Prevalence, Progression, and Modifiable Risk Factors for Diabetic Retinopathy in Youth and Young Adults With Youth-Onset Type 1 and Type 2 Diabetes: The SEARCH for Diabetes in Youth Study Free

Diabetic retinopathy (DR) is a common and serious sight-threatening complication of diabetes and the leading cause of blindness in adults in the U.S. (1). After 20 years of diabetes duration, nearly 99% and 60% of people with type 1 and type 2 diabetes, respectively, demonstrate some degree of DR (2). Among adults, known risk factors for development of DR include diabetes duration and severity, adequacy of diabetes management, and demographic factors (3–5).

There is a paucity of data to characterize the epidemiology of DR among youth and young adults (YYAs) in the U.S. and globally, particularly in those with youth-onset type 2 diabetes. Studies are limited by either a selected population or reliance on clinical data where screening practices likely influence findings. Youth-onset diabetes is becoming increasingly common in the U.S., particularly among youth from traditionally marginalized racial and ethnic groups who may be at higher risk of DR (6,7). Characterization of the epidemiology of DR can increase understanding of the public health burden and, ultimately, inform DR screening guidelines. Understanding which individuals may be at greatest risk can inform risk stratification and clinical management of YYAs with youth-onset diabetes. In the current study, data from the SEARCH for Diabetes in Youth (SEARCH) study were used to characterize the epidemiology of DR among YYAs at two follow-up points and to evaluate modifiable risk factors for the development and progression of DR.

The SEARCH study has been described in detail elsewhere (8). Briefly, it comprised a registry and longitudinal cohort component that included individuals with incident youth-onset type 1 or type 2 diabetes identified and recruited for an initial baseline visit and follow-up visits within six areas in the U.S., including four geographic-based sites in Ohio; Colorado; five counties in the Seattle, Washington, area; South Carolina, health plan enrollees from Kaiser Permanente Southern California (seven counties); and American Indian reservation–based populations in Arizona and New Mexico (8). The longitudinal cohort was recruited from patients with incident diabetes registered in the years of 2002–2006, 2008, and 2012. By design, a subset of registered patients was invited for in-person baseline and follow-up visits, with oversampling of youth from disproportionately affected groups with type 1 diabetes at the most recent follow-up. The cohort for this analysis was participants who underwent retinal imaging, which was offered at a clinical assessment visit during the third (2010–2014) and fourth (2015–2019) phases of the SEARCH study conducted at a mean diabetes duration of 7.5 and 12.5 years, respectively. Baseline assessments occurred within a mean (SD) of 9.6 (6.7) months from diagnosis for participants with type 1 diabetes and 12.1 (7.9) months from diagnosis for participants with type 2 diabetes. Study protocols were approved by each site’s institutional review board, and all participants provided informed, written consent or assent.

Retinal images were obtained across all SEARCH study sites, using the same nonmydriatic camera model (Visucam Pro N; Carl Zeiss Meditech) and a standardized protocol. DR was determined by grading 45° color digital fundus images centered on the disc and macula of both eyes. The Wisconsin Ocular Epidemiology Reading Center graded fundus images masked to clinical characteristics. Images were graded by at least two graders (preliminary grader and detail grader), with an additional adjudicator reviewing images for final determination on retinopathy severity where needed. DR categories were defined using the Airlie House/modified ETDRS classification system for denoting presence or absence of DR and severity grade of DR (10–13 indicates no DR; 14–31, mild, nonproliferative DR; 41–51, moderate nonproliferative DR; and 60–80, proliferative DR) (9–11). Retinopathy severity grade was based on the worse eye. For estimating prevalence and modifiable risk factors for developing DR, retinopathy was considered present for those with mild or moderate nonproliferative or proliferative severity grades.

Type 1 or type 2 diabetes was defined according to clinician’s documentation of diabetes type within 6 months of diagnosis. Age at diagnosis was calculated based on dates of birth and diagnosis. Other demographic factors, including health insurance status, highest education attained by either parent, race, ethnicity, and sex, were obtained through questionnaires administered at the time of enrollment. Height, weight, waist circumference, hemoglobin A_1c (A1C), and systolic and diastolic blood pressure were measured at in-person study visits using standardized study protocols. BMI z scores were calculated using age- and sex-specific Centers for Disease Control and Prevention growth references (12). Blood pressure z scores were calculated by comparing the observed blood pressure with age-, sex-, and height-adjusted externally standardized measures. For BMI and blood pressure z scores, if participants were age >20 years at the clinical visit, the age 19.999 years was imputed for calculating a z score. A1C area under the curve (AUC) measures were calculated as time-weighted averages from baseline to first image (for assessment of modifiable factors at first image) and then from baseline to second image (for assessment of modifiable factors at second image).

The analysis cohort consisted of participants with at least one retinal image of either eye at a follow-up visit, with a subset having a retinal image obtained at a second follow-up visit. Demographic characteristics, year of diagnosis, and enrollment site were summarized by presence of a first and second retinal image to compare the demographic distribution of all registered participants with incident diabetes to the analysis cohort who underwent screening. This comparison explored the potential for selection bias of having obtained a retinal image and the generalizability of the analytic sample to the population registered by the SEARCH study.

We calculated crude prevalence of DR and the 95% CI according to demographic and clinical attributes. To determine the prevalence of DR standardized to the demographic distribution of registered participants, we first calculated overall prevalence of DR at each follow-up visit in the sample and then standardized these prevalence estimates for the sex and race/ethnicity distribution in the population using Rothman’s Episheet (13). Standardized prevalence was estimated separately by diabetes type and reported for all grades of DR combined. Comparison of the difference in DR prevalence, according to diabetes type and at first and second imaging, was assessed using a two-sample binomial test.

To evaluate progression among participants with retinal images from both follow-up visits, we cross tabulated DR grade at first and second imaging and defined progression as evidence of any worsening of DR grade (e.g., from none to mild, moderate, or proliferative; from mild to moderate or proliferative; from moderate to proliferative for the worse eye).

We examined potentially modifiable factors, including A1C (at baseline and follow-up), BMI (baseline, follow-up, and by change in BMI z score from baseline to follow-up), waist/height ratio (baseline and by change from baseline to first/second follow-up), and systolic and diastolic blood pressure z scores (baseline and by change from baseline to first/second follow-up), in relation to risk of DR at ∼7.5 and ∼12.5 years after diabetes diagnosis by diabetes type and DR grade. Associations between modifiable risk factors and DR progression were estimated by calculating relative risks from generalized linear models. Crude and adjusted models were used with and without inverse weighting for probability of selection to account for loss to follow-up and the oversampling of disproportionately affected people for type 1 diabetes (14,15) (unweighted estimates used in Supplementary Tables 1–9). Weighting was based on calculation of the probability of having completed the first and second follow-up visit with retinal imaging as a function of age, sex, race/ethnicity, year of diagnosis, and study site within each diabetes type. All adjusted models were informed by directed acyclic graphs, an approach that includes development of a causal model framework for informing selection of potential confounders for adjustment, and mitigating risk of inappropriate adjustment for nonconfounders, mediators, and factors that could induce bias as a result of adjustment (16–18). All analyses were completed using R version 4.0.2 statistical software.

The data set analyzed in the current study is available in the National Institute of Diabetes and Digestive and Kidney Diseases repository (https://repository.niddk.nih.gov/studies/search/).

There were 9,374 participants registered in incident years 2002–2006, 2008, or 2012; 7,277 had type 1 diabetes, 1,919 had type 2 diabetes, and 178 had missing or unknown diabetes type. Of the 2,519 participants with type 1 diabetes who had a first retinal image obtained, 1,014 had a second retinal image. Of the 447 participants with type 2 diabetes who had a first retinal image, 239 had a second retinal image. Nearly all participants (97%) with retinal imagining had gradable images for both eyes. The total number of participants with DR at each follow-up period, by grade and diabetes type, is provided in Supplementary Fig. 1. The distribution of demographic characteristics of registered participants with incident diabetes of known type and the subset of participants in the SEARCH cohort study who underwent retinal imaging at the first or second in-person visit are presented in Supplementary Table 1.

For type 1 diabetes (Table 1), the highest proportion of participants with DR at both the first and second retinal imaging were those reporting non-Hispanic Black race/ethnicity, those whose parents (both parents) had a high school or less education, those with elevated diastolic blood pressure, and those without health insurance. For type 2 diabetes (Table 2), the highest proportion of participants with DR were observed for those reporting non-Hispanic Black race/ethnicity (at the second time point only), with elevated blood pressure (both time points), and with normal/underweight BMI (<25.0 kg/m²) (at the second time point only). Demographic factors used to estimate the standardized prevalence of DR for the overall population of registered participants at first imaging and second imaging are presented in Supplementary Table 2A and B. At first retinal imaging (mean [SD] 7.7 [2.0] and 7.5 [2.2] years from diagnosis for type 1 and type 2 diabetes, respectively), the standardized DR prevalence was estimated to be 21.3% (95% CI 19.5%, 23.0%) for those with type 1 diabetes and 31.3% (95% CI 26.8%, 35.9%) for those with type 2 diabetes. For the subset with a second retinal image (mean [SD] 12.5 [2.2] and 12.4 [2.1] years from diagnosis for type 1 and type 2 diabetes, respectively), the standardized DR prevalence was 52.0% (95% CI 48.6%, 55.3%) for type 1 diabetes and 55.7% (95% CI 48.7%, 62.7%) for type 2 diabetes. While the prevalence of DR was significantly different between those with type 1 versus type 2 diabetes at first imaging (P < 0.0001), there was no significant difference by second imaging (P = 0.51).

Retinopathy progression from first to second imaging for type 1 diabetes (Supplementary Table 3A) and type 2 diabetes (Supplementary Table 3B) was cross tabulated by retinopathy grade. For type 1 diabetes, at a mean (SD) of 4.6 (1.1) years from initial retinal imaging, 368 of 949 (38.8%) with retinal images at both time points and without proliferative retinopathy at first imaging progressed to a more severe grade. For type 2 diabetes, at a mean (SD) of 4.4 (1.0) years from initial retinal imaging, 93 of 225 (41.3%) progressed to a more severe grade. Just 44 of 1,014 (4.3%) YYAs with type 1 diabetes experienced regression of their DR. For type 2 diabetes, 12 of 239 (5.0%) regressed.

Among participants with type 1 diabetes, the associations between modifiable risk factors and DR by first and second retinal imaging are shown for weighted unadjusted and adjusted models (Table 3). For the first retinal image, baseline A1C was positively associated with DR (adjusted risk ratio [aRR] 1.10; 95% CI 1.05, 1.15), as was A1C AUC (aRR 1.22; 95% CI 1.15, 1.28), obesity (BMI ≥30.0 kg/m²) (aRR 1.26; 95% CI 1.03, 1.54), and increase in waist/height ratio (aRR 1.20; 95% CI 1.06, 1.37). Change in systolic blood pressure z score was inversely associated with DR (aRR 0.86; 95% CI 0.79, 0.94). Among youth with type 1 diabetes at second imaging, change in BMI z score was marginally inversely associated with retinopathy (aRR 0.96; 95% CI 0.92, 1.00). The mean (SD) BMI z scores at baseline, first imaging, and second imaging for those with type 1 diabetes were 0.52 (1.06), 0.59 (0.95), and 0.69 (1.01), respectively. Baseline A1C was positively associated with DR (aRR 1.04; 95% CI 1.02, 1.07), as was A1C AUC (aRR 1.15; 95% CI 1.11, 1.19), BMI z score at baseline (aRR 1.05; 95% CI 1.00, 1.11), change in systolic blood pressure z score (aRR 1.05; 95% CI 1.01, 1.09), and change in diastolic blood pressure z score (aRR 1.08; 95% CI 1.02, 1.15). Unweighted analyses generally showed similar results (Supplementary Table 4).

Among participants with type 2 diabetes, baseline A1C was positively associated with DR at first retinal image (aRR 1.16; 95% CI 1.10, 1.21) and second retinal imaging (aRR 1.07; 95% CI 1.04, 1.11), as was A1C AUC at first imaging (aRR 1.19; 95% CI 1.10, 1.29) and second imaging (aRR 1.13; 95% CI 1.09, 1.18). The mean (SD) BMI z score at baseline, first imaging, and second imaging for those with type 2 diabetes was 2.15 (0.60), 1.84 (0.81), and 1.83 (0.67), respectively. At first imaging, change in BMI z score (aRR 0.74; 95% CI 0.65, 0.85) and change in waist/height ratio (aRR 0.89; 95% CI 0.85, 0.93) were inversely associated with retinopathy. At second imaging, change in BMI z score was inversely associated with retinopathy (aRR 0.76; 95% CI 0.67, 0.87) (Table 4). Both systolic and diastolic blood pressure z score changes were associated with DR at second imaging only (aRR 1.13 [95% CI 1.03, 1.23] and 1.23 [95% CI 1.07, 1.41], respectively). Unweighted analyses generally showed similar results (Supplementary Table 5).

Associations between modifiable risk factors and retinopathy progression from first to second retinal imaging are shown in Supplementary Tables 6–9. In weighted analyses, among participants with type 1 and type 2 diabetes, A1C AUC was significantly associated with progression (aRR 1.44 [95% CI 1.26, 1.64] and 2.14 [95% CI 1.53, 2.97], respectively). For those with type 2 diabetes, a consistent state of being overweight at both imaging time points was inversely associated with risk of DR (aRR 0.51; 95% CI 0.32, 0.79). No other statistically significant associations were observed, and unweighted results were similar to weighted results.

The standardized prevalence of DR at 7.7 (2.0) years postdiagnosis was 21% among participants with type 1 diabetes. By 12.5 (2.2) years postdiagnosis, the prevalence more than doubled to 52%. For participants with type 2 diabetes, at 7.5 (2.2) years postdiagnosis, the standardized prevalence of DR was 31.0%, increasing to 55.7% at 12.4 (2.1) years postdiagnosis. Among those with type 1 diabetes, 38.8% progressed to a more severe grade of DR at a mean of just 4.6 (1.1) years from initial retinal imaging; for individuals with type 2 diabetes, 41.3% progressed at a mean of 4.4 (1.0) years. Reassuringly, just 0.8% of those with type 1 and 3.1% of those with type 2 diabetes had evidence of proliferative DR by ∼12.5 years from diagnosis. Still, extrapolating this finding to the estimated 187,000 youths with type 1 diabetes and 23,000 youths with type 2 diabetes in the U.S. (7,19) this translates to an estimated burden of 2,393 cases of proliferative DR for type 1 and type 2 diabetes combined in just over a decade from disease onset. For all grades of retinopathy combined, this finding translates to 110,051 cases.

A recent analysis from the Treatment Options for Type 2 Diabetes in Adolescents and Youth (TODAY) study reported a prevalence of 49% at a mean (SD) 12.0 (1.5) years diabetes duration among a postclinical trial cohort of individuals with youth-onset type 2 diabetes (20). In a pilot study (n = 265) conducted by the SEARCH study in 2009–2010, the prevalence of any DR was 17% for type 1 diabetes (n = 222) and 42% for type 2 diabetes (n = 43) at ∼7 years from diabetes diagnosis (21). A more recent assessment of diabetes-associated complications in SEARCH reported an age-adjusted prevalence of DR of 5.2% in type 1 diabetes and 9.1% in type 2 diabetes at the initial follow-up, but this only included cases of moderate and severe DR (22). In a claims-based assessment of incident DR in individuals age <21 years (n = 2,240 with type 1 diabetes, n = 1,768 with type 2 diabetes) from a managed care network, a higher proportion of youth with type 1 diabetes was observed to have a diagnosis of DR (31.2% vs. 10.3% with type 2 diabetes) at ∼8 years follow-up (23). One limitation of this assessment is the potential selection bias introduced when inclusion depends on having been screened by an ophthalmologist or optometrist; thus, the retinopathy status of those not screened is unknown. Among individuals with type 2 diabetes, those living in high-income households had a lower incidence of DR (23). If lower-income families were disproportionately underrepresented in the sample, DR incidence would be underestimated.

Our assessment of potentially modifiable risk factors suggests that elevated A1C, systolic and diastolic blood pressure, and higher BMI are associated with DR for YYA with type 1 diabetes. For YYA with type 2 diabetes, A1C and systolic blood pressure were associated with DR. These findings suggest that risk increases when management of A1C and blood pressure are suboptimal and that elevated blood pressure may increase risk. Blood pressure was associated with DR even with adjusting for A1C, suggesting that blood pressure management, independent of glycemic control, could further reduce risk for DR. Interestingly, a decrease in BMI z score across follow-up was associated with an increased risk of DR, particularly for type 2 diabetes. This association has been reported elsewhere in adults with type 1 diabetes and in type 2 diabetes through the TODAY study (24), although the evidence to date is mixed (25). One potential mechanism for this seemingly paradoxical association is the possibility that without insulin treatment, BMI z score may decrease with worsening disease. The change in BMI z score could reflect poorer management of diabetes and, thus, increased risk for DR. In a post hoc assessment, BMI z score change was inversely correlated with A1C AUC (Supplementary Fig. 2). Poor glycemic control, as evidenced by higher A1C levels, was positively associated with progression of DR for both type 1 and type 2 diabetes, consistent with evidence presented from the TODAY study for type 2 diabetes (20). Still, in the TODAY study cohort, no obesity-associated biomarkers were associated with DR (24). Of note, in the current study, and in contrast to BMI z score, waist/height ratio was positively associated with DR at first imaging for type 1 diabetes. These differences highlight the potential that fat mass distribution has a more consequential role in the development of DR, with higher central adiposity, as opposed to overall fat mass, increasing risk. Efforts to elucidate the risk of complications in diabetes have identified differences in risk profile according to varying fat mass indices (26–28).

While we have standardized the prevalence estimates to the demographic distribution of the registered cases of youth-onset diabetes, these reflect the cases from the six SEARCH catchment areas and, thus, may not be representative of all youth with type 1 and type 2 diabetes in the U.S. It is possible too that there are factors that we were not able to standardize that explain some of the differences in DR prevalence observed across the two imaging periods. Furthermore, while we have applied inverse probability weighting to account for differential loss to follow-up, this method is predicated on having adequately modeled loss to follow-up. Our assessment of modifiable factors, including change across follow-up, is a relatively crude assessment of change. More frequent assessments of these measures would offer greater flexibility in modeling variability across time in the management of diabetes and the potential implications for DR. More frequent assessments could also provide the ability to more clearly assess the role of BMI and whether A1C confounds or acts as a mediator between BMI and DR. With only 4.3% of participants with type 1 diabetes and 5.0% of those with type 2 diabetes experiencing regression of their retinopathy grade, data were too sparse for assessing potential factors contributing to regression. Finally, while we performed screening for retinopathy, the cases identified necessitate a complete eye examination for confirmation of diagnosis and clinical treatment.

Current American Diabetes Association recommendations for screening in youth with type 1 diabetes include an initial dilated and comprehensive eye examination at age 10 or after puberty, whichever is earlier, after 3–5 years of diabetes duration. For youth with type 2 diabetes, screening is recommended at the time of diagnosis. For both types, following the initial examination, eye examinations are recommended every 1–2 years, depending on the presence of risk factors and evidence of any early ocular changes consistent with DR (29,30). Still, the uptake of eye examinations in patients with diabetes is likely low, with an estimated 15.5% of those with type 2 diabetes and 26.3% of those with type 1 among a commercially insured population receiving examinations consistent with the American Diabetes Association guidelines (31). Furthermore, youth from underrepresented racial and ethnic groups experience significantly lower levels of screening than recommended (32). The high prevalence of DR observed for YYAs with type 1 and type 2 diabetes suggests that universal screening of YYAs with diabetes and increased access to screening services across marginalized populations (e.g., uninsured) early in disease progression could aid in early detection. These results also suggest that while management of diabetes is a key factor in mitigating DR development, specific risk factors for DR progression in YYAs remain largely unknown. Given the high prevalence observed, and potential for long-term serious consequences to sight and for quality of life resulting from DR, additional research aimed at increasing understanding of pathophysiologic mechanisms for mitigating development and progression of DR in youth-onset diabetes may be warranted.

Funding. The SEARCH study was funded by the National Institute of Diabetes and Digestive and Kidney Diseases (grants 1R01DK127208-01 and 1UC4DK108173) and supported by the Centers for Disease Control and Prevention. The Population Based Registry of Diabetes in Youth Study was funded by the Centers for Disease Control and Prevention (DP-15-002) and supported by the National Institute of Diabetes and Digestive and Kidney Diseases (grants 1U18DP006131, U18DP006133, U18DP006134, U18DP006136, U18DP006138, and U18DP006139).

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention or the National Institute of Diabetes and Digestive and Kidney Diseases.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. E.T.J. and J.R. planned the study analyses. E.T.J., J.R., and J.M.L. conceived the study objectives. J.R. executed the study analyses. J.S. and E.A.L. provided input on study analyses and critically reviewed drafts of the manuscript. K.A.R., D.D., L.M.D., R.D., B.K., S.M., M.T.M., K.R., S.M.M., A.M., and B.M.-D. provided critical review of drafts of the manuscript. All authors provided approval of the final manuscript. E.T.J. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.