1 INTRODUCTION

Diabetes is one of the most common chronic metabolic conditions worldwide with estimated 463 million individuals affected in 2019 and an estimated 629 million people expected to have diabetes in another 25 years.¹ With the rising incidence of childhood-onset type 1 diabetes by approximately 3.4% yearly reported in a pooled analysis from Europe from 1989 to 2013,² more individuals are at risk for chronic macro- and microvascular complications of diabetes, including diabetic nephropathy, neuropathy, and diabetic retinopathy (DR).^{3, 4}

DR is one of the leading causes of blindness in developed countries and one of the most common complications of type 1 diabetes.^{5, 6} A very recent analysis of the Global Burden of Disease Study has estimated that approximately 0.9 million individuals aged 50 years and older have blindness and an additional 2.9 million have moderate or serious visual impairment due to DR⁷ with the numbers stabilizing over time compared to a previous report.⁸

Modifiable and other known risk factors for DR among individuals with diabetes include hyperglycemia, with HbA1c the most commonly used surrogate marker, elevated cholesterol level, elevated albumin excretion rate, and raised blood pressure.^9-12

Younger age at diagnosis may contribute to children and young people experiencing an increased risk of developing long-term complications of diabetes, including DR and diabetes-related visual impairment before adulthood¹³; this risk, however, can be decreased substantially with a precise glycemic control.¹⁴

While there have been many studies in adults, current data regarding the prevalence of DR among children and adolescents with type 1 diabetes are scarce.^15-18

The aim of this multinational study was to estimate the prevalence of DR in children and adolescents with type 1 diabetes from several large prospective longitudinal registries and to evaluate the risk factors associated with the occurrence of DR in our study population.

2 METHODS

3 RESULTS

3.2 Multivariable analysis of DR prevalence

The adjusted odds ratio of DR decreased in the observational period (aOR_{per-1-year-increase} = 0.99, 95% CI: 0.99–1.00, p < 0.0001) after adjusting for duration of type 1 diabetes, gender, minority status, and age over time (Table 2, Figure 1). This change was driven by the improvement observed in England aOR_{per-1-year-increase} = 0.95, 95% CI: 0.94–0.97, p < 0.0001), Germany (aOR_{per-1-year-increase} = 0.92, 95% CI: 0.90–0.93, p < 0.0001), and Wales (aOR_{per-1-year-increase} = 0.84, 95% CI: 0.79–0.90, p < 0.0001). The adjusted odds ratio of DR increased with the duration of type 1 diabetes (aOR_{per-1-year-increase} = 1.04, 95% CI: 1.03–1.04, p < 0.0001). A significantly higher probability of DR was observed in females in England and Germany (p < 0.0001 and p = 0.0219, respectively), while gender-related differences in the other reported countries did not reach statistical significance (Table 2).

TABLE 2. Adjusted estimates for the risk of diabetic retinopathy, depending on unmodifiable risk factors, overall and by country

	No. of participants	aORa (95% CI)	p value	aORa (95% CI)	p value	aORa (95% CI)	p value	aORa (95% CI)	p value
		Year of the most recent eye exam		Age at exam		Duration of T1D		Gender female versus male
All	99,066	0.99 (0.98–1.00)b	0.0093	1.00 (1.00–1.01)	0.2729	1.04 (1.03–1.04)	<0.0001	1.04 (1.00–1.08)	0.0556
Australia	801	1.06 (0.95–1.19)	0.2970	1.29 (1.16–1.45)	<0.0001	1.20 (1.11–1.29)	<0.0001	1.37 (0.88–2.15)	0.1668
Austria	3407	1.01 (0.89–1.15)	0.8061	1.18 (0.95–1.46)	0.1418	1.03 (0.90–1.18)	0.6361	0.98 (0.33–2.94)	0.9755
Denmark	2283	1.00 (0.94–1.07)	0.8849	1.24 (1.04–1.49)	0.0174	1.22 (1.12–1.32)	<0.0001	1.05 (0.56–1.96)	0.8846
England	26,044	0.95 (0.94–0.97)	<0.0001	1.19 (1.16–1.21)	<0.0001	1.16 (1.15–1.17)	<0.0001	1.25 (1.17–1.34)	<0.0001
Germany	60,901	0.92 (0.90–0.93)	<0.0001	1.18 (1.13–1.24)	<0.0001	1.16 (1.13–1.19)	<0.0001	1.25 (1.03–1.52)	0.0219
The Netherlands	721	1.07 (0.65–1.77)	0.7789	0.89 (0.64–1.23)	0.4767	1.29 (1.00–1.65)	0.0484	1.46 (0.32–6.74)	0.6240
United States SEARCH	1709	1.09 (0.94–1.27)	0.2494	1.12 (1.07–1.17)	<0.0001	1.28 (1.19–1.38)	<0.0001	1.14 (0.87–1.49)	0.3390
Wales	1600	0.84 (0.79–0.90)	<0.0001	1.11 (0.99–1.25)	0.0655	1.24 (1.18–1.29)	<0.0001	1.37 (0.98–1.91)	0.0660

• Note: Odds ratios and p values are derived from multiple logistic regression analyses, adjusted for the year of the most recent eye examination, age, gender, type 1 diabetes duration, minority status, and stratified by country.
• ^a aOR: Adjusted odds ratios for binary predictor variable (gender) or expected changes in odds per one unit increase in year with 95% Confidence Intervals (95% CI).
• ^b The effect of the year of examination for all countries together is based on data from Australia, Austria, Denmark, England, Germany, and Wales for 2011–2019 (No. of included participants 73,852).

In an exploratory analysis evaluating for possible modifiable risk factors associated with DR (Table 3), the probability of DR increased with higher HbA1c (aOR for 1 mmol/mol increase in HbA1c = 1.03, 95% CI: 1.03–1.03, p < 0.0001). A strong magnitude of association was observed for hypertension (aOR = 1.24, 95% CI: 1.12–1.38, p < 0.0001). Higher diastolic (p < 0.0001) but not systolic (0.2497) blood pressure was associated with a higher probability of DR. Smokers had on average 1.30 times higher probability of DR compared to non-smokers (95% CI: 1.17–1.44, p < 0.0001), while there was no association between cholesterol (both total cholesterol and LDL) levels and DR.

TABLE 3. Adjusted estimates for the risk of diabetic retinopathy depending on modifiable risk factors

Risk factor	aORa (95% CI)	p value
HbA1c (mmol/mol) N = 96,089	1.03 (1.03–1.03)	<0.0001
Hypertension >140/90 N = 99,075	1.24 (1.12–1.38)	<0.0001
Systolic BP N = 92,781	1.00 (1.00–1.00)	0.4092
Diastolic BP N = 92,727	1.01 (1.01–1.02)	<0.0001
LDL-cholesterol (mg/dl) N = 46,280	1.00 (1.00–1.00)	0.2722
Total cholesterol (mg/dl) N = 68,799	1.00 (1.00–1.00)	0.2316
Smoking N = 73,451	1.30 (1.17–1.44)	<0.0001

• Note: Odds ratios and p values are derived from multiple logistic regression analyses (one for each risk factor) adjusted for the calendar year of the most recent eye examination, age, gender, type 1 diabetes duration, and minority status. Only complete case analyses were performed and participants with missing data were excluded from the regression models.
• Abbreviations: BP, blood pressure; LDL, low-density lipoproteins.
• ^a aOR: Adjusted odds ratios for binary predictor variables or estimated changes in odds per one unit increase for numerical predictor variables with 95% Confidence Interval (95% CI). N: Number of observations used. England and Wales did not document data regarding LDL and Australia and Italy Marche did not document data regarding smoking status.

4 DISCUSSION

To our knowledge, this is the largest investigation of DR prevalence among children, adolescents and young adults with type 1 diabetes conducted in recent years. In this cross-sectional study of more than 146 thousand young individuals with type 1 diabetes, 5.8% of them were identified with any retinopathy and 0.07% with severe retinopathy. The prevalence of DR varied between different registries. There was almost a 10-fold difference (excluding Luxemburg with none) in reported retinopathy between the countries with the highest prevalence (Australia, England, United States, and Wales) and the other five countries, which is in line with previous recent reports.^{15, 18} The key finding of this study is that there was a significant decline in odds for DR over time equating to 1% per year. Importantly, two of the registries with the highest DR rate (England and Wales) observed a significant improvement during the observational period.

Despite the ongoing effort, overall only a minority of youths achieved recommended target HbA1c level below 53 mmol/mol (7.0%), which is in line with previous reports.^{21, 22} Notably, in recent years, several studies reported improved glycemic control with better access to diabetes technology.^{23, 24} More effort is needed to further improve metabolic control and treatment of comorbidities to prevent or delay DR in this high-risk population.

There were some additional distinctions in participant characteristics between registries. Italy, the Netherlands and the United States (SEARCH) reported less than 1% of hypertension, while Austria, England, and Germany reported more than 5%, notwithstanding international guidelines recommending early hypertension management in individuals with type 1 diabetes.^{3, 20} Individuals with type 1 diabetes in this age group are frequently undertreated for hypertension. Recently, Shah and coworkers reported that only up to one out of three individuals with hypertension received antihypertensive medications.²⁵ There was less discrepancy in total cholesterol or LDL values between different registries, while Austria, Germany, Luxembourg, and the United States (SEARCH) reported higher smoking prevalence compared to other countries.

Similar to previous studies,^{9, 16, 26-28} the odds of DR increased with type 1 diabetes duration. On average, for every year living with type 1 diabetes, the odds of DR increased 1.04 times. In contrast to some previous reports,^{5, 26} in two largest datasets (Germany and England) the female gender was associated with a higher odd ratio of DR in our analysis, while it failed to reach statistical significance in other countries. As women with type 1 diabetes have a higher risk of cardiovascular disease, this cannot be fully explained by the gender differences in glycemic control and body mass index.^{25, 29}

To alleviate the risk of developing DR, it is crucial to reduce exposure to hyperglycemia as well as mitigate pronounced glucose fluctuations at the same time. Indeed, a systematic review demonstrated HbA1c variability was associated with DR.³⁰ However, glycemic excursions are not necessarily captured with HbA1c and several recent trials have demonstrated that time in range is strongly associated with the risk of microvascular complications and could complement HbA1c as an outcome measure.^{31, 32} Moreover, it is critical to optimize blood pressure screening and management. Blood pressure should be measured in accordance with established guidelines at every routine clinical visit by a trained individual, including confirmation with multiple readings and measurements on a separate day and home monitoring or 24-h ambulatory blood pressure monitoring. In addition to lifestyle management, individuals should be promptly started on pharmacological therapy to achieve recommended blood pressure targets when hypertension is confirmed.^{25, 33} While early treatment with ACE inhibitors has not slowed the progression of retinopathy in youth with type 1 diabetes, previous reports have demonstrated efficacy in adults even in individuals without hypertension.³⁴

Smoking and the use of other tobacco products have been associated independently with an increased risk of DR in individuals with type 1 diabetes.³⁵ In our study cohort, reported smoking rates were almost 10% and we observed 1.3 times higher odds for developing DR. It is imperative to screen for smoking in youth with type 1 diabetes and to address smoking avoidance or cessation during regular clinical visits.³⁶

The main limitations of our study are that the approach to screening for DR was not standardized across different countries, heterogeneity in the size of registries, and not all registries collected data for the whole observational period. Data regarding diabetes technology use was not collected. Other limitations include observational cross-sectional data as not all registries collected longitudinal information with consecutive ophthalmological examinations for each individual, and possible overestimation of the effect of age and type 1 diabetes duration, when using only the longest observation period for each participant. Two countries (England and Wales) did not document LDL and two countries (Australia and Italy) did not document information regarding the smoking status. Registries have reported only a single blood pressure measurement at each visit.

The strengths of this study include a large cohort of young individuals with type 1 diabetes from different multi-continental regions with detailed laboratory measures over a long period of time. Noteworthy, DR was confirmed by ophthalmological examination; either slit lamp or retinal photography. Standardized screening for early signs of retinal damage is imperative for early intervention, along with improvement of modifiable risk factors. Current international guidelines recommend that screening in individuals with T1D is initiated within 3–5 years of diagnosis, with follow-up exams every 1–2 years thereafter.⁴ Nevertheless, data show that only 35%–72% of youth with diabetes undergo recommended ophthalmological exams in accordance with clinical practice guidelines.³⁷ Minority youths with diabetes are even less likely to have regular eye examinations.³⁸ To reduce attrition of individuals with type 1 diabetes from DR screening programs and possibly to reduce costs, fully autonomous artificial intelligence-based systems for DR detection could be instrumental.³⁹

In conclusion, the prevalence of diabetic retinopathy in this large cohort of young individuals with type 1 diabetes varied among high-income countries, increased with longer diabetes duration, decreased over time, and was associated with HbA1c, hypertension, and smoking.

AUTHOR CONTRIBUTIONS

All authors contributed to the study design, data collection, drafting, revising, and approval of the final version of the manuscript. Marie Auzanneau and Reinhard W. Holl researched data, were responsible for the statistical analysis and take responsibility for the integrity of the data and accuracy of the data analysis. Natasa Bratina, Reinhard W. Holl, Klemen Dovc, and Kim C. Donaghue researched data and wrote the first draft of the manuscript. Natasa Bratina and Reinhard W. Holl had a full access to all the analyzed data in the study and are guarantors of this work. All data owners have given permission for publication.

CONFLICT OF INTEREST

The authors declare that there are no conflicts of interests related to this study.