Early Retinal Changes Detected Using Optical Coherence Tomography Angiography in Youth with Type 1 Diabetes Mellitus: A Case-Control Study

Abstract

Introduction

Despite contemporary management, the incidence of diabetic retinopathy in adolescents during the first 2–5 years of type 1 diabetes mellitus (T1DM) remains unchanged, underscoring the need to better understand early retinal changes. In this study, we investigated early retinal changes in children with T1DM using optical coherence tomography angiography (OCTA). Incorporating a literature review of retinal changes in other age groups, we propose a model sequence of retinal changes in patients with T1DM.

Methods

In this prospective case-control study, conducted at National Taiwan University Hospital from 2021 to 2023, we prospectively recruited children with T1DM from a pediatric endocrinology clinic and control children undergoing routine eye examinations at the Department of Ophthalmology. All patients and controls underwent comprehensive ophthalmological examinations and OCTA.

Results

In the T1DM group, the average age at diagnosis was 7.39 ± 4.29 years, average diabetes duration was 5.0 ± 3.19 years, and average glycosylated hemoglobin level was 7.41 ± 0.92%. In the control group, the mean age was 10.8 ± 4.31 years, ranging from 5 to 20 years old, and that of diabetes mellitus (DM) group was 12.27 ± 4.32 years (p = 0.08). The T1DM group showed significant decreases in macular vessel density, particularly in the superficial and deep vascular plexuses. Full retinal thickness was greater in the perifoveal annulus in the T1DM group. Prepubertal onset of T1DM negatively impacted the superficial capillary plexus in both the parafoveal and perifoveal areas. Other DM factors did not significantly correlate with vascular parameters.

Conclusion

Using OCTA, we detected early, clinically inapparent microvascular reductions in vessel density in children with T1DM, while retinal layer thickness remained preserved. These findings, along with previous literature, are consistent with a hypothetical sequence in which microvascular alterations detectable by OCTA, structural changes on OCT may occur sequentially and prior to clinically visible structural changes.

Introduction

The prevalence of diabetic retinopathy (DR) in children declined from the 1990s to 2006, but has since plateaued [1, 2]. Even with improved monitoring and treatment, adolescents with type 1 diabetes mellitus (T1DM) continue to develop DR at similar rates [3–5]. Tracking DR trends in early-onset diabetes is essential, given the long disease course and the risk of microvascular complications. Identifying retinal changes before overt microvascular damage becomes visible is particularly valuable.

Optical coherence tomography (OCT) provides faster and more detailed retinal examinations, making it feasible even for preschool-aged children with proper guidance. In clinical practice, optical coherence tomography angiography (OCTA) has become an important adjunct to fluorescence angiography for early disease detection and screening. In adults, changes in superficial and deep retinal vessel density (VD) in the macular region correlate with the severity of DR [6]. OCTA can detect differences in the deep capillary plexus (DCP) in patients with diabetes mellitus (DM) and no clinically perceivable DR [7, 8], supporting the idea that microvascular abnormalities precede clinically identifiable DR. Recent research in young adults with T1DM suggests that prepubertal onset of DM is the most powerful factor leading to retinal vascular changes, aligning partially with current recommendations for pediatric DM ocular monitoring, which state that routine eye examinations should begin at puberty [9]. However, guidelines for ocular screening in pediatric patients lack uniformity, and the earlier sequential variations detectable using OCTA before overt retinopathy remain unclear. Therefore, we proposed a prospective study targeting a younger cohort to characterize early retinal alterations in T1DM population.

Previous retrospective data from diverse populations and ages have shown trends such as thinning of the nerve fiber layer and reduced deep parafoveal vascular densities in patients with DM [10, 11]. Whereas, in the present study, we prospectively enrolled young patients with T1DM, comprehensively evaluated OCT data from a single examination, and compared differences between the eyes of children with T1DM and those of healthy controls. To contextualize our study within the existing literature, we compiled a summary of previously reported OCT and OCTA findings across cohorts of different age groups and disease durations (Table 1) [9, 12–16]. This literature-based overview highlights reported microvascular and neuronal alterations in pediatric, young adult, and adult populations with T1DM prior to clinically detectable DR, and illustrates how reported findings vary with age and disease stage [14, 17].

Table 1.

New findings in our study and summary of OCT/OCTA changes across age groups in T1DM reviews

​	Our cohort	Children	Young adult	Adult
Clinical stage	No DR	No DR	No DR	No DR
Age, years	Range: 5–20 (mean: 12.27)	Range: 8–19 (mean: 14.68) [12]	Mean: 24 [14]	Range: 22–65 (mean: 42.93) [16]
Diabetes duration, years	5.0±3.19	7.16±3.59 [12]	13.43±5.44 [9]	21.3±10.6 [15]
Vascular parameters on OCTA
Vessel densities	Decrease in SCP, trend of decrease in DCP	VD normal [12] (mean: 14-year-old)	Decrease in parafoveal VD in SCP, DCP [9] (mean: 23-year-old)	Decreased parafoveal VD in DCP [15] (mean: 42-year-old)
Foveal characteristics	FAZ, FP, FD, FA normal	FAZ normal [12]	Increase in FAZ [9]	FAZ same [15], increased [17]
Neuronal thickness on OCT
Full thickness	Thickening in perifoveal areas	​	Macular full thickness normal [14]	Thickening in full macular, parafoveal S, I [16] (mean: 43-year-old)
Inner layers (RNFL/GCC/IPL)	No decrease	Thinning of the RNFL and GCC [13] (mean: 12-year-old)	Thinning of RNFL and GCC thickness [14] (mean: 24-year-old)	Thinning in inner layers [16]
Peripapillary RNFL	No decrease	​	Decrease in sup. peripapillary RNFL [14]	​

DM, diabetes mellitus; OCT, optical coherence tomography; OCTA, optical coherence tomography angiography; T1DM, type 1 diabetes mellitus; DR, diabetic retinopathy; SCP, superficial capillary plexus; DCP, deep capillary plexus; VD, vessel density; FAZ, foveal avascular zone; FP, foveal perimeter; FD, foveal vasculature density; FA, acircularity of FAZ; S, superior; I, inferior; RNFL, retinal nerve fiber layer; GCC, ganglion cell complex; IPL, inner plexiform layer.

Subjects

Patients aged 5–20 years with T1DM were recruited consecutively from the pediatric endocrinology clinic. Dilated fundus examination was used in accordance with the AAO Preferred Practice Pattern for DR, and only those with absence of any clinical DR were enrolled. During the study period, UWF imaging was not available in our institution, and standard ETDRS 7-field photography was not feasible in younger children.

Nondiabetic controls in the same age range were recruited from routine eye examinations at the Department of Ophthalmology, National Taiwan University Hospital, and were matched by age and refractive status. To avoid confounding vascular effects from high myopia, controls were required to have spherical equivalents between −5.00 D and +3.00 D and astigmatism within 2.00 D. Participants with systemic diseases aside from DM – including vascular, neurodegenerative, inflammatory, or infectious conditions – were excluded.

Data Collection

Data collected at enrolment included height, bodyweight, resting blood pressure, glycosylated hemoglobin (HbA1c) value, and history of DM (onset age and duration). After enrolment, an ocular examination was performed, including refractometry, intraocular pressure measurement, slit-lamp biomicroscopy, axial length (AxL) measurement, extensive fundoscopic examination, and OCTA. Refractive status, intraocular pressure, and AxL were measured in triplicate, and the mean values were recorded.

OCTA (Avanti RTVue XR, Optovue Inc., CA, USA) was performed after full pupillary dilation. Image quality control was applied at the time of acquisition, and only scans meeting predefined criteria – defined as signal strength index of ≥8/10 and absence of segmentation errors on manual review – were retained for analysis. Scans below the quality threshold were not systematically recorded as they were not entered into the analytic dataset; we acknowledge that the lack of a record for the number of excluded scans per group is a limitation that may introduce selection bias in image quality. There was no obvious qualitative imbalance in image acquisition or scan quality between the T1DM and control groups at the time of imaging. Identical acquisition protocols and quality thresholds were applied across all participants.

Imaging boundaries included the Early Treatment of Diabetic Retinopathy Study (ETDRS) macular map sectors of 3-mm and 6-mm areas. The imaged VD was denoted as a percentage, separated into superficial VD (defined as vascular structures within the inner limiting layer to 10 μm above the inner plexiform layer) and deep VD (from 10 μm above the inner plexiform layer to 10 μm below the outer plexiform layer) per the original imaging protocol.

Foveal parameters included the area of the foveal avascular zone (FAZ), acircularity of the FAZ, and foveal vasculature density. The acircularity of the FAZ was denoted by the acircularity index, calculated by dividing the perimeter of the FAZ by the perimeter of a circle with an equal area [18, 19]. Foveal VD was measured as the VD in the 300-µm ring of the foveal area, calculated by dividing the number of vessel pixels by the total number of pixels, and recorded as a percentage.

OCT macular thickness measurements were obtained using the same RTVue XR device with a macular cube scan of 6 × 6 mm centered on the fovea. Automated retinal layer segmentation was performed using the device’s built-in software, and all segmentation boundaries were reviewed by trained graders. Scans with low signal quality (signal strength <8/10), motion artifacts, or segmentation errors were excluded. Full retinal thickness was defined as the distance from the internal limiting membrane to the retinal pigment epithelium, and inner retinal thickness as the distance from the internal limiting membrane to the inner plexiform layer. Outer retinal thickness was calculated as the difference between full retinal thickness and inner retinal thickness, based on automated layer segmentation.

To characterize the timing of diabetes onset relative to pubertal development, pubertal status at diabetes onset was classified using Tanner staging, with pubertal onset defined as Tanner stage of ≥2. Accordingly, “prepubertal onset” was defined as diagnosis of T1DM prior to pubertal onset (Tanner stage of <2 at the time of diagnosis). Pubertal status at the time of OCT/OCTA imaging was not systematically recorded; therefore, analyses referring to “prepubertal onset” pertain specifically to age and pubertal stage at diabetes onset rather than pubertal status at imaging.

Statistical Analysis

All analyses were conducted in R (v4.2.2). Categorical variables were compared using Fisher’s exact test, and continuous variables using the Mann-Whitney U test. To integrate bilateral data into the statistical models while accounting for the inherent correlation between eyes of the same individual, we used linear mixed-effects models with a random intercept for each participant. Group status served as the main predictor, and the models were adjusted for the prespecified covariates age, sex, BMI, AxL, and spherical equivalent, selected a priori based on their known influence on retinal microvascular and structural measurements. For exploratory analyses within the T1DM cohort, similar mixed-effects models were fitted with HbA1c, prepubertal onset (P_onset), diabetes duration, BMI, AxL, spherical equivalent, and sex specified as covariates.

To control for multiplicity among confirmatory endpoints, we defined two outcome families: capillary density (superficial capillary plexus [SCP] and DCP composites) and foveal geometry (FAZ area, acircularity index, foveal VD, and FAZ perimeter). For each eye, parafoveal and perifoveal capillary density values were extracted from the ETDRS macular grid. Sector-level values were standardized using z-scores relative to the control group distribution. The SCP and DCP composite measures were then calculated as the mean of the z-scored values across the eight included parafoveal and perifoveal sectors within each plexus. Family-wise error was controlled within each family using the Holm step-down procedure, with a gatekeeping structure in which capillary density was tested before foveal geometry. For capillary density, we used prespecified directional hypotheses (lower density expected in the DM group). Statistical significance was defined as α = 0.05.

Sample size estimation was based on prior data reporting differences between children with T1DM and controls in parafoveal DCP [9]. A power analysis for a two-sample t-test (α = 0.05, two-sided) indicated that 18 participants per group would provide 80% power to detect the expected difference.

Results

Eighty-three participants were enrolled in this study, including 40 with T1DM (80 eyes) and 43 controls (86 eyes). All participants contributed bilateral eye data, and both eyes were included in the analysis for all subjects. Linear mixed-effects models with a random intercept for each participant were used to account for the correlation between eyes. Their ages ranged from 5 to 20 years, with a mean age of 12.27 ± 4.32 years in the DM group and 10.8 ± 4.31 years in the control group (p = 0.08). The mean AxL was 23.87 ± 1.33 mm in the DM group and 23.72 ± 0.7 mm in the control group (p = 0.64). The sex ratio was 22:18 (boy:girl) in the DM group and 19:24 in the control group (p = 0.42). In the DM group, the average time since diagnosis was 5.00 ± 3.19 years, with a mean age at diagnosis of 7.39 ± 4.29 years. The mean glycosylated hemoglobin level in the DM group was 7.41 ± 0.92% (Table 2). Among participants with T1DM, diabetes onset occurred prior to puberty (Tanner stage < 2) in 28 individuals and after pubertal onset in 12 individuals. These distributions are summarized in Table 2.

Table 2.

Demographic data of DM group and control group

​	DM group	Control group	p value
Participants, n	40	43	​
Age, mean±SD, years	12.27±4.32	10.8±4.31	0.08
Sex (boy:girl)	22:18	19:24	0.42
Axial length, mean±SD, mm	23.87±1.33	23.72±0.77	0.64
Body mass index, mean±SD, kg/m2	18.94±3.5	16.73±4.47	<0.001
DM duration, mean±SD, years	5.00±3.19	​	​
Age of onset, mean±SD, years	7.39±4.29	​	​
Prepubertal onset, n	28 (16:12; boy:girl)	​	​
HbA1c, mean±SD, %	7.41±0.92	​	​

DM, diabetes mellitus; SD, standard deviation.

Our primary objective was to identify differences in VDs detectable by OCTA between participants with and without T1DM. Our cohort with T1DM exhibited significant decreases in VD in the macular area. Group comparisons by t-test are presented descriptively (Table 3); inferential conclusions are based on the prespecified mixed-effects analysis with multiplicity control. In the superficial vascular plexus, all quadrants within the 3-mm annulus showed significant decreases in VD (all p < 0.05). In the perifoveal annulus, similar decreases were observed, with significant reductions noted only in the inferior and temporal quadrants (superior [S], p = 0.196; nasal [N], p = 0.098; inferior [I], p = 0.017; temporal [T], p = 0.007). In the deep vascular plexus, decreased VD was also observed in the DM group, with significant reductions in all quadrants in the parafoveal and perifoveal areas (p < 0.05).

Table 3.

Comparison of vascular parameters on OCTA between groups

​	DM	Control	p value	​	DM	Control	p value	​	DM	Control	p value
superficial retinal capillary densities, %				deep retinal capillary densities, %				central zone vascular characteristics
Central (1 mm)	22.27±5.89	23.65±7.45	0.37	Central (1 mm)	37.64±5.97	38.73±10.26	0.28	FAZ, mm2	0.25±0.08	0.23±0.12	0.48
Parafoveal (1- to 3-mm annulus)				Parafoveal (1- to 3-mm annulus)				FP, mm	1.92±0.31	1.81±0.46	0.45
Superior	52.73±3.98	54.15±4.92	0.03	Superior	55±4.7	56.01±9.29	0.04	FD, %	52.97±5.4	54.35±7.42	0.091
Nasal	51.12±3.09	53±4.6	0.01	Nasal	54.24±4.76	56.07±7.37	0.02	FA	1.1±0.05	1.1±0.05	0.36
Inferior	51.77±5.35	53.55±4.24	0.03	Inferior	56.24±3.53	58.14±4.11	0.02
Temporal	50.98±4.81	52.73±4.68	0.03	Temporal	52.38±4.38	54.3±8.28	0.02
Parafoveal mean	51.65±3.49	53.33±4.07	0.02	Parafoveal mean	54.47±3.66	55.83±7.5	0.03
Perifoveal (3- to 6-mm annulus)				Perifoveal (3- to 6-mm annulus)
Superior	51.14±4.28	50.62±9.11	0.2	Superior	52.32±5.38	56.31±5.59	0.01
Nasal	54.46±3.31	55.29±2.62	0.1	Nasal	48.26±6.58	53.17±6.71	0.01
Inferior	50.41±4.66	52.03±3.01	0.02	Inferior	48.57±6.45	53.16±7.28	0.01
Temporal	46.65±4.64	49.21±2.97	0	Temporal	46.92±6.84	51.69±7.89	0.02
Perifoveal mean	50.66±3.58	52.17±2.51	0.01	Perifoveal mean	49.02±5.82	53.7±6.69	0.01

AxL, axial length; DM, diabetes mellitus; SD, standard deviation; OCTA, optical coherence tomography angiography; FAZ, foveal avascular zone; FP, foveal perimeter; FD, foveal vasculature density; FA, acircularity of FAZ.

When comparing the characteristics of the central FAZ, the DM group exhibited a slightly larger foveal avascular area (0.25 ± 0.08 mm² in DM, 0.23 ± 0.12 mm² in control; p = 0.48) and perimeter (1.92 ± 0.31 mm in DM, 1.81 ± 0.46 mm in control; p = 0.445). Foveal vasculature density was less in the DM group (p = 0.091; Table 3).

Retinal thickness was similar between groups. The only significant difference observed in the macular area was at the perifoveal annulus, where the DM group exhibited a thicker full retinal layer (superior [S], p = 0.063; nasal [N], p = 0.025; inferior [I], p = 0.042; temporal [T], p = 0.014). Inner retinal layer thickness and peripapillary nerve fiber layer thickness were comparable between groups (all p > 0.05; Table 4).

Table 4.

Comparison of retinal thickness between groups

	DM	Control	p value		DM	Control	p value		DM	Control	p value
inner retinal thickness, μm				full retinal thickness, μm				peripapillary RNFL thickness
Center (1 mm)	52.8±7.91	54.97±8.99	0.39	Center (1 mm)	244.72±31.22	243.5±16.38	0.62	Supero-temporal	149.13±19.08	149.5±14.01	0.75
Parafoveal (1- to 3-mm annulus)				Parafoveal (1- to 3-mm annulus)				Supero-nasal	116.51±16.1	117.16±17.87	0.91
Superior	111.35±17.63	111.74±13.33	0.96	Superior	328.15±23.8	322.97±19.89	0.13	Nasal superior	82.56±12.36	77.08±10.83	0.098
Nasal	111.22±7.56	112.03±8.23	0.67	Nasal	328.6±15.9	323.92±12.49	0.27	Nasal inferior	69.87±10.33	72.68±14.71	0.42
Inferior	113.7±6.94	112.32±14.03	0.77	Inferior	325.45±14.92	317.66±20.73	0.13	Infero-nasal	110.33±20.11	112.03±17.4	0.52
Temporal	103.17±9.06	102.87±12.06	0.87	Temporal	316.65±17.14	311.55±16.08	0.16	Infero-temporal	148.74±15.27	147.83±13.13	0.56
Perifoveal (3- to 6-mm annulus)				Perifoveal (3- to 6-mm annulus)				Temporal superior	91.38±19.28	89.03±11.07	0.90
Superior	102.11±7.58	102.03±7.14	0.89	Superior	292.7±12.92	288.14±10	0.063	Temporal lnferior	80.44±20.93	91.72±99.34	0.53
Nasal	118.86±8.05	118.06±10.08	0.65	Nasal	308.84±13.89	301.31±12.99	0.025
Inferior	99.78±5.93	99.8±9.16	0.73	Inferior	280.76±12.02	275.11±10.07	0.042
Temporal	88.7±4.95	88.03±7.89	0.39	Temporal	278.24±12.4	271.14±11.7	0.014

DM, diabetes mellitus; SD, standard deviation.

Consistent with the prespecified outcome families and multiplicity control strategy described in the Methods, SCP and DCP composite measures, defined as the mean z-scored parafoveal and perifoveal capillary density values across the ETDRS sectors for each plexus, were evaluated using regression models adjusting for prespecified covariates and multiplicity. The prespecified analysis showed that SCP composite VD was significantly reduced in the T1DM group compared with controls (p = 0.016; Holm-adjusted p = 0.031). To address clinical magnitude in interpretable units, the adjusted model showed a mean reduction of 1.96% (95% CI: −3.75% to −0.18%) in raw VD for the SCP composite, corresponding to a large effect size (Cohen’s d = −0.88). The DCP composite did not reach statistical significance after adjustment (p = 0.13; Holm-adjusted p = 0.13), although the estimated effect was in the same direction with a comparable mean reduction of 1.90% (95% CI: −5.04% to 1.24%; Cohen’s d = −0.80) (online suppl. Table S3; for all online suppl. material, see https://doi.org/10.1159/000552020). The lack of significance in the DCP despite a similar effect size reflects the higher inherent variability in the deep vascular plexus. Among foveal geometry measures, none remained significant after Holm correction. Quadrant-wise sector analyses were performed for descriptive purposes and are considered exploratory; corresponding p values are therefore unadjusted and should be interpreted cautiously. Results are summarized in Table 5; exploratory generalized linear mixed-model analyses are provided in online supplementary Table S1.

Table 5.

GLMM-adjusted differences (DM − control) in OCTA outcomes

​	Adj. diff. (DM – control)a	95% CI	p value	Holm
A. Capillary density composites (directional testing: DM < Control)	​	​	​	​
SCP composite	−0.53	[−1.02, −0.04]	0.016	0.031
DCP composite	−0.30	[−0.82, 0.23]	0.13	0.13
B. Foveal geometry endpoints
FAZ area	0.0035	[−0.0519, 0.0589]	0.9	0.9
FA (ratio)	0.033	[−0.0103, 0.0758]	0.14	0.40
FD, %	−3.01	[−5.97, −0.05]	0.046	0.18
FP, mm	0.13	[−0.110, 0.369]	0.29	0.57

GLMM, generalized linear mixed model; OCTA, optical coherence tomography angiography; SCP/DCP, superficial/deep capillary plexus; FAZ, foveal avascular zone; FA, foveal acircularity index; FD, foveal density (300-µm annulus); FP, foveal perimeter.

^aAdjusted differences estimated from linear mixed-effects models (GLMM) with a random intercept for participant to account for two eyes per subject. Composite capillary density measures (SCP composite, DCP composite) were calculated as the mean of z-scored parafoveal and perifoveal vessel densities within each plexus. Directional testing (DM < control) was prespecified for the capillary density family. Holm step-down correction was applied separately within the capillary density family and the foveal geometry family to control family-wise error.

DM-associated variables were further categorized for a detailed second round of regression analysis to explore the correlation with vascular parameters. Exploratory generalized linear mixed model controlling HbA1c, BMI, AxL, spherical equivalent, and sex was performed and result is as online supplementary Table S2. In order to address multiplicity, we used predefined capillary density composite as endpoints, as which analyzed between groups. As a result, prepubertal onset of T1DM negatively impacted the SCP (p = 0.0023; Holm-adjusted p = 0.0046) after adjusted for HbA1c, BMI, AxL, spherical equivalent, and sex. No association was observed for the DCP composite (Holm-adjusted p = 0.62) (Table 6).

Table 6.

Within-DM mixed-effects regression: association of puberty with capillary density composites

​	Adj. diff. (onset before puberty – onset after puberty)	95% CI	p value	Holm
SCP composite	−1.06	[−1.79, −0.33]	0.002	0.0046
DCP composite	0.08	[−0.43, 0.60]	0.62	0.62

Discussion

We prospectively recruited participants with T1DM and age-matched controls to compare retinal changes using OCTA. The T1DM group, particularly those with prepubertal onset exhibited significantly lower VD across the macular region, representing our primary finding.

Because all participants had no clinically detectable DR, no peripheral lesions were present by definition. Our posterior pole OCTA findings therefore represent preclinical macular microvascular alterations, which may precede or parallel the peripheral microvascular changes described in later stages of DR.

DR has traditionally been assessed based on clinically visible retinal lesions; however, advancements in imaging allow identification of subclinical microvascular and structural changes at earlier stages. This may aid earlier detection and management of this potentially blinding disease. In 2018, Abramoff et al. [20] questioned whether neurodegenerative changes, beyond purely vascular lesions, should be incorporated into screening protocols. Building on this concept, our study specifically examined early microvascular and structural markers using OCT/OCTA in youth with T1DM.

OCT has documented global thinning of the retinal nerve fiber layer (RNFL) in young patients with T1DM, specifically in the temporal superior and nasal inferior regions, compared to the healthy population [11]. These structural decreases have been linked to longer diabetes duration and higher HbA1c levels. Vascular studies have also shown parafoveal VD reductions in children aged 6–17 years, while some reports described increased perfusion in the inner superficial area [10]. In contrast to these mixed findings, our T1DM cohort demonstrated a consistent pattern of reduced VD in both parafoveal and perifoveal regions and across both the superficial and deep plexuses. This aligns with the known early microvascular effects of hyperglycemia, in which capillary rarefaction precedes neovascularization.

Notably, our results revealed perifoveal retinal thickening in the T1DM group, whereas inner retinal thickness did not differ, suggesting that the outer retinal layers contributed to this increase. This differs from several studies in older T1DM populations, which generally report retinal thinning and RNFL loss; our younger cohort instead showed early vascular dropout accompanied by preserved neuronal structure and perifoveal thickening (Table 1) [14, 21]. Neuro-retinal alterations in adults with prediabetes – such as perifoveal thinning or selective inner retinal thinning with preserved or thickened outer retinal layers – have been described before clinically overt microvascular damage [14, 16]. Our observed thickening of the outer retinal layers is consistent with hypotheses of subclinical edema related to hyperglycemia or inflammation [16]. Furthermore, peripapillary RNFL thinning was primarily observed at the temporal side, while other areas were comparable to those of the healthy population [21]. By contrast, our T1DM cohort did not exhibit thinning in the peripapillary RNFL, as noted in the EPIC-Norfolk Eye Study, suggesting that neural layer changes may not have manifested yet [22]. This discrepancy likely reflects the younger age and shorter disease duration in our cohort, suggesting that neuronal changes may not yet have manifested. These age-stratified differences support the possibility that outer retinal alterations and capillary rarefaction may precede detectable inner layer neuronal loss.

Regarding DM-associated parameters affecting retinal microvascular changes, prepubertal onset of DM was not a potent modifier in the literature. However, both HbA1c and creatinine levels reduced parafoveal superficial VD (p = 0.039) and parafoveal thickness (p = 0.003) [23]. The duration of DM is associated with overt DR and is used as a cut-off point for initiating screening [24, 25]. In our cohort, DM-related parameters, including HbA1c levels and DM duration, did not significantly correlate with vascular parameter differences in OCTA. Contrary to reported findings in the literature, prepubertal onset of T1DM strongly correlated with a decrease in SCP in our population. The poor correlation of vascular densities with HbA1c levels and DM duration may be due to younger patient ages and smaller ranges in these parameters in our cohort.

No statistically significant sex-related effects were observed. This may reflect the relatively young age distribution of the cohort and the limited proportion of participants who had entered puberty at diabetes onset [26–28]. All regression models adjusted for sex, age, and AxL to minimize confounding; however, because sex and pubertal status are interrelated during adolescence, such adjustment may attenuate detectable sex-specific effects in cross-sectional analyses of predominantly prepubertal or early pubertal populations.

Differences in OCTA parameters have been reported across DR populations. Some studies have shown decreased VD or perfusion at the DCP in young adults with T1DM without associated thickness changes [15, 29], and others noted changes in superficial parafoveal VD [9]. Our cohort displayed reductions in both the superficial and deep plexuses, with SCP reduction particularly associated with prepubertal onset. These vascular and structural alterations, detectable using OCT and OCT angiography in children with T1DM prior to clinically overt DR, provide biologic context for current guideline recommendations that advocate periodic ocular screening in this population. Our findings do not define optimal screening intervals, but underscore the presence of subclinical retinal changes during childhood that justify ongoing surveillance [2, 30]. Based on our early-stage findings, we recommend incorporating OCTA alongside indirect fundoscopy, as VD reduction may represent one of the earliest detectable indicators available on current imaging modalities. Comparing parameters with established nomograms may further help stratify individual risk and determine whether more intensive surveillance is warranted.

One limitation of our study is the modest sample size and relatively short DM duration. However, this homogeneity minimizes age-related confounders and enhances the ability to isolate diabetes-specific effects. The other strength is that our cohort comprised children followed routinely in a pediatric endocrinology clinic, representing a broad, real-world T1DM population rather than a referral-biased ophthalmic cohort. Another limitation of this study is that pubertal status was only recorded at the time of T1DM diagnosis but not at the time of imaging. This constraint limits the interpretation of sex- and puberty-related findings, as hormonal changes during this period are known to influence vascular parameters.

Together, these results support the growing need to integrate OCTA into pediatric diabetic eye care, particularly in patients with early-onset T1DM. Our findings complement existing screening guidelines and may help refine risk stratification based on both vascular and neuronal biomarkers.

Conclusion

In this prospective pediatric study, we identified early retinal alterations in youth with T1DM using OCTA. Based on integration of prior adult and prediabetes literature together with our cross-sectional pediatric findings, we propose a hypothetical model in which microvascular alterations and perifoveal outer retinal thickening may occur early, followed by inner retinal thinning and eventual peripapillary RNFL loss, ultimately preceding clinically overt DR (Fig. 1).

Fig. 1.

Hypothetical framework of retinal alterations detectable by OCT/OCTA in T1DM prior to clinically apparent diabetic retinopathy.

The combination of vascular and structural OCTA parameters may offer a sensitive means of detecting preclinical retinal involvement in pediatric T1DM. Due to the limitation of our study’s cross-sectional nature, and longitudinal follow-up of this cohort will be essential to validate the proposed sequence and to inform earlier, individualized screening protocols.

Acknowledgment

We would like to thank Editage [http://www.editage.com] for editing and reviewing this manuscript for English language.

Statement of Ethics

This prospective case-control study was conducted at National Taiwan University Hospital (NTUH) from 2021 to 2023. Prior to enrolment, all participants were required to obtain written informed consent from parents or their legal representatives. The study protocol was approved by the NTUH Research Ethics Committee (202009088RINA) and complied with the principles of the Declaration of Helsinki.

Conflict of Interest Statement

The authors declare that they have no competing interests.

Funding Sources

This study was supported by the National Taiwan University Hospital, Hsinchu Branch (Grant Nos. 110-BIH014 and 111-BIH018 to I-Hsin Ma). The funding sources supported the study execution, English language editing, and publication costs.

Author Contributions

Conceptualization: I-Hsin Ma, Yi-Ching Tung, and Tzu-Hsun Tsai; data curation: I-Hsin Ma, Wei-Lun Huang, and Tzu-Hsun Tsai; formal analysis: I-Hsin Ma, Wei-Lun Huang, and Yao-Lin Liu; funding acquisition: I-Hsin Ma; project administration: I-Hsin Ma and Tzu-Hsun Tsai; writing – original draft: I-Hsin Ma; and writing – review and editing: I-Hsin Ma, Wei-Lun Huang, Yao-Lin Liu, and Tzu-Hsun Tsai.

Funding Statement

This study was supported by the National Taiwan University Hospital, Hsinchu Branch (Grant Nos. 110-BIH014 and 111-BIH018 to I-Hsin Ma). The funding sources supported the study execution, English language editing, and publication costs.

Data Availability Statement

The data that support the findings of this study are not publicly available due to privacy reasons but are available from the data sharing committee upon request.