Roy MS, Klein R, O'Colmain BJ, Klein BEK, Moss SE, Kempen JH. The Prevalence of Diabetic Retinopathy Among Adult Type1 Diabetic Persons in the United States. Arch Ophthalmol. 2004;122(4):546-551. doi:10.1001/archopht.122.4.546
Copyright 2004 American Medical Association. All Rights Reserved.Applicable FARS/DFARS Restrictions Apply to Government Use.2004
To estimate the US prevalence of diabetic retinopathy (DR) among personswith type 1 diabetes mellitus (DM).
Prevalence data from the New Jersey 725 and Wisconsin EpidemiologicStudy of Diabetic Retinopathy were used to estimate the prevalence of DR byage, gender, and race among persons 18 years and older having type 1 DM diagnosedbefore age 30 years. Severity of DR was determined via masked grading of 7-fieldstereoscopic fundus photographs. Any DR was defined as retinopathy severitylevel of 14 or more; and vision-threatening retinopathy, as retinopathy severitylevel of 50 or more, the presence of clinically significant macular edema,or both. The estimates of the prevalence of DR among persons with type 1 DMwere applied to the estimated number of persons with type 1 DM diagnosed beforeage 30 years in the 2000 US population to obtain prevalence estimates of DRdue to type 1 DM in the general population.
Among 209 million Americans 18 years and older, an estimated 889 000have type 1 DM diagnosed before age 30 years. Among persons with type 1 DM,the crude prevalences of DR of any level (74.9% vs 82.3% in black and whitepersons, respectively) and of vision-threatening retinopathy (30.0% vs 32.2%,respectively) are high. The prevalence of DR due to type 1 DM diagnosed beforeage 30 years in the general population 18 years and older is estimated at767 000 persons having DR of any level (0.37%), and 376 000 personshaving vision-threatening retinopathy (0.18%).
Retinopathy due to type 1 DM is an important public health problem inthe United States, affecting 1 per 300 persons 18 years and older, and 1 per600 persons with advanced, vision-threatening retinopathy.
Diabetic retinopathy (DR), the most common ocular complication of diabetesmellitus (DM), is the leading cause of new cases of legal blindness in Americansaged 20 through 74 years, despite the fact that visual loss due to DR maybe preventable either through better glycemic control or photocoagulationtreatments.1- 4 Forthis article, the goal of the Eye Diseases Prevalence Research Group is toestimate the prevalence of DR in the United States specifically attributableto type 1 DM.5
The prevalence of DR among persons with type 1 DM has been previouslyreported in 2 large cohorts of patients: whites (the Wisconsin EpidemiologicStudy of Diabetic Retinopathy [WESDR])6 andblacks (the New Jersey 725).7,8 Thepurpose of this article is (1) to estimate the prevalence of DR by age, gender,and race in persons with type 1 DM using data from those 2 studies, and (2)to estimate the prevalence of DR due to type 1 DM in the general populationby applying the stratum-specific prevalence estimates of DR to the estimatednumber of persons with type 1 DM diagnosed before age 30 years derived fromthe National Health Interview Survey9 and the2000 US Census.10
Data on patients aged 18 years and older from 2 cross-sectional prevalencestudies, the New Jersey 725 and the WESDR, were used to estimate the prevalenceof DR among persons with type 1 DM.6- 8 Patientsincluded in the New Jersey 725 were identified from among 39 710 AfricanAmericans, all of whom had a diagnosis of DM and were discharged between January1, 1982, and December 31, 1996, from 31 hospitals located in 7 counties withina 20-mile radius of the New Jersey Medical School, Newark. Of 875 patientshaving type 1 DM whose condition had been diagnosed before the patient was30 years old and who were receiving insulin therapy, 725 (83%) were enrolledin the study.7 In the WESDR, 996 younger-onsetpersons diagnosed as having DM before age 30 years who were taking insulinwere identified from among 10 135 patients with DM who received primarycare in an 11-county area in southern Wisconsin from 1979 to 1980.6 In both studies, DR was ascertained by masked gradingof stereoscopic color fundus photographs of the 7 standard fields using themodified Airlie House classification of Diabetic Retinopathy and Early Treatmentof Diabetic Retinopathy Study (ETDRS) severity scale.11,12
For both the New Jersey 725 and the WESDR, color fundus photographswere graded for DR by the Wisconsin Fundus Photograph Reading Center, Madison.For each eye, the maximum grade in any of the 7 standard photographic fieldswas used to define the retinopathy level according to the ETDRS severity scale.12 Subsequently, the retinopathy level for a participantwas determined, based on the worse eye. If the retinopathy severity couldnot be graded in 1 eye, the subject was considered to have a score equivalentto that in the gradable eye. Clinically significant macular edema (CSME) wasdefined as thickening of the retina with or without partial loss of retinaltransparency at or within 500 µm of the center of the macula; or ashard exudates at or within 500 µm of the center of the macula associatedwith thickening of the adjacent retina; or as an area of retinal thickening1 disc diameter (DD) or larger, any part of which was within 1 DD of the centerof the macula.13 If macular edema could notbe graded in an eye, the participant was assigned the score of the other eye.
Eyes that could not be graded—because of opacities of the media,phthisis, or enucleation—were initially classified as "cannot grade."For such persons, review of all previous medical records was done subsequentto the study visit. When a history of panretinal photocoagulation for proliferativeDR or pars plana vitrectomy for complications of proliferative DR was identified,then the retinopathy level was scored as 85. Persons who had an ETDRS gradingof less than 50 at the time of examination and had previously received laserphotocoagulation for proliferative DR, as documented by medical record review,were classified as grade 61.
For the purpose of this article, the severity scales were collapsedinto the following categories based on the consensus of the Eye Diseases PrevalenceResearch Group:
Mild nonproliferative DR (level 14 up to, but not including level40)
Moderate nonproliferative DR (level 40 up to, but not includinglevel 50)
Severe DR (level ≥50, including severe nonproliferative andproliferative retinopathy)
Composite outcomes are defined as follows: (1) DR of any level, consistingof "mild," "moderate," or "severe" retinopathy, CSME, or any combination thereof;and, (2) vision-threatening DR (VTDR), consisting of severe retinopathy, CSME,or both.
The composite outcomes were used as the primary outcomes for this article,respectively, indicating (1) the presence of any level of DR and (2) a levelof DR likely to result in vision loss in the absence of laser photocoagulationtreatment.
Included in both studies were patients with type 1 DM whose conditionswere diagnosed and treated with insulin before the patients were 30 yearsold and who were receiving insulin therapy.6- 8,14 Inthe New Jersey 725 study, African Americans who had the onset of DM beforeage 30 years but were not receiving insulin therapy were excluded.15 Age at onset of DM was taken as the age at whichthe diagnosis was first recorded by a physician on the patient's medical recordor in a hospital medical record. Subjects' current age was defined as theage at the baseline examination. Race of the African Americans was determinedfrom the hospital record and later confirmed by self-identification.
The age-, gender-, and race-specific prevalences of DR of any leveland of VTDR among persons with type 1 DM are given for the following age categories:18 through 39, 40 through 49, and 50 years and older. Prevalence for whiteand black persons was taken directly as reported in the WESDR and New Jersey725 studies. Owing to lack of published data on the prevalence of DR due totype 1 DM for Hispanic persons and "other" race/ethnicity groups, the prevalenceof DR in these populations was estimated using an unweighted average of theprevalence estimates for white and black persons. Since other race/ethnicitygroups make up approximately 17% of the US population, these estimates werederived solely to provide some broad indication of the number of persons inthese populations with DR due to type 1 DM.
To determine the number of individuals with DR due to type 1 DM in theUS general population, an estimate of the total number of persons with type1 DM diagnosed before age 30 years in each age-, gender-, and race-stratumwas derived based on data from the National Health Interview Survey whichgathers data from a nationally representative sample of civilian, noninstitutionalizedpersons residing in the United States.9 Personsanswering yes to the questions "Have you ever been told by a doctor or healthprofessional that you have diabetes or sugar diabetes" and "Are you now takinginsulin" and also reporting age first diagnosed with DM as being younger than30 years represented 0.38%, 0.48%, and 0.41% of the national population 18years and older in 1999, 2000, and 2001, respectively. The average overallprevalence for these 3 years (4.25/1000) was used to extrapolate to the 2000US population 18 years and older, arriving at an estimated 889 000 personshaving type 1 DM diagnosed before age 30 years. The prevalence of DR amongpersons with type 1 DM was then applied to the estimates of the number ofpersons with type 1 DM in each age-, gender-, and race-specific stratum, resultingin stratum-specific estimates of the prevalence of DR (of any level and ofVTDR) among persons with type 1 DM in the 2000 US population.
Tests for age-, gender-, and race-effects were conducted for DR of anylevel and for VTDR, as well as for mild, moderate, and severe DR, and forCSME. Gender differences were evaluated using the Mantel-Haenszel stratified-adjustedestimate of the relative risk with test inverted confidence limits. Similarly,racial differences were evaluated using the Mantel-Haenszel stratified-adjustedestimate of the relative risk adjusting for age and gender. The Cochran-Armitagetest for trend and the Mantel-Haenszel χ2 test for generalassociation were used to evaluate age effects. Data analyses were conductedfollowing approval by the institutional review board of the The Johns HopkinsSchool of Medicine, Baltimore, Md, and followed the principles of the Declarationof Helsinki.
Two studies contributed data on DR in persons with type 1 DM, of whom594 were black (the New Jersey 725) and 790 were white (the WESDR). The originalstudies considered persons of all ages; in this study, only persons 18 yearsand older are included. Characteristics of the subjects with type 1 DM fromthe 2 studies are given in Table 1.
Among persons with type 1 DM, the crude prevalences of both DR of anylevel (74.9% among black and 82.3% among white persons) and VTDR (30.2% and32.2%, respectively) were high, but similar in both racial groups. The prevalencesof DR of any level and of VTDR increased significantly with age for both menand women (test for trend, P <.001, for both DRof any level and VTDR). However, the prevalence for men aged 50 years andolder tended to level off or decrease, whereas the prevalence for women aged50 years and older continued to increase, more so among black than among whitewomen (Table 2). The prevalenceof CSME increased significantly with age among white women (test for trend, P = .05), and even more so among black women (test fortrend, P = .03), but not in either white (test fortrend, P = .42) or black (test for trend, P = .78) men. After controlling for gender and race, tests of the associationbetween age and DR of any level, VTDR, severe DR, and CSME were all statisticallysignificant (χ2 test, P <.001 ineach instance).
Gender differences in the prevalence of DR among persons with type 1DM also were observed. After adjusting for age and race, women were more likelyto have DR of any level than men (relative risk [RR] = 1.04; 95% confidenceinterval [CI], 0.99-1.10), mainly owing to a higher prevalence of mild DR(RR = 1.24; 95% CI, 1.09-1.42). However, men were more likely to have VTDRthan women (RR = 1.17; 95% CI, 1.01-1.36), largely because of a higher prevalenceof CSME among black men compared with black women (RR = 1.94; 95% CI, 1.20-3.12),and a higher prevalence of severe DR among white men compared with white women(RR = 1.24; 95% CI, 1.01-1.52). The prevalence of moderate DR (RR = 1.14;95% CI, 0.86-1.53) was not significantly different for men compared with women.
After adjusting for age and gender, black persons with type 1 DM tendedto have a lower prevalence of DR of any level than did white persons withtype 1 DM (RR = 0.92; 95% CI, 0.87-0.97), but they had a similar prevalenceof VTDR (RR = 0.99; 95% CI, 0.85-1.15). However, black persons with type 1DM were more likely to have CSME than white persons with type 1 DM (RR = 1.43;95% CI ,1.01-2.03). This difference was most pronounced for black men comparedwith white men (RR = 1.67; 95% CI, 1.05-2.63). The prevalences of mild, moderate,and severe DR were slightly higher among white than among black persons, butdifferences were not statistically significant.
The estimated prevalences of DR of any level and of VTDR in the generalpopulation are given in Table 3.In the United States an estimated 889 000 persons (0.43%) have type 1DM diagnosed before age 30 years; of these, approximately 767 000 (86.4%)have DR of any level and 376 000 (42.1%) have VTDR. The estimated prevalencesof DR of any level and VTDR in the general population are 0.37% and 0.18%,respectively.
The results of this analysis indicate that DR due to type 1 DM is commonin the United States. Approximately 86% of persons with type 1 DM have DRof any level and 42% have VTDR. In the general population, DR due to type1 DM is a major public health problem, with 1 per 300 adults (767 000)having DR of any level, and 1 per 600 adults (376 000) having VTDR. Additionalpersons would have DR due to older-onset type 1 DM diagnosed at or after age30 years. However, because specific data on the prevalence of DR due to older-onsettype 1 DM are unavailable from population-based surveys, it was impossibleto obtain an overall estimate of the burden of DR because of type 1 DM. Inthis article, only estimates for persons with type 1 DM diagnosed before age30 years are reported. An accompanying article provides estimates for theoverall prevalence of DR in persons aged 40 years and older.5
The estimated population prevalences of DR of any level and of VTDRare highest in the 40- through 49-year and the 50-year-and-older age groups,most likely because older persons who had the onset of DM prior to age 30years had a longer average duration of disease. The prevalence of DR in thegeneral population declines thereafter, reflecting a decreasing prevalenceof type 1 DM in older age groups (possibly due to early mortality among personswith type 1 DM).
White women with type 1 DM were more likely to have mild DR than whitemen, and white men had a higher prevalence of VTDR than white women, supportinga previous report of greater severity of DR among men compared with womenin white persons with type 1 DM.16 Among blackand white persons with type 1 DM, the prevalence of DR of any level increasedsignificantly with age, reaching more than 95% in those aged 40 years andolder. Among persons with type 1 DM, the prevalence of VTDR and of CSME didnot increase uniformly with age across either gender or race. The prevalenceof VTDR among black and white men with type 1 DM increased with age from the18- through 39- to 40-through 49-year age group, but then declined for thoseaged 50 years and older. In contrast, the prevalence of VTDR among black womenwith type 1 DM increased significantly with age, and black women aged 50 yearsand older had a significantly higher risk of VTDR than white women of thesame age. The greater severity of DR observed among black women in New Jerseywas consistent with self-reported blindness registry data that indicated thatnonwhite women were 3 times more likely to be blind from DR than any otherrace-gender subgroup.17
There were also differences between black and white persons with type1 DM in that the frequency of CSME was much higher among black persons, particularlyamong black men. Among black men, the frequency of macular edema has beenpreviously shown to be associated with the presence of renal disease, whichitself is more prevalent among black (particularly men) than among white persons.18- 20 In a clinic-basedstudy, Arfken et al21 previously reported nosignificant differences in either the prevalence or severity of DR when comparing58 African Americans with 142 white subjects with type 1 DM. In that study,however, there were no data regarding frequency of macular edema.
The strengths of the current article are that both the New Jersey 725and WESDR had similar inclusion criteria (age at onset of DM <30 yearswas used) and used gold standard determinations of the outcomes (masked gradingby the same reading center of retinal photographs of 7 standard fields usingthe ETDRS severity scale). Limitations include the fact that the studies wereconducted about 15 years apart in different geographic locations. The olderWESDR data may overestimate the current prevalence of DR of any level andVTDR because glycemic and blood pressure control have probably improved inthe general population during the interval. These changes in clinical practicehave probably decreased the incidence of DR but may also have improved survival,which would affect the prevalence of DR in opposite directions. Also, differencesin the methods of obtaining population-based samples or other differencesbetween study protocols could have given rise to apparent racial differencesbecause neither study simultaneously evaluated both racial groups.
Estimates of the prevalence of DR due to type 1 DM in the general populationmust also be interpreted with caution since only a general estimate of theprevalence of type 1 DM in the United States was available, thus, limitingthe ability to accurately assess the total number of persons with type 1 DMfor each age-, gender-, and race-specific stratum. Racial/ethnic differencesas well as gender and age differences in the incidence of type 1 DM have beenreported, but no reliable national data on the prevalence of type 1 DM inthese populations are available.22 The estimatesof the prevalence of DR due to type 1 DM for other races/ethnicities werebased on extrapolation of the rates for white and black persons and, thus,may not accurately reflect the true prevalence in these populations. However,our results provide some measure of the burden of DR due to type 1 DM in theUnited States and may, thus, serve as a useful guide regarding public healthpolicies and research priorities until more satisfactory data regarding theprevalence of type 1 DM in the United States by age, gender, and race/ethnicityand by the prevalence of DR due to type 1 DM in racial/ethnic groups otherthan blacks and whites become available.
In addition, there may be about 0.31% of the US population aged 30 through74 years who have older-onset type 1 DM (onset at age ≥30 years) and anunknown number of adults identified with type 2 DM who slowly become insulindependent.23 Our studies only evaluated DRin individuals having type 1 DM diagnosed before age 30 years. Thus, we areunable to accurately assess the nature and severity of DR in persons witholder-onset type 1 DM. The omission of individuals with older-onset type 1DM would lead to underestimates of DR due to type 1 DM in the general population.Therefore, the burden of DR due to type 1 DM in the general population maybe as much as 2 times higher than the estimates provided here for DR due toyounger-onset type 1 DM only.
In the United States, the prevalence of DR due to type 1 DM is high.An estimated 767 000 persons aged 18 years and older in the US generalpopulation have DR of any level, 1 per 300 adult persons. An estimated 376 000persons in this age range have VTDR, 1 per 600 adult persons. The prevalenceof DR is expected to increase substantially by 2020, driven by an increasingprevalence of DM over time with the aging of the US population, in combinationwith anticipated increases in the age-specific prevalence of DM. Because DRis a substantial public health problem, public and private policy effortsdirected toward improving primary and secondary prevention programs are warranted.
Corresponding author and reprints: Monique S. Roy, MD, Departmentof Ophthalmology, University of Medicine and Dentistry of New Jersey–NewJersey Medical School, 90 Bergen St, Room 6164, Newark, NJ 07101-1709 (e-mail: Roymo@umdnj.edu).
Submitted for publication August 6, 2003; final revision received December30, 2003; accepted December 30, 2003.
This study was supported by grants RO1 EY 09860 (Dr Roy), EY03083 (DrsKlein and Klein), and EY00386 (Dr Kempen) from the National Eye Institute,Bethesda, Md; and a Lew Wasserman Merit Award from Research to Prevent Blindness,New York, NY (Dr Roy).