater than 26.5 mm, or intraocular pressure (IOP) greater than 21 mm Hg in each eye. Subjects currently undergoing myopia control treatments, such as orthokeratology or atropine, were excluded. In addition, subjects with use of medicine were also excluded.

Examination Procedures

Medical history was collected, followed by ophthalmic screening examinations (including non-cycloplegic subjective refraction and slit-lamp examination) to assess the eligibility of the participants. IOP was measured by non-contact tonometry (Canon TX-20, Tokyo, Japan). Corneal power (CP) and AL were measured with IOLMaster 700 (Carl Zeiss Meditec AG, Jena, Germany). Body weight, height, and body mass index (BMI) were measured using an Ultrasonic height- and weight-measuring instrument (HGM-300, Henan Shengyuan Industrial Co., LTD, Zhengzhou City, Henan Province, China). Choroidal imaging was performed with swept source optical coherence tomography (SS-OCT) and OCTA, as detailed below. Subjects were free of caffeine intake for at least 24 hours prior to choroidal imaging. All measurements were conducted between 13:30 and 17:00, so as to minimize any effects of circadian variations.¹⁰

SS-OCT and OCTA Imaging and Analysis

The choroidal images were acquired by SS-OCT/OCTA (VG200S; SVision Imaging, Henan, China), provided with a swept-source laser having a central wavelength of approximately 1050 nm and an eye-tracking utility. The designed power level used by VG200S was 3.4 mW, which is much lower than the maximum allowable power of 32 mW and 73 mW for 2-dimensional and 3-dimensional imaging of OCT module, respectively. The scan rate was 200,000 A-scans per second, and scan depth was 3 mm. Axial resolution and lateral resolution were 5 µm and 13 µm, respectively.

Structural OCT of the submacular region was performed with 18 radial scan lines centered on the fovea. Each scan line, generated by 2048 A-scans, was nominally 12-mm long and separated from adjacent lines by 10 degrees. Sixty-four B-scans were obtained on each scan line and automatically averaged to improve the signal-to-noise ratio.¹¹ A representative scan along the horizontal meridian was used to analyze choroidal thickness and vascularity. The images were segmented semiautomatically and binarized (Fig. 2) with Niblack's autolocal threshold, using custom-designed algorithms in MATLAB R2017a (MathWorks, Natick, MA, USA) as previously described.^12–14 After image processing, the mean ChT, total choroidal area (TCA), luminal area (LA), and stromal area (SA) in a 6-mm diameter submacular region centered on the fovea were calculated. The CVI was defined as the ratio of LA to TCA.

Figure 2.

Choroidal image binarization and illustration of choroidal vasculature analysis. (A) Original OCT B-scan in horizontal meridian. (B) Binarized image of choroidal area. (C) Overlays of binarized choroidal areas on original images. The 6-mm wide submacular area between the yellow vertical lines, centered on the fovea, was used for analysis.

OCTA fundus images were obtained with a raster scan protocol of 512 horizontal B-scans that covered an area of nominally 6 mm × 6 mm centered on the fovea. The B-scans, which contained 512 A-scans each, were repeated 4 times and averaged. The choriocapillaris layer was defined as a slab from the basal border of the retinal pigment epithelium-Bruch's membrane complex to 20 µm behind it (i.e. into the choroid).^6,15 To isolate the areas with absence of flow from choriocapillaris, a global thresholding method that utilizes the standard deviation values of a normal database (SD method) was applied to each choriocapillaris image.⁶ The percentage of CcFD was calculated in a 5-mm diameter circular region centered on the fovea (Fig. 3). The location of the fovea was determined manually by examining the B-scans vertically and horizontally (Figs. 3C1-C2).

Figure 3.

Illustration of choriocapillaris blood perfusion analysis. (A) OCTA scan region of nominally 6 mm × 6 mm, centered on the fovea. (B) Magnified en face OCTA choriocapillaris image within a 5-mm diameter circle, which was centered on the fovea through examining (C1) horizontal and (C2) vertical meridian scans across the fovea, and was segmented as a 20-µm thickness slab below Bruch's membrane (BM).

The scales of both OCT and OCTA images were adjusted for the differences in magnification due to differences in AL among the eyes.¹⁶ The choroidal measurements exhibited good repeatibility and reproducibility, as previously reported.^12,17

Statistical Analysis

The statistical analyses were performed primarily using SPSS Statistics 27.0 (IBM, Armonk, NY, USA) and R version 4.2.3 with RStudio (Posit Software; PBC, Boston, MA, USA). Because of the symmetry between the two eyes in the same individual, only the right eyes were included in the analysis, unless poor OCT/OCTA images were obtained due to fixation issues or an OCT signal strength <8, in which case, the left eyes were chosen for analysis instead (n = 57). The distribution of continuous numerical data was assessed by the Kolmogorov-Smirnov normality test. The means and standard deviations (SDs), as well as medians and ranges, of all continuous variables were calculated. The coefficients of variation for choroidal metrics were also calculated by dividing their SDs by their means. The categorical variable (i.e. gender) was expressed as number and proportion.

To understand the relationships of independent variables (age, AL, and ChT) with dependent variables (CVI, LA, and CcFD), restrictive cubic spline (RCS) regression with 3 specified knots at the 10th, 50th, and 90th centiles was used to estimate the linear/nonlinear trend with the rms R package created by Frank (https://hbiostat.org/R/rms). These knots, concomitantly, included two splines β₁ and β₂ with their respective estimates, β₁ quantified whether there was a linearity, β₂ quantified whether there was a deviation from linearity, testing β₂ = 0 was equally to test whether the assumption of linearity was statistically rejected or not.¹⁸ In terms of linear relationships, the general linear regression model was applied, whereas piecewise regression was applied for nonlinear relationships. The piecewise-regression Python package created by Pilgrim was used to identify the breakpoints of the fitted curves with curvilinear trends where the transition of linear fitted slope happened.¹⁶ The package was constructed following the general form of Muggeo,¹⁹ fitted a linear regression model to data that included one or more breakpoints.²⁰ Then, generalized estimating equations were used for multivariable analysis to explore the associations further. For associations of ChT with the aforementioned choroidal metrics, covariates such as age, gender, BMI, CP, AL, and IOP were adjusted in the multivariable model. For associations of age and AL with choroidal metrics, ChT was excluded from the regression models because there are intrinsic physiological links between ChT and these choroidal metrics (i.e. potential role of ChT as mediator rather than confounder).^21,22 In conditions having an age-defined breakpoint, as mentioned above, stratified analysis was performed and interaction effects of age group with age, AL, and ChT were tested. A value of P < 0.05 was considered statistically significant.

Results

Demographics, Ocular Biometrics, and Choroidal Characteristics

A total of 380 subjects were eligible for this study, the mean age was 16.6 ± 5.7 years, with 47.6% of male subjects, and the mean BMI was 20.67 ± 3.49 kg/m² (Table 1). Of the 380 eyes, the SER, AL, and IOP were −2.33 ± 1.80 D, 24.62 ± 0.95 mm, and 15.0 ± 2.6 mm Hg, respectively.

Table 1.

Demographics and Ocular Biometrics of the Population

	Total (n = 380)
	Mean ± SD	Median	Range
Gender, boy or male (%)	181 (47.6%)	–	–
Age, y	16.6 ± 5.7	17.0	8.0 ∼ 30.0
Weight, kg	52.3 ± 14.5	51.7	19.6 ∼ 105.0
Height, cm	157.5 ± 13.4	159.0	117.5 ∼ 186.0
BMI, kg/m2	20.67 ± 3.49	20.35	13.17 ∼ 33.14
SER, D	−2.33 ± 1.80	−2.25	−5.87 ∼ +1.00
CP, D	43.18 ± 1.36	43.10	39.65 ∼ 47.68
AL, mm	24.62 ± 0.95	24.64	21.58 ∼ 26.47
IOP, mm Hg	15.0 ± 2.6	15.0	7.5 ∼ 21.0

BMI, body mass index; SER, spherical equivalent refraction; CP, corneal power; AL, axial length; IOP, intraocular pressure; SD, standard deviation.

The mean submacular ChT was 260.4 ± 63.3 µm, TCA was 1.56 ± 0.38 mm², LA was 0.94 ± 0.25 mm², and SA was 0.62 ± 0.15 mm² (Table 2). The CVI was 60.25 ± 3.21%. The mean CcFD was 11.95 ± 1.98% in the submacular region. The coefficients of variation for ChT, TCA, LA, and SA ranged from 23.63% to 26.05%. The coefficient of variation was 16.54% for CcFD, and 5.33% for CVI, respectively. The distribution of choroidal metrics is shown in Supplementary Figure S1.

Table 2.

Submacular Choroidal Metrics in the Total Study Population

	Total (n = 380)
	Mean ± SD	Median	Range	Coefficient of Variation (%)
ChT (µm)	260.4 ± 63.3	254.3	127.0 ∼ 517.8	24.29%
TCA (mm2)	1.56 ± 0.38	1.53	0.76 ∼ 3.11	24.29%
LA (mm2)	0.94 ± 0.25	0.92	0.43 ∼ 1.94	26.05%
SA (mm2)	0.62 ± 0.15	0.60	0.30 ∼ 1.20	23.63%
CVI (%)	60.25 ± 3.21	60.39	48.91 ∼ 69.48	5.33%
CcFD (%)	11.95 ± 1.98	11.71	7.21 ∼ 20.14	16.54%

ChT, choroidal thickness; TCA, total choroidal area; LA, luminal area; SA, stromal area; CVI, choroidal vascularity index; CcFD, choriocapillaris flow deficits; SD, standard deviation.

Factors Associated With CVI, LA, and CcFD

RSC were fitted to examine the relationships of CVI with age, AL, and ChT (Figs. 4A1-A3). The fitted functions were approximately rectilinear for these associations; therefore, multivariable regression analyses were carried out using data for the entire population (aged 8 to 30 years). CVI was positively associated with age (β = 0.07%/y, 95% confidence interval [CI] = 0.00 to 0.13, P < 0.05) and ChT (β = 0.10%/10 µm, 95% CI = 0.04 to 0.15, P < 0.001), but negatively associated with AL (β = −0.74%/mm, 95% CI = −1.17 to −0.31, P < 0.001; Table 3). Similar analyses were carried out for LA and CcFD below.

Figure 4.

Scatterplots and restrictive cubic splines (RCS) for choroidal metrics. Upper row: Relationships of CVI with age (A1), AL (A2), and ChT (A3). Middle row: Relationships of LA with age (B1), AL (B2), and ChT (B3). Lower row: Relationships of CcFD with age (C1), AL (C2), and ChT (C3). The black curves and grey area represented predicted value and 95% confidence interval (95% CI). The β₁ coefficient of linear trend and β₂ coefficient of nonlinear trend was calculated by restricted cubic spline. AL, axial length; ChT, choroidal thickness; CVI, choroidal vascularity index; LA, luminal area; CcFD, choriocapillaris flow deficits.

Table 3.

Multivariable Analyses for CVI (%)

	Total (n = 380)
	β (95% CI)
Age (y)a	0.07 (0.00, 0.13)*
AL (mm)a	−0.74 (−1.17, −0.31)‡
ChT (10 µm)b	0.10 (0.04, 0.15)‡

^aModels were adjusted for age, gender, BMI, CP, AL, and IOP, except the variable itself.

^bModel was adjusted for age, gender, BMI, CP, AL, and IOP.

^* P < 0.05; ^‡ P < 0.001.

The RCS indicated a nonlinear relationship between LA and age, whereas other correlations were approximately rectilinear (Figs. 4B1-B3). Accordingly, piecewise linear regression analyses were performed for the association between LA and age. The results showed that the breakpoint occurred at 15 years where LA was negatively associated with age in participants ≤15 years old (P < 0.05), whereas it was positively associated in those >15 years old (P < 0.01). Further multivariable analyses showed that the effects of age, AL, and ChT on LA in the ≤15 years and >15 years groups were not significantly different (age group × age, P-interaction = 0.058; age group × AL, P-interaction = 0.301; and age group × ChT, P-interaction = 0.695; see Table 4). Therefore, additional analyses were performed for the effect of age, AL, and ChT on LA across the entire population; these showed that LA was negatively associated with AL (β = −0.14 mm²/mm, 95% CI = −0.17 to −0.11, P < 0.001) and positively associated with age (β = 0.01 mm²/y, 95% CI = 0.01 to 0.02, P < 0.001) and ChT (β = 0.04 mm²/10 µm, 95% CI = 0.04 to 0.04, P < 0.001; see Table 4).

Table 4.

Multivariable Analyses for LA (mm²)

	Total (n = 380)	Age ≤15 (n = 180)	Age >15 (n = 200)
	β (95% CI)	β (95% CI)	β (95% CI)	P-Interactionc
Age (y)a	0.01 (0.01, 0.02)‡	0.00 (−0.02, 0.01)	0.02 (0.01, 0.03)†	0.058
AL (mm)a	−0.14 (−0.17, −0.11)‡	−0.10 (−0.14, −0.05)‡	−0.15 (−0.20, −0.10)‡	0.301
ChT (10 µm)b	0.04 (0.04, 0.04)‡	0.04 (0.04, 0.04)‡	0.04 (0.04, 0.04)‡	0.695

^aModels were adjusted for age, gender, BMI, CP, AL, and IOP, except the variable itself.

^bModels were adjusted for age, gender, BMI, CP, AL, and IOP.

^cP-interaction indicated the interaction effects of age group with age, AL, or ChT on LA.

^† P < 0.01;

^‡ P < 0.001.

The RCS indicated that the associations between CcFD and age and between CcFD and ChT were nonlinear, but they were almost rectilinear between CcFD and AL (see Figs. 4C1-C3). Piecewise linear regression analysis showed a breakpoint at age 19 years for the association between CcFD and age. CcFD was found to be positively correlated with age in subjects ≤19 years old (P < 0.001), but negatively correlated with age in subjects >19 years old (P < 0.01). In addition, piecewise regression for the association between CcFD and ChT showed that ChT breakpoint was 290.3 µm. When the ChT was below 290.3 µm, CcFD was negatively associated with ChT (P < 0.01), but the association was not significant when ChT >290.3 µm (P = 0.379). Further multivariable regression analyses showed that the effects of age and AL on CcFD were significantly different in both age groups (age group × age, P-interaction = 0.001 and age group × AL, P-interaction = 0.015); specifically, CcFD was positively associated with age in subjects ≤19 years old (β = 0.12%/y, 95% CI = 0.04 to 0.20, P < 0.01) and negatively associated with age in subjects >19 years old (β = −0.16%/y, 95% CI = −0.31 to −0.01, P < 0.05). In addition, CcFD was positively associated with AL in both age groups, with a greater effect of AL on CcFD in the older age group (subjects ≤19 years old: β = 0.52%/mm, 95% CI = 0.24 to 0.80, P < 0.001; subjects >19 years old: β = 1.05%/mm, 95% CI = 0.67 to 1.42, P < 0.001; Table 5). In the subpopulations with ChT below 290.3 µm, the effects of ChT on CcFD in these 2 age groups were not significantly different (age group × ChT, P-interaction = 0.594); therefore, additional analyses were performed for the effect of ChT on CcFD across the entire age group. There was a negative, although insignificant, association between CcFD and ChT (β = −0.05%/10 µm, 95% CI = −0.11 to 0.01, P = 0.087; see Table 5).

Table 5.

Multivariable Analyses for CcFD (%)

	Total (n = 380)	Age ≤19 (n = 236)	Age >19 (n = 144)
	β (95% CI)	β (95% CI)	β (95% CI)	P-Interactionc
Age (y)a	0.01 (−0.02, 0.05)	0.12 (0.04, 0.20)†	−0.16 (−0.31, −0.01)*	0.001
AL (mm)a	0.84 (0.62, 1.06)‡	0.52 (0.24, 0.80)‡	1.05 (0.67, 1.42)‡	0.015
ChT (10 µm)b	−0.05 (−0.11, 0.01)	−0.03 (−0.10, 0.03)	−0.07 (−0.19, 0.05)	0.594

^aModels were adjusted for age, gender, BMI, CP, AL, and IOP, except the variable itself.

^bModels were adjusted for age, gender, BMI, CP, AL, and IOP. The datasets here included only subjects with ChT ≤290.3 µm. The sample size for this subpopulation is 272: 189 of age ≤19, and 83 of age >19 years.

^cP-interaction indicates the interaction effects of age group with age, AL, or ChT on CcFD.

^* P < 0.05;

^† P < 0.01;

^‡ P < 0.001.

Discussion

In this study, we used SS-OCT/OCTA to measure choroidal vascularity and choriocapillaris perfusion in a Chinese population of 8 to 30 years old. In the entire population, higher CVI and LA was associated with older age, thicker ChT, and shorter AL; and lower CcFD was associated with shorter AL. However, the associations between CcFD and age were not uniformly rectilinear; specifically, CcFD was positively associated with age in subjects ≤19 years and negatively associated with age in subjects >19 years.

The quantitative evaluation of choroidal circulation in vivo is difficult, because the choroidal structure and vasculature are complex. Sonoda et al.³ first reported a binarization method – to differentiate between the choroidal luminal area and the stromal area, and to quantify them. Agrawal et al.⁴ further developed this methodology and proposed a quantitative metric, CVI, to assess the structure and functional state of the choroidal vasculature. CVI, which is the ratio of vascular luminal area to total choroidal area, thus represents the relative intravascular volume in the choroid. Although CVI has been investigated in various conditions to differentiate normal from diseased states,^23,24 this is not applicable in all conditions. For example, in vivo indocyanine green angiography demonstrated delayed filling of blood vessels and loss or narrowing of choroidal vessels and capillaries in highly myopic eyes,^25,26 whereas CVI was increased.²⁷ In addition, in a condition of dramatic decrease in systemic circulatory volume after hemodialysis, the ChT and perfused vessel density in choriocapillaris decreased, but CVI remained stable.^28,29 These phenomena could be attributed to the nature of CVI estimation, which reflects a component ratio of intravascular volume and perivascular tissue content in choroid; in consequence, increased CVI could imply either greater intravascular volume or less perivascular tissue in choroid, and vice versa for reduced CVI. Therefore, the absolute vascular metric, LA, is required to extend the concept of choroidal vascularity, because it expresses the total cross-sectional intravascular area in the region of interest, which represents the intravascular volume. Although these choroidal metrics from structural OCT images mainly reflect the vascular structure in Haller's and Sattler's layers, an integration of choriocapillaris flow deficits from OCTA would comprehensively delineate the global state of the choroidal circulation.

As ChT has been extensively studied in various conditions,³⁰ it is important to establish its relationship with the new metrics (CVI, LA, and CcFD). Agrawal et al. and He et al. found that CVI was positively correlated with ChT in normal middle-aged and elderly populations.^4,31 A recent study on interocular asymmetry in anisomyopic adults showed that the interocular differences in CVI, LA, and ChT were positively correlated.¹² Our current study showed that ChT is consistently correlated with CVI and LA in healthy children and young adults, in that greater choroidal thickness accompanied higher values for choroidal vascular components, in both proportion and absolute amount. Data on the relationship between ChT and choriocapillaris perfusion in normal populations are still scarce. Borrelli et al. reported that CcFD decreased with initial increases in LA and SA, followed by an increase with progressive increment of LA and SA characterized by a U-shaped curve fitted by a quadratic function in healthy adults.³² The results of our analyses of these associations were consistent with the above studies (Supplementary Figs. S2–S4), as was the effect of increasing ChT on choriocapillaris perfusion. Based on the piecewise linear regression, if ChT was less than the breakpoint (e.g. 290.3 µm), which are close to the mean ChT levels of cohorts of similar age reported in previous studies,^33–37 increasing choroidal thickness was accompanied by lower flow deficits (i.e. higher perfusion); when ChT was greater than this breakpoint value, however, this was not observed. In other words, a thinner choroid was associated with lower choriocapillaris perfusion when ChT was within a certain (lower) range. It has been reported that hyperopes have greater choroidal thickness than emmetropes and myopes, with a strong negative correlation between ChT and CVI^38,39; thus, ChT and CcFD might be positively associated in hyperopes. Multivariable analyses suggested that ChT and CcFD are affected simultaneously by factors such as age and AL within the lower range, whereas their physiologic link warrants further validation. These observations show that choroidal thickness alone may not be optimal or even sufficient as an indicator of choroidal circulation.

Age is another critical factor that is associated with choroidal changes. Several studies have shown thinning of the choroid with aging, in adults up to ages in their 80s,^4,40 accompanied by loss of choroidal vascular and stromal components.^3,7 CcFD also increases with aging in adults.^8,9 Although such findings suggest age-related attenuation of choroidal circulation in middle-aged and elderly life, the proportion of vascular volume in the choroid (i.e. CVI) has not been consistently shown to vary with age. For example, He et al. reported CVI was negatively correlated with age in middle-aged and elderly adults,³¹ whereas Zhou et al. reported a lack of association of CVI with age in healthy adults by univariable analyses,⁷ and a study on Singapore's healthy adult population showed that the significant associations of CVI with age and AL disappeared when ChT was included as one of the independent variables.⁴ Because ChT was reported to be associated with age and AL in various studies,^4,37,40,41 these inconsistences could be due to a role of ChT as mediator in the associations of CVI with age and AL. Considering this possibility, we constructed two regression models to analyze roles of these independent variables, separately with each one of the choroidal metrics. We found that CVI increased at a rate of 0.07% per year during childhood and early adulthood, and LA increased at a rate of 0.01% per year. In addition, changes in CcFD were positively associated with age in subjects ≤19 years old, whereas negatively associated with age in an older age group. These results indicated that the development of choroidal vasculature was distinct between medium- and large-sized vascular layer and choriocapillaris across 8 to 30 years old. Interestingly, myopia develops mainly during childhood, and its progression stops around the age of 16 years^42,43 – that is to say, the developmental changes of choroidal circulation occur in parallel with myopia development. Whether this link could explain the mechanism underlying myopia progression and stabilization, however, needs further study.

An increase in AL, a critical factor in myopia development, was also associated with choroidal thinning – characterized by reduced choroidal vascularity and choriocapillaris perfusion – in emmetropes and low-to-moderate myopes.^3,12,17,31 Choroidal vascularity and choriocapillaris perfusion further deteriorated in high myopes with extreme ocular elongation,^27,44,45 – despite the increase in CVI, which was attributed to greater loss of stromal volume than of intravascular volume.²⁷ These findings suggested that greater AL was a risk factor of attenuated choroidal circulation.

We acknowledge several limitations of the present study. Although we aimed to establish normal reference values for choroidal metrics, the application of different algorithms from different laboratories and manufacturers, as well as diurnal fluctuation, would likely produce variability in absolute values of these metrics in different studies. What's more, the exclusion criterion in the current study was relatively stringent aiming to characterize the developmental changes in choroid from childhood to young adults. Therefore, our results could not be extrapolated to hyperope, high myope, or older populations. Nevertheless, the associations between these variables may still reflect the intrinsic links or underlying mechanisms. In addition, choroidal vasculature was analyzed only with horizontal scans, whereas a volumetric analysis would have provided additional information on choroidal vasculature. Finally, given the cross-sectional nature of this study, the implications of these results for causality and underlying mechanisms must be interpreted with caution, and further longitudinal studies will be required for complete characterization of these associations.

Conclusions

To conclude, in a Chinese population aged 8 to 30 years, the development of medium- and large-sized choroidal blood vessels and choriocapillaris was found to differ during childhood and early adulthood. Both the proportions and absolute contents of choroidal vascular components were positively associated with choroidal thickness, but the association between choriocapillaris perfusion and choroidal thickness was not significant. Finally, greater AL was associated with attenuation of choroidal circulation. Future studies, involving comprehensive assessment of the choroidal vasculature and choriocapillaris perfusion with quantitative metrics, will greatly facilitate our understanding of the physiological and pathological mechanisms responsible for ocular homeostasis and abnormal conditions such as myopia.