OCT Angiography of Chorioretinal Microvasculature in Children with Type 1 and Type 2 Diabetes without Retinopathy

Abstract

Objective

To evaluate chorioretinal microvasculature using OCT angiography (OCTA) in children with type 1 diabetes mellitus (T1DM) and type 2 diabetes mellitus (T2DM) without diabetic retinopathy (DR), compared with healthy controls.

Design

A cross-sectional, observational study.

Participants

Subjects with T1DM and T2DM without DR were recruited from the pediatric diabetes center, and nondiabetic controls were recruited from the pediatric eye clinic from 2022 to 2024.

Methods

All participants underwent color fundus and 1050 nm swept-source OCTA imaging (3 × 3, 6 × 6, and 12 × 12 mm²). Quantitative OCTA analyses were performed using a custom MATLAB algorithm. Retinal segmentation was performed for the superficial retinal layer, the deep retinal layer, the whole retina, and the choroid. A linear mixed effects regression model was used to compare the results between youth with diabetes and controls, adjusting for age.

Main Outcome Measures

Mean values for retinal thickness, vessel skeleton density (VSD), vessel diameter index (VDI), mean choroidal thickness (MCT), choroidal vascularity index (CVI), choroidal vascular volume (CVV), and choriocapillaris thickness.

Results

A total of 32 subjects (44% females, 57 eyes) were included as follows: 10 T1DM, 7 T2DM, and 15 without diabetes. For youth with T1DM, age was 16 ± 2 years, glycated hemoglobin was 7.7 ± 0.9%, and diabetes mellitus (DM) duration was 8.0 ± 4 years. For youth with T2DM, age was 16 ± 2 years, mean glycated hemoglobin was 8.7 ± 3.5%, and mean DM duration was 2.4 ± 1.3 years. Age for youth without diabetes was 12 ± 5 years. Choroidal vascularity index and CVV were significantly higher in T1DM compared with controls (CVI: 0.69 ± 0.04 vs. 0.63 ± 0.05, P = 0.001; CVV: 18.9 ± 5.0 vs. 16.2 ± 2.0 mm³, P = 0.001) but not in T2DM. Choriocapillaris thickness was also significantly higher in T1DM (9.5 ± 1.2 μm) compared with controls (8.4 ± 1.0 μm, P = 0.003), with no significant difference in T2DM. Retinal thickness, MCT, VSD, VDI, and flux were not different among groups (P > 0.05).

Conclusions

Children with T1DM without DR exhibited larger choroidal and choriocapillaris vascularity than controls and no contemporaneous differences in retinal vascularity measures. This suggests that subclinical choroidal changes are present in pediatric diabetic patients before clinical or subclinical signs of DR.

Financial Disclosure(s)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Diabetic retinopathy (DR) is a serious and progressive complication of diabetes mellitus (DM) that can lead to vision loss. According to the SEARCH (The SEARCH for Diabetes in Youth) study, DR is present in 52% of youth with type 1 diabetes mellitus (T1DM) and 56% with type 2 diabetes mellitus (T2DM) within 12 years of diagnosis.¹ Although visual impairment in DR is primarily attributed to retinal pathology, there is evidence from histologic,² angiographic,3, 4, 5 and laser Doppler flowmetry⁶ studies in adults that shows diabetic choroidopathy concurrent with retinopathy and even preceding DR. Until recently, in vivo evaluation of the choroid has been limited to indocyanine green angiography, laser Doppler flowmetry, or spectral-domain OCT angiography (OCTA). Indocyanine green angiography and laser Doppler flowmetry are not indicated in the clinical assessment of DR at all and are not clinically available in most settings. Spectral-domain OCTA has poor signal penetration into the choroid and is suboptimal for reproducible measurements or analyses of subretinal pigment epithelium structures. With the advent of US Food and Drug Administration-approved widefield and swept-source OCT angiography (SS-OCTA), noninvasive in vivo evaluation of the choroid and choriocapillaris is now clinically feasible. Rapid and noninvasive imaging in children is of particularly high value given the difficulty of performing any invasive testing in this population and the importance of early disease detection. OCT angiography has the additional benefit of being depth-resolved and amenable to quantitation, unlike dye-based angiography studies.

Importantly, SS-OCTA measures can uniquely illustrate contemporaneous pathology in both retinal and choroidal vasculature as well as surrounding nonvascular structure (e.g., choroidal stroma), thereby allowing a comprehensive understanding of the temporal relationship between pathology in these regions. Quantitative assessment of the retinal and choroidal microcirculation has improved our understanding of preclinical vascular changes in various retinal diseases including DR.7, 8, 9 Studies on adults with T1DM have shown reduced vessel density or other measures of perfusion, particularly in the deep capillary plexus.10, 11, 12 Similar findings in T2DM suggest that reductions in vessel density may precede visible microvascular damage.^8,13, 14, 15, 16, 17 In pediatric T1DM, studies using OCTA have shown inconsistent results, with some studies reporting reduced capillary perfusion in patients with DM without DR,18, 19, 20, 21, 22, 23, 24 while others have found no significant differences.^25,26 Some authors suggest earlier perfusion changes in the disc than in the macula.²⁷ In general, there appears to be an inverse association between retinal capillary perfusion and DM duration or severity in youth,^20,25,27 suggesting there is an impact of DM on retinal parameters in youth similar to adults. Early changes in the choroid of T1DM patients have also been suggested in very few studies.^28,29

While there is limited data on preclinical OCTA changes in youth with T1DM, there are no studies describing such changes in youth with T2DM. There are also no studies that contemporaneously assess retinal and choroidal changes in these populations. Previous studies demonstrating alarming rates of retinopathy after a short duration of diabetes in youth with T2DM underscore the importance of early detection of DR and methods for screening.^1,30 The aim of this study was to characterize contemporaneous changes in retinal and choroidal vascular metrics using OCTA in children with T1DM or T2DM without clinically detectable DR and compare the results to those obtained from youth without DM.

Methods

Participants

This was a cross-sectional study performed from 2022 to 2024 in youth with diabetes being prospectively followed at the Johns Hopkins Pediatric Diabetes Center and youth without diabetes from the Wilmer Eye Institute. The study was approved by the institutional review board of the Johns Hopkins Hospital. Written informed consent was obtained from all participants aged ≥18 years or from the parents of those <18 years of age. The protocol was conducted according to the principles in the Declaration of Helsinki and in compliance with the Health Insurance Portability and Accountability Act of 1996.

Participants with DM were youth aged 8 to 21 years, recruited from the ACCESS 2 (AI for Children's diabetiC Eye ExamS) study (NCT05463289). Eligibility criteria included (1) a diagnosis of T1DM or T2DM based on the American Diabetes Association diagnostic criteria, (2) meeting the American Diabetes Association guidelines for DR screening (a diagnosis of T1DM for ≥3 years, at age ≥11 or in puberty, or a diagnosis of T2DM), (2) not having received a diabetic eye examination within the last year, and (3) no known DR.³¹ Participants with DM were matched for comorbidities, if any were present. Participants with no history of diabetes or other vascular disease were recruited from the Pediatric Ophthalmology Clinic of the Wilmer Eye Institute. Participants in either cohort were excluded if they had a best-corrected visual acuity ≤20/60, developmental delay, inability to consent, poor cooperation, psychiatric comorbidities, myopia >6.00 diopters, or a history of any of the following: glaucoma or ocular hypertension, ocular media opacities, intraocular surgery or laser treatment, extraocular muscle surgery within 6 months, systemic vascular diseases (e.g., hypertension), other forms of diabetes, psychiatric comorbidities, vasculitis, or any visually threatening ocular disease.

The following data were obtained from the electronic medical records: glycated hemoglobin at the study visit, duration of disease (defined as the time from the diagnosis of DM to the time of the study), and any other potentially relevant past medical history. All participants completed a study questionnaire with self-reported information regarding medical history, caffeine consumption, and medication history. This information was verified and supplemented by information from available medical records.

Image Acquisition

All participants were imaged at the Wilmer Eye Institute. Imaging data were acquired using a commercially available swept-source OCT and OCTA instrument (PLEX Elite 9000, Carl Zeiss Meditec) with an A-scan speed of 200 kHz, a center wavelength of 1060 nm, and a motion tracking mechanism. All subjects' scans contained both OCT and OCTA data simultaneously. The device provides an axial resolution of 6 μm and a lateral resolution of 20 μm and uses a previously described^32,33 phase- and intensity-based algorithm (optical microangiography) to differentiate static tissue from moving particles. Duplicate scans were acquired in a 3 × 3, 6 × 6, and 12 × 12 mm² pattern centered on the fovea in all patients. Images that met the following quality control criteria were included: signal strength ≥7 (as assessed by the manufacturer's commercially available software), centered on the fovea or decentered by less than half of the foveal diameter, minimal motion artifacts without vessel displacement, and floaters or media opacities occupying <5% of the image.

Segmentation

OCT angiography images were automatically segmented using the device manufacturer's software (PLEX Elite version 2.1.1.107), which defines the superficial retinal layer as the slab between the internal limiting membrane and the posterior boundary of the inner plexiform layer and the deep retinal layer as the slab between the posterior boundaries of the inner plexiform layer and the outer plexiform layer. The choriocapillaris was defined using custom segmentation software as a 16-micron-thick layer with its anterior boundary located 4 microns below the Bruch membrane.³⁴

The choroidal slab was automatically segmented via a previously described and validated algorithm by Zhou et al.³⁵ The optic discs in the en face projections of the entire volume of the 12 × 12 mm² scans were automatically detected and were excluded as previously described.³⁵ The color-coded enface choroidal thickness maps were then generated representing the distance between the detected boundaries of Bruch membrane and the choroid–scleral interface. All choroidal segmentation results were manually reviewed to ensure accuracy. Any segmentation that was not accurate was either manually corrected or excluded from analysis. Examples of the segmentation schemes used for the retinal and choroidal analyses are illustrated in Figure S1 (available at www.ophthalmologyscience.org).

Image Analysis

A previously validated, semiautomated custom MATLAB (MATLAB R2023a; The MathWorks, Inc) algorithm was used to calculate the OCTA measures, including vessel skeleton density (VSD), vessel diameter index (VDI), and flux.^7,36,37 The foveal avascular zone (FAZ) area was included in the primary analysis, and the large vessels were excluded. Secondary analyses excluding the FAZ were also performed. Briefly, VSD is a measure of overall perfused vessel length within an OCTA image and is computed by reducing all vessel segments in a binarized image to 1-pixel width and then calculating the ratio of the total length of vessels to the total area of the image. Vessel diameter index reflects the average vessel diameter and is calculated by dividing the nonskeletonized vessel density area by VSD. Flux is a measure that is proportional to the red blood cell content of capillaries and is calculated by averaging the absolute (nonbinarized) decorrelation intensity values from regions of the image occupied by vascular structures and is described in extensive detail in Abdolahi et al.³⁶ All retinal OCTA measures are calculated and presented in arbitrary units and previously described in detail.^7,36

Mean choroidal thickness (MCT) is the average choroidal thickness over the entire scanning region, excluding the optic disc.³⁵ The optic disc was manually defined using a set of landmark points placed along the visible disc margin on the en face projection image. A larger MCT value indicates thicker choroidal vascular or stromal tissue but does not differentiate between the 2 types of tissue. In each volume scan, choroidal vascular volume (CVV) is calculated as the product of the number of pixels in the segmented vessels in the choroidal slab, $N_{vessel}$, and the volume of a single pixel, $v$ (i.e., $N_{vessel}\ast v$). Choroidal vessel volume represents only the vascular component of the choroid. Choroidal vascularity index (CVI) is calculated as the ratio of CVV to the total volume of the segmented choroid slab, $V_{total}$ (i.e., $\frac{CVV}{V_{total}}$) or the ratio of the luminal area to total choroidal area. Accordingly, the choroidal stromal volume can be derived as the complement of the CVV (1-CVV), and a corresponding choroidal stromal index (CSI) may be represented as CSI = 1 – CVI.^38,39 A larger CVI (or smaller CSI) value indicates a higher density of choroidal vasculature in comparison to stromal tissue. A smaller CVI (or larger CSI) value indicates relatively more stromal tissue than choroidal vascular tissue. By convention, only CVI values will be reported. The corresponding CSI values are reported in Table S1 (available at www.ophthalmologyscience.org). The choriocapillaris thickness (CCT) measurements were obtained using a peak detection algorithm as previously described.⁴⁰

Due to the differences in pixel size with different scan patterns and differences in the regional anatomy of the retinal and choroidal vessels, retinal measures were derived from foveal-centered 3 × 3 mm² OCTA scans, choriocapillaris measures were derived from foveal-centered 6 × 6 mm² OCTA scans, and choroidal measures were derived from foveal-centered 12 × 12 mm² scans. Secondary analyses with retinal measures derived from 6 × 6 mm² OCTA scans were also performed.

Retinal thickness measures were derived from the OCT data that were simultaneously acquired with OCTA. Full retinal thickness was defined as the distance from the internal limiting membrane to the retinal pigment epithelium, determined by the device manufacturer's software. For the 3 × 3 and 6 × 6 mm² scans, the automatically generated thickness values from the device were averaged to obtain mean retinal thickness measurements for analysis. Retinal thickness values from the 12 × 12 mm² scans were not available for this study.

Statistical Analysis

All statistical analyses were performed using SPSS Statistics (IBM, version 29.02.0). Categorical variables were analyzed using the Fisher exact test and continuous variables were analyzed using a 1-way analysis of variance or independent t test. One-way analysis of variance was used to compare unadjusted retinal and choroidal measures across the 3 groups. Data were then analyzed using multivariate, linear mixed-effects regression models (LMM) to investigate whether quantitative OCTA measurements differ between T1DM and T2DM subjects and those without diabetes. In each model, fixed effects included the category of subjects (T1DM, T2DM, and controls), with age as a covariate. Subject ID was included as a random effect to account for the correlation between measurements from the same individual. No statistically significant effects of gender or eye laterality on the OCTA measures were observed (P > 0.05), so these factors were not included as covariates. As there were small numbers of varying ethnicities in each group, we could not adjust our analyses for race or ethnicity.

Controls were set as the reference category in the main model. For each respective measure: $OCTAvascularmetric={\beta }_0+{\beta }_1\left(age\right)+{\beta }_2\left(category\right)+Randomeffects\left(SubjectID\right)$, where ${\beta }_0$ represents the intercept, ${\beta }_1$ and ${\beta }_2$ are fixed effects estimates, and category represents T1DM, T2DM, or controls.

In all analyses, a P value of <0.05 was considered statistically significant unless specified otherwise. All data are presented as mean ± standard deviation, and all outputs of the LMM are presented as mean ± standard error, with 95% confidence intervals (CIs), unless stated otherwise. The final values for the different retinal and choroidal measures are presented as weighted means to account for the fact that not all subjects had data available from both eyes.

Results

Subject Demographics

As shown in Table 1, a total of 32 participants (57 eyes) were enrolled in the study (age 14 ± 4 years, range 6–19 years, 44% female, 44% non-Hispanic White), including 10 individuals with T1DM (31%), 7 with T2DM (22%), and 15 without diabetes (47%). The mean ages of the T1DM, T2DM, and control groups were 16 ± 2, 16 ± 2, and 12±5 years, respectively (P = 0.015). The mean duration of diabetes in youth with T1DM was longer compared with those with T2DM (8.0 ± 4.0 vs. 2.4 ± 1.3 years, P = 0.003). In youth with T1DM and T2DM, the mean glycated hemoglobin was 7.7 ± 0.9% and 8.7 ± 3.5%, respectively (P = 0.36). Right eyes comprised 51% of the eyes imaged. Participants had no significant comorbidities or known complications of DM.

Table 1.

Characteristics of Study Participants

Characteristic	Total	T1DM	T2DM	Controls	P Value∗
Number of subjects, n (%)	32	10 (31)	7 (22)	15 (47)	0.22
Age, years (mean ± SD)	14 ± 4	16 ± 2	16 ± 2	12 ± 5	0.015†
Age, range	6–19	13–18	14–18	6–19
Sex at birth, n (%)
Males	18 (56)	4 (40)	5 (71)	9 (60)	0.43
Females	14 (44)	6 (60)	2 (29)	6 (40)
Race, n (%)
White	14 (44)	9 (90)	1 (14)	4 (27)	0.004†
Black	6 (19)	1 (10)	2 (29)	3 (20)
Asian and other	12 (38)	0	4 (57)	8 (53)
Ethnicity, n (%)
Hispanic	2 (6)	0	2 (29)	0	0.042†
Non-Hispanic	29 (94)	10 (100)	5 (71)	15 (100)
Duration of DM, years (mean ± SD)	-	8.0 ± 4.0	2.4 ± 1.3		0.003†
HbA1c at last visit, % (mean ± SD)	-	7.7 ± 0.9	8.7 ± 3.5	-	0.36
Number of eyes, n (%)	57	20 (35)	14 (25)	23 (40)
Right eye (OD)	29 (51)	10 (50)	7 (50)	12 (52)	1.00
Left eye (OS)	28 (49)	10 (50)	7 (50)	11 (48)
Number of scans, n (%)
Retina (3 × 3 mm2)	57	20 (35)	14 (25)	23 (40)
Choriocapillaris (6 × 6 mm2)	52	20 (38)	12 (23)	20 (38)	N/A
Choroid (12 × 12 mm2)	48	18 (38)	12 (25)	18 (38)

DM = diabetes mellitus; HbA1c = glycated hemoglobin; N/A = not applicable; SD = standard deviation; T1DM = type 1 diabetes mellitus; T2DM = type 2 diabetes mellitus.

^∗Categorical variables were compared using the Fisher exact test. Continuous variables were compared using an independent t test or analysis of variance (ANOVA).

^†P values in bold denote statistical significance.

Choroidal and Choriocapillaris Measures

Figure 1 illustrates representative choroidal vascular maps from participants with T1DM, T2DM, and controls without diabetes. The corresponding B-scans showing the segmentation used to define the choroid in the representative maps are illustrated in Figure S2 (available at www.ophthalmologyscience.org). These qualitatively demonstrate the differences in the choroidal vascular measures across the 3 groups. Namely, choroidal vascular measures were generally higher in T1DM than in controls or T2DM. These results are described in detail here and summarized in Figure 2. A detailed overview is provided in Table S1 (available at www.ophthalmologyscience.org). The results of the LMMs are summarized in Table 2.

Figure 1.

Representative choroidal vascular maps qualitatively illustrating choroidal vascular metrics from the right eye of a 17-year-old without DM (A–D), a 16-year-old with type 1 DM (E-H), and a 16-year-old type 2 DM subject (I–L). A, Choroidal vascular map from a 17-year-old nondiabetic subject, showing a CVV of 16.1 mm³, with corresponding pseudocolored CVI heat map (CVI = 0.60) in (B), MCT heat map (MCT = 297 μm) in (C), and CCT map (CCT = 7.8 μm) in (D). E, Choroidal vascular map from a 16-year-old type 1 DM subject, showing a CVV of 27.4 mm³, with corresponding pseudocolored CVI heat map (CVI = 0.69) in (F), MCT heat map (MCT = 442 μm) in (G), and CCT map (CCT = 9.9 μm) in (H). (I) Choroidal vascular map from a 16-year-old type 2 DM subject, showing a CVV of 12.5 mm³, with corresponding pseudocolored CVI heat map (CVI = 0.63) in (J), MCT heat map (MCT = 220 μm) in (K), and CCT map (CCT = 8.7 μm) in (L). Colors higher in saturation represent higher values. CC = choriocapillaris; CCT = choriocapillaris thickness; CVI = choroidal vascular index; CVV = choroidal vascular volume; DM = diabetes mellitus; MCT = mean choroidal thickness.

Figure 2.

Summary of retinal and choroidal vascular OCTA measures in controls, subjects with type 1 diabetes, and subjects with type 2 diabetes. Footnote: Depicted P values were calculated using a 1-way ANOVA to compare the weighted means among controls, T1DM subjects, and T2DM subjects. An asterisk (∗) denotes statistically significant univariate differences between groups (P < 0.05). A double asterisk (∗∗) denotes an individual significant difference observed between specific groups after Bonferroni correction (P < 0.05). (A), Mean choroidal thickness. (B), Mean CVI. (C), Mean CVV. D, Mean CC thickness. (E), Mean VSD in the full thickness retina in the 3 × 3 mm² scans, including the FAZ. F, Mean VSD in the full thickness retina in the 3 × 3 mm² scans, excluding the FAZ. ANOVA = analysis of variance; AUs = arbitrary units; CC = choriocapillaris; CVI = choroidal vascularity index; CVV = choroidal vascular volume; FAZ = foveal avascular zone; MCT = mean choroidal thickness; OCTA = OCT angiography; T1DM = type 1 diabetes mellitus; T2DM = type 2 diabetes mellitus; VSD = vessel skeleton density. Error bars represent positive standard deviation.

Table 2.

Linear Mixed Effects Regression Model for Choroidal Vascular Measures in Subjects with Type 1 and Type 2 Diabetes Compared with Controls^∗

Parameter	Category	Estimate ± SE (95% CI)	P Value†
MCT (μm)	T1DM	27 ± 26 (–27, 82)	0.31
	T2DM	–16 ± 30 (–78, 47)	0.61
CVI (AUs)	T1DM	0.07 ± 0.02 (0.03, 0.11)	0.001
	T2DM	–0.01 ± 0.02 (–0.05, 0.04)	0.77
CVV (mm3)	T1DM	3.5 ± 1.6 (0.2, 6.8)	0.038
	T2DM	–1.2 ± 1.8 (–5.0, 2.6)	0.52
CC thickness (μm)	T1DM	1.6 ± 0.5 (0.6, 2.5)	0.003
	T2DM	–0.3 ± 0.6 (–1.4, 0.9)	0.66

AUs = arbitrary units; CC = choriocapillaris; CI = confidence interval; CVI = choroidal vascularity index; CVV = choroidal vascular volume; MCT = mean choroidal thickness; SE = standard error; T1DM = type 1 diabetes mellitus; T2DM = type 2 diabetes mellitus.

^∗Age was included as a covariate in the model. All estimates represent the adjusted mean difference in each vascular measure in comparison to the control group, which was used as a reference category by the model.

^†P values in bold denote statistical significance.

MCT

Mean MCT for controls, T1DM, and T2DM subjects were 284 ± 37 μm, 303 ± 79 μm, and 250 ± 37 μm, respectively (P = 0.17). Linear mixed-effects regression model analysis showed no significant difference for T1DM (β = 27 ± 26, 95% CI: –27, 82; P = 0.31) or T2DM (β = –16 ± 30, 95% CI: –78, 47; P = 0.61) compared with controls.

CVV

Mean CVV for controls, T1DM, and T2DM subjects were 16.2 ± 2.0 mm³, 18.9 ± 5.0 mm³, and 13.5 ± 2.1 mm³, respectively (P = 0.013). Linear mixed-effects regression model analysis showed a significantly larger CVV for T1DM (β = 3.5 ± 1.6, 95% CI: 0.2, 6.8; P = 0.038), but not for those with T2DM (β = –1.2 ± 1.8, 95% CI: –5.0, 2.6; P = 0.52) compared with controls.

CVI

Mean CVI for controls, T1DM, and T2DM subjects were 0.63 ± 0.05 arbitrary units (AUs), 0.69 ± 0.04 AUs, and 0.61 ± 0.02 AUs, respectively (P = 0.002). Linear mixed-effects regression model analysis showed a significantly larger CVI for T1DM (β = 0.07 ± 0.02, 95% CI: 0.03, 0.11; P = 0.001), but not for those with T2DM (β = –0.01 ± 0.02, 95% CI: –0.05, 0.04; P = 0.77) compared with controls. Mean CSI values for controls, T1DM, and T2DM subjects showed similar but opposite trends (Table S1).

CCT

Mean CCT for controls, T1DM, and T2DM subjects were 8.4 ± 1.0 μm, 9.5 ± 1.2 μm, and 7.6 ± 1.1 μm, respectively (P = 0.004). Linear mixed-effects regression model analysis showed a significantly larger CCT for T1DM (β = 1.6 ± 0.5, 95% CI: 0.6, 2.5; P = 0.003), but not for those with T2DM (β = –0.3 ± 0.6, 95% CI: –1.4, 0.9; P = 0.66) compared with controls.

T1DM Compared with T2DM

When the choroidal and choriocapillaris measures were compared only between T1DM and T2DM, the above findings persisted. Namely, CVV, CVI, and CCT were all significantly higher in T1DM compared with T2DM.

Retinal OCTA Measures

Figure 3 illustrates representative OCTA scans from a healthy control, a T1DM subject, and a T2DM subject. These scans are provided to qualitatively demonstrate the differences in VSD measures across the 3 groups. For 3 × 3 mm² scans, overall retinal vascular measures were trending higher in T1DM subjects compared with T2DM subjects and controls, but these differences did not reach statistical significance. A detailed overview of these results is provided in Tables S1 and S2 (available at www.ophthalmologyscience.org). Similar results were observed when analyzing data for the superficial retinal layer and the deep retinal layer and when analyzing the 6 × 6 mm² scans (Tables S3 and S4, available at www.ophthalmologyscience.org). Overall VSD among the 3 groups is summarized in Figure 2. The main results of the LMM are summarized in Table 3.

Figure 3.

Representative 3 × 3 mm² OCTA scans qualitatively illustrating VSD from the right eye of a 17-year-old without DM (A–C), a 16-year-old with type 1 DM (D–F), and a 16-year-old type 2 DM subject (G–I). A, En face OCTA scan of the full-thickness retina, with representative skeletonized vessel (B) and pseudocolored VSD maps (C) from a 17-year-old subject without DM, with a VSD of 0.134. (D) En face OCTA scan of the full-thickness retina, with representative skeletonized vessel (E) and pseudocolored VSD maps (F) from a 16-year-old subject with type 1 DM, with a VSD of 0.147. G, En face OCTA scan of the full-thickness retina, with representative skeletonized vessel (H) and pseudocolored VSD maps (I) from a subject with type 2 DM, with a VSD of 0.143. Colors higher in saturation represent higher values. DM = diabetes mellitus; OCTA = OCT angiography; VSD = vessel skeleton density.

Table 3.

Linear Mixed Effects Regression Model for the 3 × 3 mm² Full-Thickness Retinal Vascular Measures in Subjects with Type 1 and Type 2 Diabetes Compared with Controls^∗

Parameter	Category	Estimate ± SE (95% CI)	P Value†
FAZ included
VSDtot (AUs)	T1DM	0.006 ± 0.004 (–0.001, 0.014)	0.10
	T2DM	0.000 ± 0.004 (–0.008, 0.008)	0.97
VDItot (AUs)	T1DM	–0.06 ± 0.05 (–0.15, 0.04)	0.21
	T2DM	0.02 ± 0.05 (–0.09, 0.12)	0.75
Fluxtot (AUs)	T1DM	0.010 ± 0.006 (–0.002, 0.022)	0.10
	T2DM	–0.001 ± 0.007 (–0.014, 0.013)	0.92
FAZ excluded
VSDtot (AUs)	T1DM	0.005 ± 0.003 (–0.002, 0.012)	0.16
	T2DM	0.000 ± 0.004 (–0.008, 0.008)	0.98
VDItot (AUs)	T1DM	–0.07 ± 0.05 (–0.18, 0.04)	0.19
	T2DM	0.02 ± 0.06 (–0.10, 0.14)	0.73
Fluxtot (AUs)	T1DM	0.008 ± 0.006 (–0.004, 0.020)	0.17
	T2DM	–0.001 ± 0.006 (–0.014, 0.012)	0.91

AUs = arbitrary units; CI = confidence interval; FAZ = foveal avascular zone; SE = standard error; T1DM = type 1 diabetes mellitus; T2DM = type 2 diabetes mellitus; tot = total; VDI = vessel diameter index; VSD = vessel skeleton density.

^∗Age was included as a covariate in the model. All estimates represent the adjusted mean difference in each vascular measure in comparison to the control group, which was used as a reference category by the model.

^†Significance set at P < 0.05.

Retinal VSD

Mean VSD values for controls, T1DM, and T2DM were 0.135 ± 0.008, 0.140 ± 0.007, and 0.133 ± 0.009, respectively (P = 0.20). Linear mixed-effects regression model analysis showed no significant difference for T1DM (β = 0.006 ± 0.004, 95% CI: –0.001, 0.014; P = 0.10) or T2DM (β = 0.000 ± 0.004, 95% CI: –0.008, 0.008; P = 0.97), when compared with controls. In an analysis where the FAZ area was excluded, this finding was unchanged. In this analysis, mean VSD values for controls, T1DM, and T2DM were 0.141 ± 0.008, 0.144 ± 0.006, and 0.139 ± 0.009, respectively (P = 0.32). Linear mixed-effects regression model analysis showed no significant difference for T1DM (β = 0.005 ± 0.003, 95% CI: –0.002, 0.012; P = 0.16) or T2DM (β = 0.000 ± 0.004, 95% CI: –0.008, 0.008; P = 0.98), when compared with controls. These findings were also unchanged when assessing only the superficial or deep capillary layers separately.

Retinal VDI

Mean VDI values for controls, T1DM, and T2DM were 3.0 ± 0.1, 2.9 ± 0.1, and 3.0 ± 0.1, respectively (P = 0.25). Linear mixed-effects regression model analysis showed no significant difference for T1DM (β = –0.06 ± 0.05, 95% CI: –0.15, 0.04; P = 0.21) or T2DM (β = 0.02 ± 0.05, 95% CI: –0.09, 0.12; P = 0.75) when compared with controls. In an analysis where the FAZ area was excluded, this finding was unchanged. Mean VDI values for controls, T1DM, and T2DM were 3.1 ± 0.1, 3.0 ± 0.1, and 3.1 ± 0.2, respectively (P = 0.20). Linear mixed-effects regression model analysis showed no significant difference for T1DM (β = –0.07 ± 0.05, 95% CI: –0.18, 0.04; P = 0.19) or T2DM (β = 0.02 ± 0.06, 95% CI: –0.10, 0.14; P = 0.73), when compared with controls. These findings were also unchanged when assessing only the superficial or deep capillary layers separately.

Retinal Capillary Flux

Mean flux values for controls, T1DM, and T2DM were 0.138 ± 0.013, 0.147 ± 0.014, and 0.136 ± 0.011 AUs, respectively (P = 0.156). Linear mixed-effects regression model analysis showed no significant difference for T1DM (β = 0.010 ± 0.006, 95% CI: –0.002, 0.022; P = 0.10) or T2DM (β = –0.001 ± 0.007, 95% CI: –0.014, 0.013; P = 0.92), when compared with controls. In an analysis where the FAZ area was excluded, this finding was unchanged. Mean flux values for controls, T1DM, and T2DM were 0.145 ± 0.013, 0.153 ± 0.013, and 0.144 ± 0.011, respectively (P = 0.27). Linear mixed-effects regression model analysis showed no significant difference for T1DM (β = 0.008 ± 0.006, 95% CI: –0.004, 0.02; P = 0.17) or T2DM (β = –0.001 ± 0.006, 95% CI: –0.014, 0.012; P = 0.91), when compared with controls. These findings were unchanged when assessing only the superficial or deep capillary layers separately.

T1DM Compared with T2DM

When the retinal OCTA measures were compared only between T1DM and T2DM, the aforementioned findings persisted. Namely, there was no difference between any of the measures when comparing subjects with T1DM and T2DM.

Retinal Thickness Measurement

Overall retinal thickness (internal limiting membrane to retinal pigment epithelium) measurements were trending higher in T1DM subjects compared with T2DM subjects and controls, but these differences did not reach statistical significance. For 3 × 3 mm² scans, the mean retinal thickness was 325 ± 14, 330 ± 15, and 316 ± 13 μm (P = 0.12) for controls, T1DM, and T2DM, respectively. Linear mixed-effects regression model analysis showed no significant difference for T1DM (β = 5 ± 6, 95% CI: –8, 18, P = 0.42) or T2DM (β = –10 ± 7, 95% CI: –24, 5, P = 0.19) when compared with controls. For 6 × 6 mm² scans, the mean retinal thickness was 301 ± 10, 299 ± 12, and 290 ± 8 μm (P = 0.09) for controls, T1DM, and T2DM, respectively. Linear mixed-effects regression model analysis showed no significant difference for T1DM (β = –0.9 ± 5, 95% CI: –11, 8, P = 0.85) or T2DM (β = –10 ± 5, 95% CI: –20, 1, P = 0.08) when compared with controls. There was also no statistically significant difference in retinal thickness when comparing subjects with T1DM and T2DM.

Discussion

In this cross-sectional study, we used a US Food and Drug Administration-approved SS-OCTA device to investigate the retina and choroid in a pediatric population. Our study is novel in the following ways: (1) contemporaneous assessment of retinal and choroidal vascular measures in a pediatric population using widefield SS-OCTA, (2) application of volumetric choroidal measures rather than 2-dimensional measurements that are historically performed on a limited set of B-scans in the volume, and (3) comparison of diabetic subjects (type 1 and type 2) without retinopathy and controls in a single study. In comparison to control subjects, we show that subjects with T1DM have a more vascular choroid as demonstrated by a significantly increased CVV (a measure of absolute vascular volume in the choroid) and CCT (a measure of choriocapillaris thickness). Notably, these findings occur in the absence of any detectable changes in the retinal vasculature. Our findings support histopathologic findings of choroidal remodeling and inflammation in T1DM² and suggest the timing of such changes precedes the more widely recognized retinal changes. We also show that choroidal thickening may be unique to T1DM or the duration of disease because similar changes are not observed in our cohort of T2DM who have a shorter duration of disease.

Our contemporaneous retinal and choroidal assessment using SS-OCTA revealed several novel findings. We show that the volume of vascular structures in the choroid as compared with choroidal stroma (i.e., CVI) is significantly greater in T1DM than in controls or T2DM. This suggests that the volume of blood in the choroid at any point in time is greater in T1DM without retinopathy than in controls, and this presumably explains the increasing trend in choroidal thickness that we and others^18,24,28,41, 42, 43 observe in T1DM as well. The increase in blood volume may be due to physical remodeling of the choroidal stroma that mechanically facilitates larger vascular structures. This may explain ultrasonographic findings of decreased blood flow in subjects with DM^6,44 due to decreased vascular resistance from choroidal vascular dilation. Our findings also show thickening of the choriocapillaris in T1DM, which is corroborated by histopathological studies of diabetic eyes that reveal an increased deposition of basal laminar deposits, polymorphonuclear cells, and intercellular adhesion molecule 1 in the choriocapillaris.²

Although there was no statistically significant difference in any of the retinal vascular metrics between youth with and without diabetes, we observed that children with T1DM showed a nonsignificant but consistently increased VSD in the superficial retinal layer, deep retinal layer, and the whole retina, as well as an increase in flux, compared with T2DM subjects and controls. As we briefly discussed in the introduction, studies evaluating retinal vascular density in T1DM have reported controversial results,^10,18,20,45, 46, 47, 48, 49 with a few studies reporting insignificant results similar to ours.25, 26, 27 In a meta-analysis of early microvascular changes detected by OCTA in adults without DR, subgroup analysis showed no statistically significant differences in perfusion density between T1DM patients without DR and healthy controls, except in the parafoveal perfusion density within the deep capillary plexus.¹⁷ The best explanation for these conflicting findings is that perfusion varies with severity, duration, and type of DM, as well as other physiological variables that are not controlled. Glycemic fluctuations and postprandial hyperglycemic peaks may trigger the increase of local blood flow, in analogy to what was reported in other tissues, such as the kidney, muscle, and skin. Hemodynamic changes seem to be mediated by the increased nitric oxide activity and vasodilatory prostaglandins due to local retinal acidosis, which is prominent in the early stages of DR.50, 51, 52, 53 These findings support the possibility that subclinical choroidal changes in T1DM may precede, or at least be more easily detectable than, retinal vascular changes. Further longitudinal studies will be required to understand whether this also applies to pediatric DM.

Our study is unique in assessing youth with T2DM. In youth with T2DM, there is a high prevalence of DR and other diabetes-related complications after a short duration of diabetes.^1,30 In the TODAY (Treatment Options for Type 2 Diabetes in Adolescents & Youth) study of youth with T2DM, OCT imaging was performed at 2 time points 7 years apart, demonstrating decreases in total retinal thickness,⁵⁴ but OCTA was not performed. Studies in adults with T2DM have consistently demonstrated that OCTA can identify microvascular alterations in patients without clinically detectable DR. These include but are not limited to microaneurysms, deformation of the FAZ, and reductions in various vascular density measures.7, 8, 9^,16,17 In our study, however, we did not observe statistically significant differences between T2DM children and controls, likely due to our limited sample size and a shorter duration of DM in these children. Future studies with a larger sample size of T2DM children will be important in order to understand how these factors contribute to DR development in this high-risk population.

This study has several limitations. Although the cohort is diverse and includes T1DM, T2DM, and pediatric controls, a combination not previously reported to our knowledge, the sample size is small and therefore we are underpowered to detect subtle changes. Despite this, we were able to detect differences in the groups that are consistent with histopathology and potential biomarkers of early disease if validated in larger studies. Second, the relatively short disease duration in T2DM subjects limits our ability to observe potential long-term vascular changes, but this problem is inherent in studying children and can be overcome with the prospective collection that we are planning. Most T2DM patients in our dataset had a disease duration of <5 years (mean: 2.4 ± 1.3 years), which may be too early to see any changes, whereas T1DM patients had a relatively longer disease duration (mean: 8.0 ± 4.0 years). Future studies should be longitudinal and include a more ethnically diverse cohort and a broader range of disease durations to validate these findings in a larger pediatric population.

Conclusion

This study describes contemporaneous retinal and choroidal vascular changes using widefield OCTA in a cohort of youth with T1DM, T2DM, and controls using state-of-the-art volumetric choroidal analyses. Similar to the value of OCTA as a noninvasive tool for detecting early retinal microvascular changes in adult DM patients, our findings underscore its potential as a rapid and noninvasive tool to detect subclinical changes in pediatric DM patients, a more vulnerable and at-risk cohort.

Manuscript no. XOPS-D-25-00278.