Prevalence of Diabetic Retinopathy in the United States, 2005-2008 FREE

Diabetic retinopathy is the leading cause of new cases of legal blindness among adults aged 20 to 74 years in the United States.¹ Vision loss due to diabetic retinopathy occurs through a variety of mechanisms, including retinal detachment, preretinal or vitreous hemorrhage, associated neovascular glaucoma, and macular edema or capillary nonperfusion.² The presence of diabetic retinopathy may indicate microcirculatory dysfunction in other organ systems.^3- 4 Diabetes-related blindness is a personal catastrophe to the individual and costs the United States approximately $500 million annually.⁵ However, risk of vision loss due to diabetic retinopathy can be reduced by effective control of serum glucose and blood pressure and by its early detection and timely treatment.^6- 8 The efficacy and cost-effectiveness of early detection and treatment of diabetic retinopathy is well established.^9- 10

Investigating the prevalence of diabetic retinopathy is important because it is a key indicator of systemic diabetic microvascular complications, and as such, a sentinel indicator of the impact of diabetes. Despite the documented increase in the prevalence of diabetes in the US population,¹¹ national population-based data on the prevalence and severity of diabetic retinopathy remain scarce, with previous nationwide prevalence estimates dating back to 1988-1994 (National Health and Nutrition Examination Surveys III [NHANES III]).¹² In 2004, the Eye Diseases Prevalence Research Group estimated the prevalence of diabetic retinopathy from the compilation of 8 separate population-based studies from the United States and elsewhere conducted in the late 1980s or early 1990s.¹³ Their report recommended that more recent estimates of diabetic retinopathy prevalence be obtained from the nationally representative sample of NHANES.

Moreover, several other population-based studies reported a decrease in the prevalence and incidence of severe diabetic retinopathy and related visual impairment.^14- 16 However, these findings were limited to regional cohorts and the status of diabetic retinopathy at the national level remains unknown. Thus, the principal aim of this study is to describe the most recent prevalence and risk factors of diabetic retinopathy in the US population aged 40 years and older using NHANES 2005-2008.

NHANES are national representative surveys conducted by the National Center for Health Statistics, part of the Centers for Disease Control and Prevention. The data consist of samples of the US noninstitutionalized civilian population, which were obtained using a stratified multistage probability design with planned oversampling of certain age and racial/ethnic groups.¹⁷ There were 6797 individuals aged 40 years and older interviewed for sociodemographic, medical, and family information and had a full medical examination at the medical examination center in NHANES 2005-2008. The NHANES 2005-2008 response rate for the interview sample aged 40 years and older was 71% and 69% for the examined sample.¹⁸ The NHANES protocol was approved by a human subjects review board and written informed consent was obtained from all participants.

NHANES 2005-2008 used the Canon CR6-45NM ophthalmic digital imaging system and Canon EOS 10D digital camera (Canon, Tokyo, Japan) to take 2 digital images per eye (total 4 images per participant) through a nonpharmacologically dilated pupil. Participants were seated in a windowless room with the lights turned off to allow the pupils to dilate naturally in preparation for the retinal imaging examination. One image was centered on the macula and the second on the optic nerve. The digital images were graded by masked photo graders at the University of Wisconsin Ocular Epidemiologic Reading Center, Madison, using a modification of the Airlie House classification system.^19- 21 Capture and grading of digital images and quality control by the Wisconsin group have been described in detail previously.²²

Survey participants who had no light perception or severe visual impairment in both eyes or had a severe infection in one or both eyes were excluded (n = 13). Complete data of fundus photographs of both eyes were obtained for 5371 (79%) participants aged 40 years and older who had full medical examinations.

Reasons for having incomplete data (n = 1426, 21%) included insufficient time to finish the examination (ie, arrived late or left early; n = 514; 42%), physical limitation (n = 238; 19%), eye-specific limitation (n = 193; 16%), participant’s refusal (n = 119; 10%), communication problems (n = 40; 3%), and others. Those individuals with incomplete data were more likely to be older, non-Hispanic black, with less than a high school education, higher systolic blood pressure, higher glycated hemoglobin A_1c level, and a history of using insulin than participants with complete gradable photographs (all P < .001). We further examined the potential influence of nonresponse bias due to the exclusion of participants without complete gradable photographs by adjusting the original sampling weights using the standard weighting–class method.^23- 24 Findings using these adjusted weights led to only minor differences in point and variance estimates (0%-0.5%), indicating minimal impact of nonresponse; therefore, we present all estimates using the original sampling weights.

Diabetic retinopathy was defined as the presence of 1 or more retinal microaneurysms or retinal blot hemorrhages with or without more severe lesions (hard exudates, soft exudates, intraretinal microvascular abnormalities, venous beading, retinal new vessels, preretinal and vitreous hemorrhage, and fibroproliferans) using the Early Treatment Diabetic Retinopathy Study (ETDRS) grading standards.¹⁹

Diabetic retinopathy was further categorized as nonproliferative and proliferative determined by assessment of the presence of retinal neovascularization or abnormal growth of new retinal blood vessels into the vitreous. Vision-threatening diabetic retinopathy, a level that may soon result in vision loss if left untreated, was defined as the presence of severe nonproliferative diabetic retinopathy, proliferative diabetic retinopathy, or clinically significant macular edema. Clinically significant macular edema was considered present when edema involved the fovea or was within 500 microns of the fovea, or when a 1+disc area of edema was present with at least a portion of it within the macula. Outcomes for this study were defined on the basis of the worse of the 2 eyes.

Diabetes was defined as self-report of a previous diagnosis of the disease by a clinician (excluding gestational diabetes mellitus) or hemoglobin A_1c of 6.5% or greater (American Diabetes Association's new diagnostic criterion for undiagnosed diabetes).²⁵ Although hemoglobin A_1c does not capture completely the increased risk of microvascular complications due to diabetes,²⁶ the diagnostic hemoglobin A_1c cut point of 6.5% was determined to be an inflection point for retinopathy prevalence, as is also true for the diagnostic thresholds of the glucose-based test.^25,27- 28 The final analytic sample consisted of 1006 individuals with diabetes aged 40 years and older (n = 795 for diagnosed diabetes; n = 211 for undiagnosed diabetes).

All participants were asked about their age, sex, race/ethnicity (non-Hispanic white, non-Hispanic black, Mexican American, and others [including those who selected multiple races and non-Mexican American Hispanics]), educational attainment (less than high school, high school education, or higher), and health insurance status. Consistent with previous epidemiologic studies,^29- 31 risk factors for diabetic retinopathy and vision-threatening diabetic retinopathy (hemoglobin A_1c, duration of diabetes, insulin use [yes/no], systolic and diastolic blood pressure, body mass index [BMI, calculated as weight in kilograms divided by height in meters squared], current smoking status [yes/no], and history of cardiovascular diseases [CVD; yes/no]) were examined. Hemoglobin A_1c, duration of diabetes, and systolic and diastolic blood pressure were used as continuous variables. Hemoglobin A_1c was used as the surrogate for blood glucose level and measured by a high-performance liquid chromatographic assay as used in the Diabetes Control and Complications Trial.⁷ Insulin therapy indicated that the participant had type 1 diabetes or their diabetes could not otherwise be controlled without insulin. We used measured height and weight to calculate BMI and divided respondents into 3 groups: normal/underweight (BMI<25), overweight (BMI 25-<30), and obese (BMI ≥30). Prior history of CVD was ascertained by self-report of coronary heart disease, angina, myocardial infarction, stroke, or congestive heart failure.

Statistical analyses were conducted using SAS version 9.2 (SAS Institute, Cary, North Carolina) and SUDAAN version 10.1 software (Research Triangle Institute, Research Triangle Park, North Carolina) to calculate national estimates and their standard errors while accounting for the complex survey design of the survey. Taylor series linearization was used for variance estimation.³² The NHANES 2005-2008 study has sufficient sample size to detect a relative difference of 6% (effective sample size = sample size/design effect = 1006/1.7 = 591) at 85% power and an α level of .05.

Characteristics of the study population are described using means for continuous variables and percentages for categorical variables. For continuous variables, t tests were used and for categorical variables the χ² test. We estimated the crude prevalence of diabetic retinopathy and vision-threatening diabetic retinopathy by age, sex, and race/ethnicity in the diabetic and overall US population. Multiple logistic regressions were used to assess the association between diabetic retinopathy and vision-threatening diabetic retinopathy, vs clinical potential risk factors for diabetic retinopathy and vision-threatening diabetic retinopathy after controlling for age, sex, race/ethnicity, and education attainment.

Predictive margins, odds ratios (OR), and 95% confidence intervals (CI) for each were calculated. Associations were considered to be significant if the P value was less than .05. Additionally, we compared the prevalence of diabetic retinopathy and vision-threatening diabetic retinopathy in NHANES 2005-2008 with NHANES III by using the right eye if the last digit of the participant identification number was even and the left eye if it was odd. The prevalence estimates were age standardized to the 2000 US census population.

In 2005-2008, the estimated (weighted) crude prevalence of diabetic retinopathy and vision-threatening diabetic retinopathy was 28.5% (95% CI, 24.9%-32.5%) and 4.4% (95% CI, 3.5%-5.7%), respectively, among persons with diabetes aged 40 years and older (Table 1). Extrapolating to the overall US population in the same period, the prevalence nation wide would be 3.8% (95% CI, 3.2%-4.5%) and 0.6% (95% CI, 0.5%-0.8%). Approximately 1.5% (95% CI, 1.1%-2.2%) of adults with diabetes had proliferative diabetic retinopathy and 2.7% (95% CI, 1.8%-4.0%) had clinically significant macular edema. In other words, approximately 0.2% (95% CI, 0.1%-0.3%) of adults aged 40 years or older had proliferative diabetic retinopathy and 0.4% (95% CI, 0.2%-0.5%) had clinically significant macular edema.

Among individuals with diabetes, no significant difference was found in the prevalence of diabetic retinopathy between those aged 40 to 64 years and those aged 65 years and older (28.0%; 95% CI, 23.0%-33.6%; vs 29.5%; 95% CI, 25.4%-33.9%; P = .64). Approximately 31.6% (95% CI, 26.8%-36.8%) of men with diabetes had diabetic retinopathy and approximately 25.7% (95% CI, 21.7%-30.1%) of women with diabetes had diabetic retinopathy (P = .04). Approximately 26.4% (95% CI, 21.4%-32.2%) of non-Hispanic white individuals, 38.8% (95% CI, 31.9%-46.1%) of non-Hispanic black individuals, and 34.0% (95% CI, 26.7%-42.1%) of Mexican American individuals with diabetes had diabetic retinopathy (P = .008). Prevalence of vision-threatening diabetic retinopathy was not statistically different between individuals aged 40 to 64 years and those aged 65 years and older (4.1%; 95% CI, 2.8%-5.8%; vs 5.1%; 95% CI, 3.5%-7.3%; P = .41). There was no significant difference in the prevalence of vision-threatening diabetic retinopathy between men and women (4.2%; 95% CI, 2.8%-6.1%; vs 4.7%, 95% CI, 3.2%-6.9%; P = .67). Approximately 3.2% (95% CI, 2.0%-5.1%) of non-Hispanic white individuals, 9.3% (95% CI, 5.9%-14.4%) of non-Hispanic black individuals, and 7.3% (95% CI, 3.9%-13.3%) of Mexican American individuals with diabetes had vision-threatening diabetic retinopathy (P = .006).

Extrapolating survey findings to the entire US adult population in the same period (without regard for diabetes status), the prevalence of diabetic retinopathy was significantly higher among individuals who were aged 65 years or older than those younger than 65 years of age (6.1%; 95% CI, 5.1%-7.3%; vs 3.1%; 95% CI, 2.4%-3.9% P < .001). Approximately 4.3% (95% CI, 3.5%-5.3%) of adult men in the United States had diabetic retinopathy compared with 3.3% (95% CI, 2.7%-4.1%) of adult women (P = .046). Non-Hispanic black individuals and Mexican American individuals had a higher prevalence of diabetic retinopathy than non-Hispanic white individuals (9.6%; 95% CI, 7.7%-11.9%; 6.7%; 95% CI, 5.4%-8.4%; vs 2.9%; 95% CI, 2.2%-3.9%; both P < .001).

Prevalence of vision-threatening diabetic retinopathy was higher among people aged 65 years or older than those aged 40 to 64 years (1.0%; 95% CI, 0.7%-1.5%; vs 0.4%; 95% CI, 0.3%-0.7%; P = .009). There was no significant difference in the prevalence of vision-threatening diabetic retinopathy between men and women observed (0.6%; 95% CI, 0.4%-0.9%; vs 0.6%; 95% CI, 0.4%-0.9%; P = .81). Approximately 0.4% (95% CI, 0.2%-0.6%) of non-Hispanic white individuals, 2.3% (95% CI, 1.5%-3.6%) of non-Hispanic black individuals, and 1.4% (95% CI, 0.8%-2.7%) of Mexican American individuals had vision-threatening diabetic retinopathy (P < .001).

Among individuals with diabetes, those with diabetic retinopathy were more likely to be men (53.7%; 95% CI, 47.4%-59.9%; vs 46.5%; 95% CI, 41.5%-51.6%; P = .04) than those without diabetic retinopathy (Table 2). Diabetic individuals with diabetic retinopathy had a longer duration of diabetes (15.0 years; 95% CI, 13.4-16.5; vs 7.3 years; 95% CI, 6.5-8.1; P < .001), higher systolic blood pressure (134.2 mm Hg; 95% CI, 131.6-136.9; vs 130.1 mm Hg; 95% CI, 127.9-132.4; P = .04), and higher hemoglobin A_1c level (7.9%; 95% CI, 7.6%-8.1%; vs 7.0%; 95% CI, 6.8%-7.1%; P < .001) than those without diabetic retinopathy. Diabetic individuals with diabetic retinopathy were more likely to use insulin than those with diabetes but no diabetic retinopathy (44.6%; 95% CI, 38.5%-50.9%; vs 10.2%; 95% CI, 8.1%-12.7%; P < .001).

Table 3 shows the associations of various risk factors with diabetic retinopathy and vision-threatening diabetic retinopathy among individuals with diabetes. In multivariate analysis, independent risk factors for diabetic retinopathy include male sex (38.1%; 95% CI, 32.6%-43.6%; vs 27.1%; 95% CI, 22.4%-31.8%; OR, 2.07; 95% CI, 1.39-3.10), higher hemoglobin A_1c level (OR, 1.45; 95% CI, 1.20-1.75), longer diabetes duration (OR, 1.06 per year duration; 95% CI, 1.03-1.10), use of insulin (47.4%; 95% CI, 39.1%-55.8%; vs 26.7%; 95% CI, 21.9%-31.5%; OR, 3.23; 95% CI, 1.99-5.26), and higher systolic blood pressure (OR, 1.03 per mm Hg; 95% CI, 1.02-1.03). Moreover, longer diabetes duration (OR = 1.03 per year duration; 95% CI, 1.01-1.05), use of insulin (7.5%; 95% CI, 4.8%-10.1%; vs 3.3%; 95% CI, 1.9%-4.7%; OR = 2.63; 95% CI, 1.34-5.15), and higher systolic blood pressure (OR, 1.03 per mm Hg; 95%CI, 1.01-1.06) were also associated with increased odds of vision-threatening diabetic retinopathy. The odds of vision-threatening diabetic retinopathy were significantly higher among non-Hispanic black individuals than in non-Hispanic white individuals, even after controlling for all other factors considered (9.8%; 95% CI, 5.7%-13.8%; vs 3.3%; 95% CI, 1.6%-5.0%; OR, 3.77; 95% CI, 1.47-9.69).

Additionally, we found that the age-standardized prevalence of diabetic retinopathy was statistically significantly different between 2 periods (15.8%; 95% CI, 12.7%-19.6% in NHANES III vs 22.1%; 95% CI, 18.4%-26.3% in NHANES 2005-2008; P = .01). A statistically significant difference of vision-threatening diabetic retinopathy was also observed; 1.3% (95% CI, 0.7%-2.3%) had vision-threatening diabetic retinopathy from the NHANES III data and 3.3% (95% CI, 2.4%-4.4%) from NHANES 2005-2008 (P = .002).

This study provides the latest nationally representative estimates on the prevalence and risk factors for diabetic retinopathy in the United States. Earlier population-based studies showed that almost all individuals with type 1 diabetes and more than 60% of those with type 2 diabetes develop diabetic retinopathy during the first 2 decades of the disease.² With the expected increased prevalence of type 2 diabetes in the population, due in part to increasing rates of obesity and decreasing physical activity, the burden of diabetic retinopathy might be expected to increase as well.³³ However, improved access to screening for and treatment of diabetic retinopathy may reduce the burden of diabetes-related vision loss.

Our prevalence estimates are somewhat lower than those from a previous meta-analysis that compiled data from 8 population-based studies of non-Hispanic white, non-Hispanic black, and Hispanic people with diabetes from the United States and elsewhere that reported an overall estimate of prevalence of diabetic retinopathy of 40% and a prevalence of vision-threatening diabetic retinopathy of 8%.¹³ However, the studies that were included in the meta-analysis were from regional cohorts, many were from prior decades, and most did not include individuals with undiagnosed diabetes. If undiagnosed diabetes were excluded from our sample, our estimates for diabetic retinopathy and vision-threatening diabetic retinopathy among self-reported diagnosed diabetes would be 32.8% (95% CI, 28.6-37.2) and 5.2% (95% CI, 4.0-6.7), respectively.

It is also possible that the lower prevalence reported in this study reflects a true reduction in the prevalence of diabetic retinopathy. Rates of other diabetic complications have declined during recent decades. For example, Geiss et al³⁴ found that hospitalization rate for lower extremity amputations among individuals with diabetes began decreasing in 1997. Another recent study found that age-adjusted diabetes-related end-stage renal disease decreased between 1996 and 2006.³⁵ Improved diabetes care, such as effective management of blood glucose levels, blood pressure, and serum lipid levels, is likely to reduce the incidence of diabetic retinopathy. Conversely, it might also lead to increased survival resulting in higher prevalence of diabetic retinopathy. Moreover, a recent longitudinal cohort study suggested that low vision and blindness could be substantially reduced among individuals with diagnosed diabetes who received guideline-recommended levels of care.³⁶

We found that the age-standardized prevalences of diabetic retinopathy and vision-threatening diabetic retinopathy were statistically significantly different between NHANES III and NHANES 2005-2008. These differences between surveys may be real or may be attributed to improved methods used to photograph the fundus in the more recent NHANES 2005-2008 compared with NHANES III. Two digital 45° color images of both eyes were taken in the NHANES 2005-2008 while in NHANES III only one 45° color film image from 1 eye was taken. Although some studies have shown grading of digital images to have similar sensitivity for detecting diabetic retinopathy as grading of film images, it is possible that the current method led to increased detection of both diabetic retinopathy and vision-threatening diabetic retinopathy because of more retina being assessed in 2 images compared with 1 and higher-quality digital images compared with earlier film images used in NHANES III, thus limiting our ability to directly compare results from NHANES III and current NHANES 2005-2008 study (R.K., unpublished data, 2009).

A previous analysis of NHANES III data by Harris et al¹² suggests that the prevalence of diabetic retinopathy was 46% higher in non-Hispanic black individuals and 84% higher in Mexican American individuals than in non-Hispanic white individuals. Although not statistically significant at a .05 level, we also found that among individuals with diabetes, non-Hispanic black individuals (47% higher) and Mexican American individuals (29% higher) had a higher crude prevalence of diabetic retinopathy than their non-Hispanic white counterparts. Moreover, the prevalence of vision-threatening diabetic retinopathy in individuals with diabetes was 190% higher in non-Hispanic black individuals and 130% higher in Mexican American individuals than in non-Hispanic white individuals. This may be due to individuals of non-Hispanic black and Mexican American heritage being more likely to have poorer glycemic control and being less likely to be screened and treated for diabetic retinopathy.³⁷ Data from National Health Interview Survey also suggest that non-Hispanic black individuals and Hispanics are less likely to use eye care services.³⁸ These findings lend further insight to inform national efforts to reduce disparities in care among racial/ethnic and socioeconomic groups and preserve sight for all adults in the United States.

Consistent with previous research, we found that higher levels of hemoglobin A_1c, longer duration of diabetes, insulin use, and higher systolic blood pressure were independently associated with diabetic retinopathy in the NHANES data.^29,39- 41 This is consistent with findings from randomized controlled clinical trials that showed that modifying identified risk factors such as glycemic control and blood pressure control could reduce the burden of diabetic retinopathy and also prevent vision loss caused by it.^7- 8 Prevention efforts may also need to target individuals with longer duration of diabetes and those using insulin. The new availability of care for vulnerable sections of the population should have demonstrable effect on risk of blindness in diabetes. Furthermore, primary prevention of diabetes, including identifying and protecting individuals at risk (ie, by reduced body weight and increased physical activity), may also help delay the onset of type 2 diabetes and reduce complications of diabetic retinopathy. We also found an inverse relationship between diastolic blood pressure and the presence of diabetic retinopathy. The underlying reason remains unexplained and requires further exploration. It could be due to selective participation bias. However, previous studies from Singapore⁴² and Africa⁴³ also suggested a possible association between pulse pressure (difference between systolic and diastolic blood pressure) and diabetic retinopathy.

The strengths of this study include its population-based national sample, its inclusion of individuals with undiagnosed diabetes, and improved detection of retinopathy due to use of digital fundus images of both eyes. This improved methodology results in estimates that are less biased than those obtained from NHANES III. However, individuals without diabetes may have retinopathy because of higher glucose level or hypertension, which is not assessed in the current study. Also, due to the infrequency of proliferative diabetic retinopathy (n = 23) and clinically significant macular edema (n = 37), we were unable to provide meaningful estimates by age, sex, and race/ethnicity.

Our study is subject to several limitations. First, individuals who were institutionalized (eg, nursing home residents) were not included in the NHANES, which may have led to an underestimate of diabetic retinopathy prevalence. Second, we could not distinguish between type 1 and type 2 diabetes and the specific risk of diabetic retinopathy complications. Third, there were substantial numbers of eligible individuals with diabetes who did not have photographs that could be graded, which may negatively bias estimates of diabetic retinopathy prevalence. Survey participants who had no light perception or severe visual impairment in both eyes, or a severe infection in 1 or both eyes were excluded—this might negatively bias the prevalence estimates. Small sample sizes might have prevented us from detecting differences, if they existed, between and among subgroups. Due to limitations inherent with the NHANES sampling frame, we were unable to estimate the prevalence of diabetic retinopathy and vision-threatening diabetic retinopathy among racial/ethnic groups other than non-Hispanic white individuals, non-Hispanic black individuals, and Mexican American individuals.

Our data demonstrate that a high prevalence of diabetic retinopathy and vision-threatening diabetic retinopathy in the United States exists, especially among racial/ethnic minorities. Male sex, higher hemoglobin A_1c level, longer duration of diabetes, insulin use, and higher systolic blood pressure were independently associated with the presence of diabetic retinopathy. These estimates provide policy makers updated information for use in planning eye care services and rehabilitation. With the aging of the population and the increasing proportion of the population with diverse racial/ethnic heritage, the number of cases of diabetic retinopathy and vision-threatening diabetic retinopathy will likely increase. Furthermore, the need for eye care and for culturally appropriate interventions that can reduce disparity and improve access to eye care among diverse populations is also likely to increase.