Retinal Neurophysiological Parameters in Young Patients With Type 1 Diabetes

Abstract

Purpose

Diabetic retinopathy has prominent microvascular manifestations, but preclinical studies also find neuroglial cell abnormalities. This study aimed to examine retinal electrophysiological characteristics and their relation to diabetes duration and glycemia history in type 1 diabetes (T1D).

Methods

This cross-sectional study included 69 patients with T1D aged 6-32 (mean 16) years in equally sized subgroups of T1D duration <5 months, approximately 5 years, 10 years or 15 years. Exclusion criteria were chronic disease other than T1D, and retinal disease other than fundus photographic evidence of diabetic retinopathy. Comparison was made with 54 age-matched healthy subjects aged 8-34 (mean 17) years. Examinations included fundus imaging, electroretinography (ERG), dark adaptometry, and structural and angiographic optical coherence tomography.

Results

In patients with T1D, ERG implicit times of the dark-adapted 85 troland seconds (Td·s) a-wave and light-adapted flicker 85 Td·s wave increased with T1D duration in simple linear regression (P = 0.041 and 0.027, respectively). In patients with >4 years T1D duration, implicit times of flicker 16 Td·s and flicker 32 Td·s in dim ambient lighting increased with higher aggregate HbA_1c means (P = 0.018 and 0.036, respectively). For all T1D patients, implicit times of dark-adapted 85 Td·s a-wave and b-wave increased with higher blood glucose (P = 0.041 and 0.023, respectively). Dark adaptation rod intercept and retinal thicknesses were comparable between the groups.

Conclusions

This study of young T1D patients with little or no fundus photographic signs of retinopathy showed that selected ERG implicit times were longer with higher previous HbA_1c levels and longer diabetes duration.

Diabetic retinopathy usually develops slowly and asymptomatically. In principle, diabetes should be identified early enough and treated to such standards that retinopathy does not become a problem. With conventional fundus photographic and vision monitoring techniques, the first signs of diabetic retinopathy appear years or decades after the onset of type 1 diabetes (T1D). In a cohort study, more than half of adolescents with T1D had developed diabetic retinopathy after a mean diabetes duration of 12 years.¹ Although this detection enables early ophthalmic intervention, opportunities to optimize diabetes care may have been missed. Earlier detection of diabetic retinopathy by methods that are more sensitive than color fundus photography may therefore lead to better long-term outcomes.²

Clinical diagnosis of diabetic retinopathy relies on detection of microvascular abnormalities. However, numerous studies indicate that these are initiated by neural dysfunction. In preclinical studies of diabetic rodents, retinal ganglion cells have been shown to become dysfunctional³ by impairment of, for example, protein and lipid synthesis,⁴ indicating a disturbance of their high metabolic requirements.⁵ Also, leukostasis and inflammation have been suggested to lead to gliosis and neurodegeneration.⁶ In clinical studies, the velocity of optic nerve conduction, measured by visual evoked potentials is diminished already within the first year after the diabetes diagnosis.⁷ After longer diabetes duration, electrical signal transmission through the retina, measured using electroretinography (ERG), may have increased latencies and diminished amplitudes.^8,9 Notably, diminished amplitudes of the ERG oscillatory potentials have been suggested to predict the development of proliferative diabetic retinopathy.^9–11 Furthermore, thinning of the retinal nerve fiber layer on optical coherence tomography (OCT)¹² is a sign of neurodegeneration,¹³ and it has been colocalized with non-perfused regions on OCT angiography (OCTA).¹⁴

In this study, we examined if early functional and structural abnormalities in the retina can be detected using noninvasive clinical examination methods in young patients with T1D. We further assessed the value of these methods by examining their correlation with glycemia history. This report covers the baseline examination of participants in a cohort study that is intended to assess long-term retinal complications in relation to past glycemia in T1D.

Methods

Study Design and Participants

This was a cross-sectional analysis of baseline data of an observational, prospective cohort study initiated in August 2020, and the last of the present data were collected in January 2023. All patients were recruited via phone contact after approval by their patient responsible medical doctor or by advertisement at the Department of Pediatrics and Adolescent Medicine, Herlev and Gentofte Hospital, and Steno Diabetes Center Copenhagen, where those under and above the age of 18 years, respectively, had their ongoing diabetes follow-up. For the patients, the inclusion criteria used were T1D onset ≤ 18 years of age and age ≥5 years, and the exclusion criteria were significant chronic systemic disease other than T1D and ocular disease other than diabetic retinopathy in the study eye. The T1D patients were recruited with the aim of having equal numbers of participants for each of the following subgroups: T1D duration <5 months, 5 years, 10 years, and 15 years. Age-matched healthy subjects were recruited by advertisement in hospitals and schools and at official websites for recruitment of research participants. For the healthy subjects, the inclusion criteria were age ≥5 years and intended approximately ± 1 year to a T1D patient, and the exclusion criteria were significant chronic systemic diseases or ocular disease in the study eye (diabetes was excluded by measuring HbA_1c < 48 mmol/mol). Participants were not offered any compensation besides coverage of transportation.

All previous HbA_1c values of the T1D patients with T1D durations >4 years were registered retrospectively from the time of diagnosis until the examination date by review of their health records. A time-weighted average of previous HbA_1c values was calculated by dividing the cumulated area under the time curve of HbA_1c values from the first value recorded when HbA_1c had stabilized after diabetes onset to the HbA_1c measured in the present study (see below), with the duration of this period in days. Throughout the text, this measure is referred to as “previous HbA_1c.” The standard deviation (SD) of the mean of these previous HbA_1c values was used to express the stability of the blood glucose regulation.

The study was conducted according to the Declaration of Helsinki and was approved by the Regional Committee on Health Research Ethics (H-19080620). Written informed consent was obtained from the subjects aged ≥ 18 years, or from the legal guardians of subjects <18 years.

Examination Program

All participants underwent a same-day examination at the Eye Department, Rigshospitalet, Glostrup, or at Steno Diabetes Center Copenhagen, Herlev, consisting of best-corrected Early Treatment Diabetic Retinopathy Study (ETDRS) visual acuity measurement, fundus imaging (Optos California, Monaco, or Silverstone, Dunfermline, Scotland, UK), swept-source OCT and OCTA (Topcon Triton, Tokyo, Japan), full-field ERG (RETeval, LKC Technologies Inc., Gaithersburg, MD, USA), and dark adaptometry (AdaptDx, MacuLogix, Inc., Harrisburg, PA, USA). Protocols are specified below. Both pupils were dilated with 1.0% tropicamide eye drops after visual acuity measurement. As default, the right eye was chosen for analysis, and the left eye was only chosen if image or examination quality of the right eye was insufficient.

In all participants, blood glucose was measured before dark adaptometry (either by finger prick blood samples or by a continuous glucose monitoring device in the T1D patients having such), and in the patients additionally after the last ERG measurements. Also, the participants had measurements of blood pressure, height and weight. Body mass index was calculated as weight in kg divided by height in m². Additionally, the T1D patients had a fasting blood sample within a week before or after the eye examination for measurement of HbA_1c, plasma glucose and lipids, amongst others, whereas healthy subjects had a point-of-care finger prick blood sample for measurement of HbA_1c (Tosoh G8 HPLC Analyzer, Tosoh Corporation, Tokyo, Japan).

Grading of Diabetic Retinopathy and Fundus Imaging Protocol

Diabetic retinopathy was graded from fundus images according to the ETDRS abbreviated summary severity scale.¹⁵ Fundus imaging was acquired by ultra-widefield scanning laser ophthalmoscopy using a red laser with wavelength 635 nm and a green laser with wavelength 532 nm for 200° color imaging and the latter for autofluorescence imaging. Images were obtained so that at least one image with each modality of each eye was gradable.

OCT/OCTA Protocols

OCT scans were obtained for analysis of inner retinal layer thicknesses, and OCTA scans were obtained for analysis of retinal capillary densities. The OCT/OCTA instrument used an infrared beam with an average wavelength of 1050 nm, a scan rate of 100 kHz, and had a nominal axial resolution of 8 µm and lateral resolution of 20 µm. The fast swept-source OCT instrument was chosen because short examinations are more child friendly and because the chosen instrument and the accompanying software package could calculate superficial capillary plexus densities.

The OCT protocol consisted of a 9 mm transfoveal line scan and a 12 × 9 mm combined macular and optic disc raster and radial scan. From the latter, retinal layer thickness of a segmentation including the nerve fiber- and ganglion cell layer (called “GCL++”), by a slab from the inner limiting membrane to the border between the inner plexiform layer and the inner nuclear layer (ILM-IPL/INL), were automatically provided for nine zones of a fovea-centered ETDRS grid.

The OCTA protocol consisted of at least three 3 × 3 mm macular scans each comprising 320 horizontal B-scans spaced 10.6 µm apart with an averaging of 4 frames per B-scan. For each participant, the OCTA scan with best image quality was chosen for analysis. OCTA en face angiograms of the superficial and deep capillary plexus were segmented automatically by a slab from, respectively, 2.6 µm to 15.6 µm and 15.6 µm to 70.2 µm from the inner limiting membrane. For the superficial capillary plexus only, capillary densities in percent could be automatically provided for four zones of a fovea-centered ETDRS grid (superiorly, inferiorly, nasally, and temporally of the fovea) using the proprietary software (IMAGEnet6, Topcon Corporation, Tokyo, Japan). According to previous studies, ischemia has been suggested to begin in the temporal sector, which is why that region was chosen for correlation analyses with ERG outcomes.¹⁶ Measures of the deep capillary plexus were not included as we found in another study that they were less significantly different from controls than those of the superficial (Torm M. et al. 2025, unpublished data).

ERG Protocol

ERG measures the retinal neuronal impulse reactivity in terms of implicit time and amplitudes in response to light stimulation. The RETeval ERG device was handheld, and its display was set to a dim red background to prepare for recording dark-adapted rod responses. The electrode was positioned 2 mm under the lower eyelid. Measurements were conducted with one eye occluded. In participants <18 years of age, only the right eye was examined to limit examination time, and in participants ≥ 18 years of age, both eyes were examined, but only one eye, typically the right, was chosen for analysis.

The full-field ERG protocol consisted of first flicker measurements performed under ambient dim lighting (flicker 16 Troland^a seconds (Td·s) at 28.3 Hz, as well as flicker 32 Td·s at 28.3 Hz) and then the International Society for Clinical Electrophysiology of Vision's (ISCEV) five-step protocol (default in the ERG device) after 30 minutes of dark adaptation (during dark adaptometry described below) (scotopic measurements of 0.28 Td·s b-wave at 0.5 Hz, 85 Td·s a-wave and b-wave at 0.1 Hz, and 85 Td·s total oscillatory potentials at 0.1 Hz) and then after 10 minutes of light adaptation (photopic measurements of 85 Td·s a-wave and b-wave at 2 Hz with a background of 850 Td, and flicker 85 Td·s at 28.3 Hz with a background of 850 Td). Implicit time (ms) and amplitude (µV) were noted. Examples of outputs of the ERG protocol are shown in the Figure.

Figure.

Examples of the outputs of the used full-field ERG protocol consisting of flicker measurements (A, B) and the International Society for Clinical Electrophysiology of Vision's five-step protocol (C–G) performed with a handheld ERG device. The example is taken from a 15-year-old participant with a diabetes duration of 10 years and very mild nonproliferative diabetic retinopathy (two microaneurysms) in the study eye. ISCEV, International Society for Clinical Electrophysiology of Vision; ms, milliseconds; OP, oscillatory potentials; Td·s, troland seconds; µV, microvolts.

Dark Adaptometry Protocol

Dark adaptometry was performed to examine the time course of the rods to reach a specific sensitivity in dark (rod intercept) and their maximum sensitivity after full dark adaptation. Dark adaptometry was conducted with one eye occluded and, according to the refraction measurement, a spherical equivalent lens correction in front of the examined eye. The default protocol used was “Research 83-50”: At test initiation, participants were bleached briefly with a flicker equivalent to 83% bleaching level for rods, after which sensitivity recovery was measured for the following 30 minutes. The zero log unit stimulus intensity was 1.0 (7.40e^0.1 scotopic milli-lux^b).¹⁷ Approximately every 30 seconds, thresholds were measured by stimulus lights of 200 milliseconds (ms) duration located 5° superior to the fovea. Rod intercept was determined by the proprietary software as duration (minutes) required for rod sensitivity to recover consistently to a criterion sensitivity value below 5.0 × 10⁻³ scotopic cd/m².¹⁷

Statistics and Statistical Considerations

The data were analyzed using RStudio 2024.04.2 version 4.4.1 (R Core Team (2024), R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing, Vienna, Austria). Data are shown as mean and SD or median and interquartile range. Statistical significance level was set to a P-value ≤ 0.050.

Prior to patient recruitment there was a lack of robust data for a power calculation, but it was carefully estimated that the post hoc strength of finding the OCTA abnormality diminished capillary density, according to Mameli et al.¹⁴, was close to 80% with an α value of 5% for a two-sided t-test. This power calculation was based on a standard deviation of 0.018 calculated from the confidence interval (CI) and a clinically relevant difference of 0.01, meaning a standardized difference of 0.55 and a required sample size of minimum 50 participants in each group.

The distribution of the data was assessed by visual inspection of histograms and residual plots. Because of the small sample sizes of the T1D duration subgroups, the Shapiro-Wilk test was also applied to ascertain whether data were normally distributed. If there were missing observations for some independent factors, the specific participant(s) was/were not included in the respective analysis.

Simple linear regression analysis was used to examine correlations in the T1D group between clinical outcomes and covariates, and multiple linear regression analysis was used to adjust for potential confounding effects. A t-test was used to determine if independent variables were statistically significant in the regression models. For post hoc analysis of significant correlations from simple linear regression, the Benjamini-Hochberg method was performed to a vector of the P-values from all four correlations of the specific outcome, controlling the expected false discovery rate for minimizing the risk of Type 1 error. Only the significant correlations from simple linear regression analysis are shown in this article, and the remaining can be found in Supplemental Material. The healthy subjects were omitted from the regression analyses as they cannot be categorized as having a “duration” of diabetes because they do not have diabetes.

Additionally, each of the T1D duration subgroups were compared with the healthy control group using pairwise Student's t-test if the outcomes were normally distributed or Wilcoxon test if they were not. As a correction for multiple comparisons of the latter tests, a post hoc Bonferroni correction was applied adjusting the P-values by multiplying them by the number of comparisons which was 4 (each subgroup compared with the healthy control group).

The neurophysiology branch of the study is exploratory only and performed for generating hypotheses as previous literature was too sparse for prespecified hypotheses about the correlations. In future, the study will do follow-up examinations of an expanded study population.

Results

Study Population

This cross-sectional analysis of observational baseline data from a T1D cohort included 123 eyes in 123 participants, 69 with T1D (age range 6–32 years) and 54 healthy age-matched volunteers (age range 7–34). All participants were eligible after examination. Of the diabetes patients, two declined participation in follow-up examinations, but we were allowed to use their baseline-data. Patients and healthy subjects had comparable best-corrected visual acuity (median 88 ETDRS letters for study eyes with range 73–91 and 79–91 ETDRS letters, respectively, shown in Snellen notation in Table 1), refraction, blood pressure, and body mass index (Table 1).

Table 1.

Demographic and Clinical Characteristics of Study Participants

	Patients With Type 1 Diabetes (n = 69)				Healthy Control Subjects (n = 54)
Sex (n)
Female	36				36
Male	33				18
Age (years), median (Q1; Q3)	15.0 (12.0; 18.0)				15.5 (12.0; 19.8)
Diabetes duration (years), median (Q1; Q3)	6.5 (4.7; 11.5)				—
Current HbA1c (mmol/mol), median (Q1; Q3)	61 (53; 70)				34 (32; 36)
Best-corrected visual acuity, median (Q1; Q3)	1.25 (1.0; 1.25)				1.25 (1.0; 1.25)
Refraction (diopters), median (Q1; Q3)	0.00 (−0.25; 0.00)				0.00 (−0.25; 0.00)
Systolic blood pressure (mmHg), mean ± SD	119 ± 10				116 ± 13
Diastolic blood pressure (mmHg), mean ± SD	75 ± 9				72 ± 10
Body mass index (kg/m2), mean ± SD	21.9 ± 3.8				21.9 ± 4.6
Diabetes Duration Subgroup Division	<5 Months (n = 17)	∼5 Years (n = 18)	∼10 Years (n = 17)	∼15 Years (n = 17)	—
Diabetes duration (years), median (Q1;Q3), range	0.12 (0.10; 0.23), 0.049–0.42	5.7 (5.3; 5.9), 4.5–6	10.0 (9.6; 10.4), 9–11.5	15.5 (14.6; 16.0), 14.5–17.5	—
Current blood glucose (mmol/L) at examination day, median (Q1;Q3)	6.7 (5.2; 8.2)	6.5 (5.5; 7.6)	7.6 (5.0; 10.6)	9.4 (7.9; 11.3)	5.1 (4.6; 5.4)
Diabetic retinopathy ETDRS severity*					—
No detectable retinopathy (no. of patients)	14	15	13	6	—
Very mild NPDR (no. of patients)	3	2	3	8	—
Mild NPDR (no. of patients)	0	1	1	2	—
Moderate NPDR (no. of patients)	0	0	0	1	—

ETDRS, Early Treatment Diabetic Retinopathy Study; NPDR, Nonproliferative Diabetic Retinopathy; Q1, the first quartile (25^th percentile); Q3, the third quartile (75^th percentile).

^*Total prevalence of diabetic retinopathy: Of 69 type 1 diabetes patients, 48 had no detectable retinopathy (22 male and 26 female), 16 had very mild NPDR (9 male and 7 female), 4 had mild NPDR (2 male and 2 female), and 1 patient had moderate NPDR (1 female).

The right eye was by default chosen for analysis of the ERG results, except for in five healthy subjects (the left eye was chosen in two of them because of accidental light exposure to the right eye after the dark adaptometry examination before initiation of the dark-adapted ERG measurements (as the left eye was occluded, it was not affected by this), in two of them because image quality of the OCTA scans used for comparison was only sufficient on the left eye, and in one of them because the participant only accepted medical pupil dilation in the left eye due to a contact lens in the right eye). In one healthy subject and in four diabetes patients, an OCTA scan of the left eye was included in the analysis because of insufficient image quality of the OCTA scans of the right eye, although ERG had only been performed in the right eye (because of age < 18 years, see “ERG protocol”). At least one OCTA scan of at least one eye was gradable except for in one healthy subject who was excluded from the OCTA analysis.

Regression Analyses Regarding Diabetes Duration, Glycemia History, and Blood Glucose

ERG testing among T1D patients found increasing implicit times of the dark-adapted 85 Td·s a-wave and the light-adapted 85 Td·s flicker wave with longer T1D duration as assessed by simple linear regression analysis (Table 2). ERG implicit times of dark-adapted 85 Td·s a-wave and b-wave increased with higher blood glucose (Table 2). Dark-adapted maximum rod sensitivity decreased with higher blood glucose, though not significantly (Supplemental Material).

Table 2.

Significant Correlations in the Diabetes Patients Between Specific Outcomes and Diabetes Duration (n = 69), Glycemia History (for Patients With Five, 10, or 15 Years of Diabetes Duration, n = 52) and Blood Glucose Before Examination (n = 62) From Simple and Multiple Linear Regression Analyses

		Coefficients From Simple Linear Regression					Multiple Linear Regression
Outcome*	Independent Factor	r 2	β	CI (2.5%; 97.5%)	t	P-Value†	Additional Independent Factor(s)	P-Value
Flicker 16 Td·s implicit time (ms)	Time-weighted average of previous HbA1c (mmol/mol)	0.13	0.056	0.018; 0.093	2.98	0.018	Blood glucose (mmol/L) (+ diabetes duration [years])	0.036 (0.079)
Flicker 32 Td·s implicit time (ms)	Time-weighted average of previous HbA1c (mmol/mol)	0.11	0.036	0.0094; 0.063	2.71	0.036	Blood glucose (mmol/L)	0.048
							Diabetes duration (years)	0.014
Dark-adapted 85 Td·s a-wave implicit time (ms)	Diabetes duration (years)	0.054	0.047	0.0044; 0.089	2.21	0.041	Blood glucose (mmol/L)	0.027
	Standard deviation of previous HbA1c (mmol/mol)	0.096	0.085	0.018; 0.15	2.53	0.041	Diabetes duration (years) and blood glucose (mmol/L)	0.021
	Blood glucose (mmol/L)	0.060	0.076	0.0074; 0.15	2.21	0.041	See above	—
Dark-adapted 85 Td·s b-wave implicit time (ms)	Blood glucose (mmol/L)	0.11	0.54	0.16; 0.91	2.86	0.023	Diabetes duration (years)	0.0099
Light-adapted flicker 85 Td·s implicit time (ms)	Diabetes duration (years)	0.091	0.032	0.0092; 0.055	2.8	0.027	Blood glucose (mmol/L)	0.0085
Capillary density (%) of temporal sector of SCP	Time-weighted average of previous HbA1c (mmol/mol)	0.24	−0.11	−0.16; −0.055	−4.10	0.0006	Diabetes duration (years)	0.00029

The simple linear regression analyses above have been repeated with the addition of the interaction term “diabetes duration subgroup” and the respective outcome. None of these interaction tests found significant interactions, and the outcomes all remained significant.

See Supplemental Material for all simple linear regression analyses of ERG and dark adaptometry outcomes, because only the significant correlations are listed above.

^*Implicit times were obtained by ERG and capillary density was obtained by OCTA.

^†Post hoc adjusted P-values by the Benjamini-Hochberg method.

For the patients with T1D duration > 4 years, ERG implicit times of flicker 16 Td·s and flicker 32 Td·s waves increased with higher previous HbA_1c levels. Furthermore, ERG implicit times of dark-adapted 85 Td·s a-wave increased with higher standard deviation of mean previous HbA_1c (Table 2). There was no significant correlation between previous HbA_1c levels and T1D duration (P = 0.34; r² = −0.0011; β = 0.046; t = 0.97; CI, −0.049 to 0.14).

Outcomes of the Diabetes Duration Subgroups Compared With the Healthy Control Group

ERG outcomes were largely comparable between T1D subgroups and healthy controls as assessed by post hoc analyses (Tables 3, 4). Dark adaptometry showed lower maximum rod sensitivity (full dark adaptation) only in the shortest and longest T1D duration subgroups, <5 months and 15 years, compared with the healthy subjects (post hoc analysis, Table 5). Rod intercepts were comparable between the T1D subgroups and the healthy subjects.

Table 3.

Implicit Times and Amplitudes of ERG Flicker Measurements in Dim Lighting

Outcomes, Mean ± SD (Post Hoc Adjusted P-Values*)	T1D Duration < 5 Months (n = 17)	T1D Duration ∼ 5 Years (n = 18)	T1D Duration ∼ 10 Years (n = 17)	T1D Duration ∼ 15 Years (n = 17)	Mean of T1D Patients (n = 69)	Healthy Subjects (n = 54)
Flicker 16 Td·s implicit time (ms)	28.3 ± 1.3 (1.00)	28.1 ± 1.7 (1.00)	27.3 ± 1.8 (0.051)	27.6 ± 1.6 (0.18)	27.8 ± 1.6	28.5 ± 1.5
Flicker 16 Td·s amplitude (µV)	30.5 ± 5.8 (1.00)	24.0 ± 7.5 (0.13)	22.7 ± 7.9 (0.054)	28.2 ± 6.9 (1.00)	26.4 ± 7.0	30.0 ± 10.9
Flicker 32 Td·s implicit time (ms)	26.0 ± 0.9 (1.00)	26.4 ± 1.2 (1.00)	25.9 ± 1.3 (0.89)	26.0 ± 1.1 (1.00)	26.1 ± 1.1	26.3 ± 1.3
Flicker 32 Td·s amplitude (µV)	33.7 ± 6.6 (1.00)	26.6 ± 6.2 (0.046)	27.1 ± 8.8 (0.42)	30.5 ± 8.4 (1.00)	29.5 ± 7.5	32.5 ± 12.7

Simple linear regression analyses of each of the above outcomes as a function of diabetes duration were not significant (Supplemental Material).

^*Bonferroni-adjusted P-values from Student's t-test or Wilcoxon (P-values have been multiplied by 4 as a correction for multiple comparisons).

Table 4.

Implicit times and amplitudes of ERG ISCEV's Five-Step Protocol Measurements After, Respectively, Dark and Light Adaptation

Outcomes, Mean ± SD (Post Hoc Adjusted P-Values*)	T1D Duration <5 Months (n = 17)	T1D Duration ∼5 Years (n = 18)	T1D Duration ∼10 Years (n = 17)	T1D Duration ∼15 Years (n = 17)	Mean of T1D Patients (n = 69)	Healthy Subjects (n = 54)
Dark-adapted 0.28 Td·s b-wave implicit time (ms)	95.9 ± 8.6 (1.00)	91.1 ± 12.1 (1.00)	100 ± 21.4 (1.00)	93.4 ± 12.1 (1.00)	95.0 ± 14.4	92.6 ± 13.4
Dark-adapted 0.28 Td·s b-wave amplitude (µV)	65.0 ± 18.9 (0.50)	49.5 ± 14.8 (0.35)	55.6 ± 14.0 (1.00)	57.2 ± 19.1 (1.00)	56.8 ± 16.7	57.3 ± 17.2
Dark-adapted 85 Td·s a-wave implicit time (ms)	14.3 ± 0.6 (1.00)	14.4 ± 0.7 (1.00)	14.6 ± 1.0 (1.00)	15.1 ± 1.4 (0.68)	14.6 ± 1.0	14.5 ± 0.7
Dark-adapted 85 Td·s a-wave amplitude (µV)	−56.1 ± 9.9 (0.94)	−44.0 ± 10.7 (0.064)	−48.1 ± 11.2 (0.98)	−49.1 ± 10.8 (1.00)	−49.3 ± 10.7	−52.1 ± 12.5
Dark-adapted 85 Td·s b-wave implicit time (ms)	47.8 ± 5.4 (1.00)	46.4 ± 5.5 (1.00)	49.2 ± 6.0 (1.00)	49.0 ± 6.2 (1.00)	48.0 ± 5.7	47.3 ± 5.8
Dark-adapted 85 Td·s b-wave amplitude (µV)	94.1 ± 22.3 (0.26)	73.2 ± 17.5 (0.59)	77.8 ± 17.6 (1.00)	82.0 ± 20.6 (1.00)	81.8 ± 19.5	82.0 ± 23.4
Dark-adapted total oscillatory potentials’ implicit time (ms)	154.0 ± 15.1 (0.65)	153.3 ± 17.7 (1.00)	154.5 ± 15.0 (0.64)	147.4 ± 18.3 (1.00)	152.3 ± 16.5	149.3 ± 16.0
Dark-adapted total oscillatory potentials’ amplitude (µV)	67.5 ± 19.2 (0.19)	49.1 ± 17.0 (0.85)	48.9 ± 14.9 (0.76)	59.2 ± 16.5 (1.00)	56.2 ± 16.9	56.1 ± 20.7
Light-adapted 85 Td·s a-wave implicit time (ms)	12.3 ± 1.0 (1.00)	12.1 ± 0.9 (1.00)	12.2 ± 1.1 (1.00)	11.8 ± 0.9 (0.22)	12.1 ± 1.0	12.3 ± 1.0
Light-adapted 85 Td·s a-wave amplitude (µV)	−9.4 ± 2.6 (1.00)	−7.6 ± 2.6 (0.32)	−8.1 ± 1.8 (0.27)	−8.3 ± 2.6 (1.00)	−8.4 ± 2.4	−9.4 ± 3.9
Light-adapted 85 Td·s b-wave implicit time (ms)	27.4 ± 0.9 (0.70)	28.2 ± 0.7 (0.65)	27.9 ± 0.7 (1.00)	28.0 ± 0.9 (1.00)	27.9 ± 0.8	27.8 ± 0.9
Light-adapted 85 Td·s b-wave amplitude (µV)	46.4 ± 12.3 (0.77)	35.4 ± 11.5 (0.40)	35.9 ± 7.4 (0.48)	39.5 ± 11.1 (1.00)	39.3 ± 10.6	41.4 ± 13.8
Light-adapted flicker 85 Td·s implicit time (ms)	24.3 ± 0.6 (1.00)	24.5 ± 0.7 (1.00)	24.5 ± 0.5 (1.00)	24.8 ± 0.4 (0.19)	24.5 ± 0.6	24.5 ± 0.6
Light-adapted flicker 85 Td·s amplitude (µV)	42.2 ± 7.7 (1.00)	33.5 ± 9.4 (0.28)	36.7 ± 6.5 (1.00)	38.8 ± 8.8 (1.00)	37.8 ± 8.1	39.2 ± 11.9

^*Bonferroni-adjusted P-values from Student's t-test or Wilcoxon (P-values have been multiplied by 4 as a correction for multiple comparisons).

Table 5.

Dark Adaptometry

Outcomes, Mean ± SD (Post Hoc Adjusted P-Values*)	T1D Duration <5 Months (n = 17)	T1D Duration ∼ 5 Years (n = 18)	T1D Duration ∼ 10 Years (n = 17)	T1D Duration ∼ 15 Years (n = 17)	Mean of T1D Patients (n = 69)	Healthy Subjects (n = 54)
Rod intercept (min)	9.76 ± 1.8 (1.00)	9.08 ± 2.0 (1.00)	9.20 ± 1.8 (1.00)	8.94 ± 1.5 (1.00)	9.24 ± 1.8	9.52 ± 2.2
Maximum rod sensitivity (log units)	4.4 ± 0.1 (0.035)	4.6 ± 0.2 (1.00)	4.6 ± 0.2 (1.00)	4.4 ± 0.1 (0.0040)*	4.5 ± 0.2	4.6 ± 0.2
Fixation error rate (%)	15.4 ± 11.7 (1.00)	18.6 ± 14.9 (1.00)	15.4 ± 8.7 (1.00)	10.2 ± 8.8 (0.074)	14.9 ± 11.5	17.5 ± 12.3

^*Bonferroni-adjusted P-values from Student's t-test or Wilcoxon (P-values have been multiplied by 4 as a correction for multiple comparisons).

Inner retinal layer thicknesses, including the inner limiting membrane, the retinal nerve fiber layer, the ganglion cell layer and the inner plexiform layer, were all comparable between each of the T1D duration subgroups and the healthy subjects (Table 6). Superficial capillary plexus densities were numerically lower in all T1D duration subgroups compared with the healthy subjects in all but one parafoveal ETDRS sector (the temporal sector of the <5 months-group), but reached statistically significantly lower density primarily in the 15-year duration subgroup (Table 7).

Table 6.

Retinal Thickness Segmented From the Inner Limiting Membrane to the Border Between the Inner Plexiform Layer and The Inner Nuclear Layer Corresponding to Nine Zones of an Overlayed ETDRS Grid Centered on the Fovea of the 12 × 9 mm OCT Scan

Outcomes, Mean ± SD (Post Hoc Adjusted P-Values*)	T1D Duration <5 Months (n = 17)	T1D duration ∼5 years (n = 18)	T1D Duration ∼10 Years (n = 17)	T1D Duration ∼15 Years (n = 17)	Mean of T1D Patients (n = 69)	Healthy Subjects (n = 54)
Foveal zone (µm)	57.4 ± 13.4 (1.00)	54.7 ± 9.7 (1.00)	61.1 ± 14 (0.48)	55.1 ± 11.2 (1.00)	57.1 ± 12.1	55.3 ± 9.2
Superior inner zone (µm)	124.8 ± 7.3 (1.00)	121.5 ± 7.9 (0.30)	128.1 ± 8.3 (1.00)	125.9 ± 7.0 (1.00)	125.0 ± 7.8	125.4 ± 8.0
Superior outer zone (µm)	106.6 ± 5.9 (0.84)	106.8 ± 8.9 (1.00)	109.8 ± 8.9 (1.00)	112.2 ± 9.4 (0.68)	108.9 ± 8.6	109.2 ± 7.4
Inferior inner zone (µm)	124.9 ± 6.8 (1.00)	123.1 ± 8.2 (0.52)	127.4 ± 8.8 (1.00)	125.9 ± 7.2 (1.00)	125.3 ± 7.8	126.2 ± 6.9
Inferior outer zone (µm)	105.3 ± 5 (0.12)	107.4 ± 8.6 (1.00)	108.5 ± 8.6 (1.00)	109.9 ± 9.3 (1.00)	107.8 ± 8.1	109.2 ± 7.9
Nasal inner zone (µm)	120.9 ± 7.9 (1.00)	119 ± 9.0 (0.72)	126.3 ± 9.0 (0.14)	121.3 ± 7.5 (1.00)	121.9 ± 8.6	121.8 ± 6.9
Nasal outer zone (µm)	124.4 ± 6.9 (1.00)	122.3 ± 8.7 (0.52)	127.8 ± 11.5 (1.00)	127.3 ± 10.4 (1.00)	125.4 ± 9.6	125.6 ± 8.1
Temporal inner zone (µm)	115.4 ± 6.5 (1.00)	111.6 ± 7.3 (0.48)	116.1 ± 8.4 (1.00)	113.6 ± 7.1 (1.00)	114.1 ± 7.4	114.3 ± 5.9
Temporal outer zone (µm)	93.9 ± 7.3 (0.84)	95.3 ± 6.7 (1.00)	98.2 ± 5.3 (1.00)	98.9 ± 8.0 (1.00)	96.6 ± 7.0	96.5 ± 7.4

^*Bonferroni-adjusted P-values from Student's t-test or Wilcoxon (P-values have been multiplied by 4 as a correction for multiple comparisons).

Table 7.

Retinal SCP Densities^* Corresponding to Four Parafoveal Zones of an Overlayed ETDRS Grid Centered on the Fovea of the 3 × 3 Mm OCTA Scan

Capillary Densities of SCP, Mean ± SD (Post Hoc Adjusted P-Values†)	T1D Duration <5 Months (n = 17)	T1D Duration ∼5 Years (n = 18)	T1D Duration ∼10 Years (n = 17)	T1D Duration ∼15 Years (n = 17)	Mean of T1D Patients (n = 69)	Healthy Subjects (n = 53)
Superior sector (%)	49.0 ± 2.0 (0.15)	48.6 ± 2.3 (0.028)	49.4 ± 2.7 (1.00)	47.2 ± 2.2 (16.8 × 10−7)	48.5 ± 2.4	50.1 ± 1.82
Nasal sector (%)	47.1 ± 1.9 (1.00)	47.2 ± 1.8 (1.00)	46.8 ± 1.9 (1.00)	45.1 ± 2.5 (0.0022)	46.5 ± 2.2	47.3 ± 1.48
Inferior sector (%)	49.4 ± 1.6 (1.00)	49.5 ± 1.8 (1.00)	47.1 ± 3.9 (0.011)	47.3 ± 3.5 (0.058)	48.3 ± 3.0	49.9 ± 2.21
Temporal sector (%)	47.8 ± 1.3 (1.00)	47.1 ± 2.1 (1.00)	47.3 ± 3.0 (1.00)	45.5 ± 2.3 (0.0064)	46.9 ± 2.4	47.6 ± 1.92

^*Shown in part in reference by Torm et al. 2025 (unpublished data).

^†Bonferroni-adjusted P-values from Student's t-test or Wilcoxon (P-values have been multiplied by 4 as a correction for multiple comparisons).

Correlation With Capillary Density

Of the ERG implicit times that showed significant correlation with either T1D duration (dark-adapted 85 Td·s a-wave and light-adapted 85 Td·s flicker wave) or previous HbA_1c (flicker 16 Td·s and flicker 32 Td·s wave), only that of the light-adapted 85 Td·s flicker wave was significantly correlated with capillary density of the temporal sector of the superficial capillary plexus (P = 0.022; r² = 0.062; β = −1.17; t = −2.34; CI, −2.17 to −0.17).

Discussion

In this cross-sectional analysis, a total of 69 patients with T1D duration from a few months to 17 years with mild or no fundus photographic signs of retinopathy and 54 age-matched healthy control subjects were examined using ERG, dark adaptometry, OCT and OCTA. This study found positive correlations between increased ERG implicit times of dark-adapted 85 Td·s a-wave or light-adapted 85 Td·s flicker wave and longer T1D duration, and of flicker 16 Td·s or flicker 32 Td·s waves and higher previous HbA_1c levels. These findings might suggest beginning mild retinal neural dysfunction. Additionally, there were capillary density deficits. Retinal nerve fiber and ganglion cell layer thickness did not differ between the diabetes duration subgroups and the control group.

Previous prospective studies using ERG found a reduced sum of oscillatory potential amplitudes in patients with T1D compared with age-matched healthy control subjects.^9,10 These data may reflect impaired activity in the bipolar, amacrine and ganglion cells of the retina.¹⁸ Both studies suggested that this measure could predict patients at risk of developing proliferative diabetic retinopathy,^9,10 (e.g., within six years given that T1D duration was more than five years, based on follow-up 6–8 years and 13–15 years after baseline).⁹ Regarding flicker ERG, one study found that only amplitudes of high-frequency flicker rates (62 or 100 Hz) were diminished in diabetes patients without retinopathy or with non-proliferative diabetic retinopathy compared with healthy subjects.¹⁹ Other studies^7,20 found reduced amplitudes in the pattern ERG of patients with less than one year of T1D compared with age-matched healthy control subjects, expressing slower electrical transmission in the inner retina,²⁰ but no differences in the amplitudes of the oscillatory potentials using regular or flash ERG. Similarly as in the present study, delayed implicit times on flicker ERG have been correlated with previous HbA_1c in adults with diabetes having non-proliferative retinopathy,²¹ and they were longer and the amplitudes diminished compared with healthy subjects.²¹ Also, increased implicit times on multifocal ERG²² have been observed in adult diabetes patients with no retinopathy.²³ In the present study, we found no significant reductions of the sum of oscillatory potential amplitudes, and the ERG technique used does not perform pattern or multifocal ERG.

In a previous study using the same handheld ERG device as in the present study, only the implicit times had acceptable intra-individual variabilities, whereas the amplitudes depended on placement of the electrode below the eye that could be difficult to standardize between individuals.²⁴ Another study similarly found that implicit times were comparable between the handheld ERG device and a standard ERG device, but that the amplitudes were two- to three fold higher using the standard ERG device.²⁵ The handheld ERG has though been shown to diagnose diabetic retinopathy with around 70% accuracy according to a receiver-operation characteristic curve.²⁶

Dark adaptation in diabetes patients has been sparsely examined using a dark adaptometer. One study of diabetes patients having either no retinopathy or increasing stages of retinopathy suggested that rod recovery rate was affected in early-stage retinopathy while cone sensitivity and rod intercept were only affected in proliferative diabetic retinopathy.¹⁷ Another study of diabetes patients with no retinopathy or non-proliferative diabetic retinopathy and healthy subjects found a borderline significant difference in dark adaptation speed between the three groups with a P-value of 0.07. Rod intercept in the nonproliferative diabetic retinopathy group was, on average, one minute slower than for the other two groups, but it was not correlated with HbA_1c or blood glucose levels.²⁷ A study of a group of diabetic patients with mainly no or mild diabetic retinopathy and a group of healthy subjects found no difference in rod intercept between the two groups.²⁸ In these studies, maximum rod sensitivity was not measured as the examination was interrupted after the rod intercept had been reached. Another method of measuring adaptation to dark, nyctometry, has been proposed to have clinical value in prognostication of proliferative diabetic retinopathy.^29,30

Early signs of peripheral and autonomic neuropathy have been shown to be highly prevalent in children and adolescents having similar T1D duration and glycemic history as in the present study.³¹ Furthermore, the portion of subjects with very mild or mild nonproliferative diabetic retinopathy (NPDR) in our study was similar to that in the TODAY³² and SEARCH¹ studies. Rosenthal et al.³³ recently emphasized the serious implications of early diabetic retinal disease in youth. Thus, the current study contributes to the understanding of this problem and the need to identify and treat it before vision is threatened. By using ERG, patients at risk of developing late complications may be identified earlier.

Strengths of our study include analysis of young patients with well-characterized histories of T1D under good long-term metabolic control for defined durations and the use of multiple tests of neural retinal function and microvascular structure. The patients were not having any clinically measurable visual dysfunction. The use of a handheld ERG system was also an advantage to facilitate examination of children. The sample size of the study is larger than or comparable to that of previous similar studies.^{19,21,22,25,27} Limitations of the study was its exploratory nature and small subgroups. The number of patients with mild or moderate non-proliferative diabetic retinopathy limited statistical analyses about the role of clinical retinopathy. The investigation of multiple correlations may increase the risk of Type 1 errors. Missing values may have introduced selection bias. The present cross-sectional baseline data cannot define if impairment of retinal neural tissue precedes microvasculature abnormalities.

Electrophysiological examination of the retina provides a wealth of numerical data that is not yet directly interpretable.³⁴ More research and longitudinal studies are needed to identify a narrow selection of quantitative measures, notably of early and potentially reversible subclinical abnormalities. Data from the present study may assist the generation of specific hypotheses and the selection of endpoints for use in such studies. Measures of neuronal processing speed are particularly attractive as they seem more reproducible than measures of amplitude.^24,25 Notably, we found a correlation between flicker ERG implicit times and HbA_1c history. The large interindividual variation prompts an early baseline examination after the onset of T1D. Longitudinal data will then enable analysis of change based on individual references and provide a potentially higher sensitivity to investigate if early alteration of retinal neural function may increase as capillary perfusion is decreasing. The hypothesis to be tested is that differences between diabetes patients and healthy controls increase in proportion to diabetes duration, mean glycemia and glycemia instability.