Diabetic Retinopathy in African Americans With Type 1 Diabetes: The New Jersey 725: Title and subTitle BreakII. Risk Factors FREE

CLINICAL and epidemiological studies of risk factors for diabetic retinopathy in patients with type 1 diabetes have been conducted among predominantly white populations, with little representation of African Americans.^{1 - 14} Thus, little is known about risk factors for diabetic retinopathy and its severity in this ethnic group.

Among white patients with type 1 diabetes, long duration of diabetes, poor glycemic control, high blood pressure, and the presence of renal disease are associated with the frequency and severity of diabetic retinopathy.^{1 - 10 ,12} Whether these are also risk factors for retinopathy in African Americans with type 1 diabetes is unknown. Among patients with diabetes, African Americans have higher prevalence rates of high blood pressure and are 2.6 to 5.6 times more likely to have diabetic nephropathy, including diabetic end-stage renal disease, than white patients.^{13 ,15} African Americans have lower mean incomes, fewer physician contacts, and poorer general health, and those with diabetes have poorer glycemic and blood pressure control than white patients with diabetes, all of which may put African Americans with type 1 diabetes at additional increased risk for development of diabetic retinopathy.^{13 ,16 - 19} Also, low socioeconomic status has been shown to be a strong predictor of morbidity for a number of diseases, including diabetes.^{20 - 21} Finally, US Blindness Registry data indicate that among nonwhite patients with mostly type 2 diabetes, women have higher rates of blindness than men, suggesting that female sex may be a risk factor for diabetic retinopathy in this ethnic group.²²

To examine whether some or all of these are risk factors for diabetic retinopathy in African Americans, a large cohort of African Americans with type 1 diabetes was studied (The New Jersey 725). Details of study methods, patient characteristics, frequency of diabetic retinopathy by age, and associated visual impairment are presented in the companion article.²³ This article presents data regarding 6 putative risk factors for diabetic retinopathy in The New Jersey 725 cohort. Specifically, it is hypothesized that the frequency and severity of diabetic retinopathy in African Americans with type 1 diabetes would be associated with long duration of diabetes, poor glycemic control, presence of systemic hypertension and renal disease, low socioeconomic status, and female sex.

Patients included in The New Jersey 725 were studied. Patients and methods are described in the companion article.²³ Briefly, patients were identified from a listing of 68 455 African Americans admitted to hospitals in New Jersey from January 1, 1982, through December 31, 1996, and listed on the New Jersey Hospital Discharge Data with a diagnosis of diabetes mellitus. To access this information, institutional review board approval was obtained from the New Jersey Department of Health. To obtain patients' most recent address, the last hospital admission was selected and a random listing of patients by hospital was generated. Since there was no reliable computer information regarding type 1 (insulin-dependent) or type 2 (non–insulin-dependent) diabetes mellitus, the last known medical charts were reviewed to identify patients with type 1 diabetes. To this aim, institutional review board approval was obtained in 31 of 40 hospitals located in the 7 counties (Bergen, Hudson, Passaic, Union, Middlesex, Essex, and Morris) located within a 20-mile radius of the New Jersey Medical School, Newark, and that listed at least 100 African Americans with a discharge diagnosis of diabetes. Of 39 710 African Americans with diabetes listed for the 7 counties for the study years, 34 941 (88.0%) had been admitted to the 31 hospitals where the chart review was performed. Of the 875 eligible patients with type 1 diabetes mellitus diagnosed before 30 years of age and currently receiving insulin therapy, 725 patients were enrolled.

Clinical evaluation included structured clinical interview, ocular examination, 7-field stereoscopic fundus photography, and blood pressure measurements. Retinopathy severity was determined for each eye via masked grading of fundus photographs using the modified Airlie House classification scheme and Early Treatment of Diabetic Retinopathy Study Final Scale.^{24 - 25} Severity of diabetic retinopathy for each patient was determined from the grading of the worse eye. Level 10 represents no retinopathy; level 15, questionable retinopathy; levels 20 through 53, nonproliferative retinopathy of increasing severity; and levels 61 through 85, proliferative retinopathy of increasing severity.²⁵ Any diabetic retinopathy was defined as level 20 or higher, and proliferative diabetic retinopathy as level 61 or higher.²⁵ Patients with a grading of no higher than 53 who had previously received laser photocoagulation for proliferative retinopathy received a grade of level 61, and those with documented vitreous hemorrhage or traction retinal detachment secondary to proliferative retinopathy, level 85.

Blood pressure was measured twice, in the sitting and standing positions, using a random zero sphygmomanometer following the Hypertension Detection and Follow-up Program protocol.²⁶ The average of both measurements in each position was used. As part of the structured clinical interview, a detailed medical history was obtained. Socioeconomic factors recorded included occupation, education, and family and personal incomes. The classification of occupations of Goldthorpe and Hope²⁷ was used to divide patients into middle (levels 1-25) or lower socioeconomic class (levels 26-33). Socioeconomic status was also determined from the Green index, using the education of the head of household, family income, and occupation of the main earner.^{28 - 29} Green index scores were divided into tertiles.

Patients were classified as currently receiving short- or long-acting insulin or a combination. Information regarding glucose control during the past year was recorded as follows. Missing an insulin injection was defined as often if at least once a week and as none if less than that. Frequency of home glucose urine testing was defined as never or rarely if performed a few times a year, as sometimes if performed once a month or some months per year, and as often if performed at least a few times a month per year. Frequency of home blood glucose testing was defined as sometimes if performed no more than once a week and as often if performed at least a few times a week.

Renal disease was considered to be present if the albumin-creatinine ratio in either of the 2 urine specimens was greater than 0.03, and/or the albumin excretion rate in the 4-hour timed urine specimen was at least 20 µg/min, and/or the patient was undergoing dialysis or had received a renal transplant for diabetic renal disease.³⁰ Microalbuminuria was defined as present if the albumin excretion rate was 20 to 200 µg/min; macroproteinuria, if the albumin excretion rate was more than 200 µg/min and/or the patient was undergoing dialysis or had received a kidney transplant.

A venous blood sample was obtained. Blood glucose level was measured using the hexokinase method; total (A₁ and A₂) glycosylated hemoglobin level, by means of high-pressure liquid chromatography (Biorad; Labcorp Laboratory, Hercules, Calif); serum creatinine level, by means of the alkaline picrate method; and plasma C peptide level, by means of a radioimmunoassay. The lower limit of detection of glycosylated hemoglobin is 0.02, and the coefficients of variation are 3% for values of 0.04 to 0.07, 2% for values of 0.08 to 0.12, 1.5% for values of 0.12 to 0.15, and 1% for values of more than 0.15. The lower limit of detection of C peptide is 0.03 nmol/L, and the interassay coefficient of variation is 5.7% for low, 6.7% for medium, and 6.1% for high values. Plasma high-density lipoprotein and total cholesterol levels were measured using an enzymatic assay, and low-density lipoprotein cholesterol levels, using immuno-separation spectrophotometry (Genzyme Diagnostics, Cambridge, Mass).

The morning first voided urine specimen brought by the patient and a 4-hour timed urine specimen collected in the clinic were assayed for creatine level using the alkaline picrate method and for albumin using a radioimmunoassay (SmithKline Beecham Clinical Laboratory, Philadelphia, Pa). Creatine level in the 4-hour timed urine collection was used to determine adequacy of collection. Leukocyte/nitrite dip stick reagent (Chemstrip; Boehringer Mannheim Corporation, Indianapolis, Ind) was used to exclude patients with urinary tract infection.

Data were entered into a database that used SPSS data entry (SPSS Inc, Chicago, Ill). The files were transformed into SPSS Windows-based databases for the actual analyses. To examine the relationship between putative risk factors and the presence of any retinopathy or proliferative retinopathy, logistic regression was used. Duration of diabetes served as the covariate, the putative risk factor as the independent variable, and any retinopathy or proliferative retinopathy as the dependent variable. For dichotomous variables, the odds ratio (OR) and 95% confidence interval (CI) for the predictor (abnormal vs normal or present vs absent) are presented; for categorical variables, the Wald test was used, and the OR for every level of the variable vs the normal category is presented. For continuous variables, the increased risk (as an OR) corresponding to a single-unit increase in the predictor is shown. In addition, for each risk factor a P value for a test of the null hypothesis that the OR is 1.0 is reported.

Descriptive analyses, including cross tabulations (stratified within various cohorts), were performed to identify potential confounders for the putative risk factors and to determine whether these might have been responsible for observed relationships between the risk factor and the retinopathy. Multiple logistic regression was used to isolate the impact of specific risk factors by controlling for the effect of potential confounders. The dependent variable in this regression was the presence or absence of retinopathy subsequent to a given period at risk (ie, the duration of diabetes). Independent variables were entered in a predetermined sequence, which allowed identification of the unique impact of each factor.

Details of the demographic characteristics of the 725 patients are presented in the companion article.²³ The cohort included a wide range of ages (3-80 years) and duration of diabetes (0.1-62 years). Female (58.3%) and male (41.7%) patients were well represented. Among the 724 patients in whom grading of the retinopathy was obtained, 463 (64.0%) had any retinopathy and 137 (18.9%) had proliferative diabetic retinopathy. The frequency and severity of any retinopathy or proliferative retinopathy were examined in relation to the 6 putative risk factors.

The frequency and severity of diabetic retinopathy were significantly associated with longer duration of diabetes (OR, 1.30 [95% CI, 1.25-1.36] and OR, 1.16 [95% CI, 1.13-1.19] for any retinopathy and proliferative retinopathy, respectively) (Table 1 and Figure 1). Microaneurysms were seen in patients with less than 2 years' duration of diabetes. The frequency of any retinopathy increased from 8.9% in patients with less than 2 years' duration to 94.7% in those with at least 30 years' duration. Proliferative retinopathy was seen in only 1 patient with 7 to 8 years' duration of diabetes, and its frequency increased from 10.2% in patients with 9 to 10 years' duration to 28.6% in those with 15 to 16 years' duration, and to 57.9% in those with at least 30 years' duration.

To determine the effect of duration of diabetes before and after puberty, age at onset of diabetes of younger than or at least 13 years was examined in relation to frequency of retinopathy after adjusting for duration of diabetes. Patients receiving a diagnosis at or after 13 years of age, compared with those receiving a diagnosis before 13 years of age, were on average twice as likely to have any retinopathy (OR, 1.85 [95% CI, 1.19-2.88]).

Glycemic control, as evaluated by total glycosylated hemoglobin values, was generally poor. The mean (± SD) glycosylated hemoglobin value was very high, 0.14 ± 0.04 (range, 0.04-0.29). Only 10.2% of patients had glycosylated hemoglobin values within the reference range (0.04-0.08), and 35.8% had values of at least 0.15. Glycosylated hemoglobin values were significantly and negatively associated with age (F₃ = 4.59; P<.003), and duration of diabetes (F₃ = 8.13; P<.001), but not with sex.

Mean (± SD) fasting blood glucose level obtained in a subset of patients (n = 134) was high, 14.3 ± 7.9 mmol/L (257 ± 142 mg/dL), with 49.3% of patients having blood glucose levels of more than 13.6 mmol/L (246 mg/dL). Twenty percent of patients had only 1 insulin injection per day. Only 8 patients (1.1%) had 4 insulin injections per day, and only 2 (0.3%) used a subcutaneous insulin infusion pump. A quarter of the patients missed their daily insulin injection at least once a week. More than half of the patients (60.9%) followed a specific diabetic diet no more than half of the time, and 34.8% had not received any diabetic dietary advice during the preceding year. Fifteen percent of patients never tested their blood glucose level, and among those who did, 23.2% reported fasting blood glucose level usually at 11.6 mmol/L (210 mg/dL) or more. Most patients (54.2%) rarely adjusted the amount of insulin taken. Also, 30.6% of patients reported 1 admission and 8.4% 1 emergency department visit for high blood glucose level during the previous year.

Various measures of glycemic control were examined in relation to the frequency and severity of retinopathy, after adjusting for duration of diabetes (Table 2 and Table 3). Total glycosylated hemoglobin values were significantly and positively associated with any retinopathy (P = .003). Patients with glycosylated hemoglobin values in the highest quartile were, on average, 3 times more likely to have any retinopathy compared with patients with values in the lowest quartile (OR, 2.94 [95% CI, 1.64-5.28]). Patients using more than 60 U/d of insulin were, on average, twice as likely to have any retinopathy compared with those using less than 30 U/d (OR, 2.11 [95% CI, 1.18-3.78]), and those who missed their insulin injection once a week or more were more likely to have proliferative retinopathy than those who never did (OR, 1.89 [95% CI, 1.19-2.99]). Patients with plasma C peptide levels of at least 0.13 nmol/L were more likely to have proliferative retinopathy than those with values of less than 0.13 nmol/L (OR, 1.62 [95% CI, 1.02-2.58]). Since C peptide levels were significantly associated with body mass index (calculated as weight in kilograms divided by the square of height in meters) (χ²₁, 83.67; P<.001), the association between C peptide level and proliferative retinopathy was reexamined after adjusting for body mass index and found not to be significant. Glycosylated hemoglobin values were not significantly associated with the frequency of proliferative diabetic retinopathy (Table 3).

Systemic hypertension, defined as systolic blood pressure of at least 160 mm Hg and/or diastolic blood pressure of at least 95 mm Hg (in sitting or standing position), and/or current use of antihypertensives, was present in 34.3% of patients and in 69.3% of those with proliferative retinopathy. Of the 188 patients who were receiving medication for systemic hypertension, only half (48.9%) were normotensive. A family history of hypertension in either parent was found in 61.7% of patients. Presence of systemic hypertension was significantly associated with older age (F₁ = 103.1; P<.001), longer duration of diabetes (F₁ = 93.1; P<.001), and male sex (χ²₁, 4.34; P = .04).

After adjusting for duration of diabetes, hypertensive patients were, on average, 4 times more likely to have proliferative retinopathy than those without hypertension (OR, 4.16 [95% CI, 2.61-6.62]). Patients in the highest quartile of systolic or diastolic sitting blood pressure were, on average, 2½ times more likely to have any retinopathy (OR, 2.48 [CI, 1.31-4.71] and OR, 2.56 [95% CI, 1.49-4.39], respectively), and, on average, 8 and 4 times more likely to have proliferative retinopathy than those in the lowest quartile (OR, 7.92 [95% CI, 3.88-16.16] and OR, 4.34 [95% CI, 2.39-7.90], respectively) (Table 4). There were similar associations between standing blood pressure and the frequency of proliferative retinopathy (data not shown).

In the cohort, 49.8% of patients had evidence of renal disease; 65 (9.2%) were undergoing dialysis (11 of whom had had a kidney transplant); 145 (20.4%) had microproteinuria; and 206 (29.1%) had macroproteinuria. Among patients with microproteinuria or macroproteinuria, 84.1% had any retinopathy and 35.3% proliferative retinopathy. Among patients with proliferative retinopathy, 73.7% had macroproteinuria and 16.8% had microproteinuria. Patients with renal disease were significantly older (F₁ = 128.3; P<.001) and had longer duration of diabetes (F₁ = 129.5; P<.001), and more were male (χ²₁, 12.27; P<.001).

After adjusting for duration of diabetes, patients with any renal disease were, on average, 3 times more likely to have any retinopathy than patients without renal disease (OR, 2.88 [95% CI, 1.88-4.34]). Patients with microproteinuria were, on average, 2 times and those with macroproteinuria were 4 times more likely to have any retinopathy than patients who had neither (OR, 2.37 [95% CI, 1.44-3.90] and OR, 4.20 [95% CI, 2.23-7.81], respectively) (Table 5).

After adjusting for duration of diabetes, patients who had any renal disease were, on average, 14 times more likely to have proliferative retinopathy than those who had no renal disease (OR, 13.94 [95% CI, 6.50-29.80). Patients with microproteinuria were, on average, 5 times and those with macroproteinuria were 20 times more likely to have proliferative retinopathy than those who had neither (OR, 5.28 [95% CI, 2.18-12.71] and OR, 20.14 [95% CI, 9.35-43.14], respectively) (Table 5).

Compared with patients with a college education or more, those with no more than a high school education had higher glycosylated hemoglobin values (χ²₃, 10.3; P = .02) and higher frequencies of renal disease (χ²₁, 4.8; P = .03). Unemployed patients had higher frequencies of renal disease (χ²₁, 12.3; P<.001) and systemic hypertension (χ²₁, 8.39; P = .004).

After adjusting for duration of diabetes, educational level, employment status, social class, and family and personal income were not significantly associated with the frequency of any retinopathy or proliferative retinopathy (Table 6).

Among the 302 men, 185 (61.3%) had any retinopathy and 53 (17.5%) proliferative diabetic retinopathy compared with 278 (65.9%) and 84 (19.9%), respectively, of the 422 women (differences were not significant). There was no significant association between frequency of any retinopathy or proliferative retinopathy and sex after adjusting for duration of diabetes (data not shown).

To evaluate the relative contribution of the various risk factors to the frequency and severity of diabetic retinopathy, models based on logistic regression were developed. Age at examination, age at diagnosis of diabetes, duration of diabetes, glycosylated hemoglobin values, sitting systolic and diastolic blood pressures, presence of renal disease, socioeconomic status, and sex were entered in the model. Presence of renal disease, poor glycemic control, and longer duration of diabetes were found to be independently associated with the frequency of any retinopathy. Presence of renal disease, the highest levels of systolic blood pressure, and longer duration of diabetes were independently associated with the frequency of proliferative retinopathy (Table 7). When the multivariate analysis was repeated excluding possible confounding factors, ie, glycosylated hemoglobin levels and presence of renal disease, there still was no significant relationship between diabetic retinopathy and socioeconomic factors.

The data indicate that the frequency and severity of diabetic retinopathy in African Americans with type 1 diabetes mellitus are significantly associated with longer duration of diabetes and renal disease, but not with socioeconomic factors or sex. Poor glycemic control was significantly and independently associated with the frequency of any retinopathy, and the highest levels of systolic blood pressure, with the frequency of proliferative retinopathy.

The frequency and severity of diabetic retinopathy were significantly associated with the duration of diabetes. Frequency of any retinopathy increased sharply between 5 and 10 years' duration of diabetes. After 15 years' duration, 90.5% of patients had some evidence of diabetic retinopathy, and 28.6% had proliferative disease, rates that are similar to those reported for white patients with type 1 diabetes.^{2 - 10 ,12} Proliferative retinopathy was first seen in African Americans after a relatively short (approximately 8 years) duration of diabetes, and its frequency in those with at least a 30-year duration of diabetes was high (57.9%).

Also, patients receiving a diagnosis at 13 years of age or older compared with those receiving a diagnosis at younger than 13 years had, after adjusting for duration of diabetes, a 1.8 higher risk for having any retinopathy. This finding has also been reported previously in white patients with type 1 diabetes and is thought to reflect increases in growth factors and/or hormonal changes and/or deterioration in glycemic control at the time of puberty.^{31 - 33} For example, Murphy et al³³ reported a 4.8 higher risk for diabetic retinopathy in postpubescent compared with prepubescent or pubescent patients with type 1 diabetes. Mortality data for this cohort also indicate that 31 (79.5%) of the 39 patients who have died since the study ended had received a diagnosis of diabetes at 13 years of age or older.

Glycemic control in this population was very poor. Ninety percent of patients had total glycosylated hemoglobin values outside the normal range, and 7.7% of patients had values of at least 0.20. This was consistent with the fact that 14.9% of them never tested their blood glucose level, 27.2% missed their insulin injection at least once a week, 54.2% rarely adjusted their insulin dose themselves, 60.9% followed a specific diabetic diet no more than half of the time, and 30.6% had had 1 admission to the hospital for high blood glucose level during the previous year. Poor glycemic control was significantly associated with younger age and shorter duration of diabetes. Among persons with type 1 diabetes mellitus, significantly poorer glycemic control has been previously reported in African American adolescents and adults when compared with white patients.^{11 ,18 ,34 - 35}

In this study, poor glycemic control was significantly associated with the presence of any retinopathy, as previously found for white patients.^{4 ,6 - 7 ,10 ,12 ,36} Such an association between glycemic control and retinopathy is also supported by results of the Diabetes Control and Complications Trial, which showed a strong relationship between improved glycemic control following intensive insulin treatment and reduced incidence and progression of diabetic retinopathy during a 10-year period.³⁷ Sixteen percent of patients in the Eurodiab Complications Study had glycosylated hemoglobin values within the normal range³⁶ ; this may explain the lower frequency of diabetic retinopathy in their patients compared with patients in this study (46.2% vs 64.0%, respectively).²³ In the African American patients, higher daily insulin doses were associated with a higher frequency of any retinopathy, a finding that may reflect greater severity of diabetes. However, this factor was not significant when other potential confounders were added to the statistical model.

In this study, there was no association between glycemic control and proliferative retinopathy, as reported in another cross-sectional study.¹⁰ Possible explanations for the lack of association are that renal complications may have developed in patients with the worst glycemic control, and after seeking medical care they may have been advised to achieve better glycemic control, or more likely, they may have died before being seen in the study. This is supported by mortality data for this cohort showing that 64.1% of patients who died since the study ended had proliferative diabetic retinopathy.

THE FREQUENCY of hypertension in the African American patients was high (34.3%), and was significantly higher than the 17.3% reported for white patients with type 1 diabetes in the Wisconsin Epidemiologic Study of Diabetic Retinopathy (WESDR) (χ²₁, 65.57; P<.01).³⁸ Furthermore, in only 48.9% of the patients who received antihypertensive medication was the blood pressure actually controlled, as previously reported for African Americans with type 2 diabetes mellitus.¹³ There was a statistically very strong association (on univariate analysis) between higher levels of systolic and diastolic blood pressures and frequency and severity of retinopathy. When other risk factors are taken into account, only the highest levels of systolic (and not diastolic) blood pressure remain a significant and independent risk factor for proliferative retinopathy. This suggests that the importance of systolic blood pressure may be greatest at higher levels or operates at a level important enough to lead to proliferative diabetic changes. In the WESDR, the 10-year follow-up of patients with younger-onset diabetes showed that baseline systolic blood pressure was associated with a small increased risk for incident retinopathy.³⁹ Other investigators have reported a significant association between high diastolic blood pressure and frequency of proliferative retinopathy.^{9 - 10} It may also be that diabetes affects blood pressure and retinopathy with no causal link.

The frequency of renal disease in the African American patients in this study was also high (49.8%). Nine percent of patients were undergoing dialysis alone or have undergone renal transplantation with dialysis, more than 3 times the 2.8% reported in the WESDR (χ²₁, 30.47; P<.01).⁴⁰ This high prevalence of renal disease is in agreement with the higher rates of macroproteinuria and diabetic end-stage renal disease reported among African Americans with diabetes when compared with white patients with diabetes.^{15 ,41 - 42} It is also consistent with the high prevalence of hypertension in the study patients (34.3%) and with previous reports indicating that diabetic nephropathy is more prone to develop in diabetic patients with a known family history of systemic hypertension (61.7% of the patients in this study) than in patients without such a history.^{43 - 44}

Renal disease was significantly associated with higher frequencies of any retinopathy and proliferative retinopathy. In diabetic patients, renal disease is thought to increase the risk for diabetic retinopathy because of associated hypertension or blood abnormalities, eg, increased plasma levels of lipoproteins and fibrinogen.¹⁰ In this study, the association between retinopathy and renal disease was very strong (P<.001) and remained significant even after taking into account other confounders, including systemic hypertension. The data also showed that although only 35.0% of patients with renal disease had proliferative disease, 93.3% of those with proliferative retinopathy had renal disease, suggesting that in African Americans with type 1 diabetes mellitus, diabetic nephropathy may precede retinopathy. This differs from incidence data for white patients with type 1 diabetes, which indicate that severe retinopathy usually precedes the development of proteinuria.^{40 ,45}

In this study, the significant association between hypertension or renal disease and male sex is expected, since higher frequencies of renal disease have been reported previously in white and African American men with type 1 diabetes mellitus.^{15 ,45} One might speculate that, in this cohort, the high prevalence of systemic hypertension and renal disease in male patients—conditions that are predictors of death in diabetic patients—may have led to an underestimation of the frequency of proliferative retinopathy in men.^{46 - 48}

In this cohort of patients, there was no association between diabetic retinopathy and socioeconomic factors as reported in some, but not all, studies.^{6 ,49 - 53} In the WESDR, higher educational attainment in women aged 25 years or older was the only socioeconomic factor significantly associated with a lower incidence of proliferative diabetic retinopathy.⁵¹ Similarly, in the European clinic-based Eurodiab IDDM Complications Study, proliferative diabetic retinopathy was more common in primary school–educated than in college-educated men, even after adjusting for other risk factors such as glycemic control.⁵³ African American patients with lower education had poorer glycemic control, and more of them had evidence of renal disease than patients with higher educational attainment. However, when the multivariate analysis was repeated excluding both confounding factors, there was still no significant association between retinopathy and education. The lack of association between diabetic retinopathy and socioeconomic factors may have resulted from selective mortality of those with the lowest socioeconomic status. This again is supported by mortality data for this cohort, indicating that 74.4% of those who died since the study ended had low socioeconomic status.

In this study, there was also no difference in the frequency or severity of retinopathy between men and women, a finding consistent with the lack of gender difference in rates of blindness.²³

This study demonstrates that African Americans with type 1 diabetes mellitus have poor glycemic control and high frequencies of renal disease and systemic hypertension. The high frequency of the latter 2 risk factors, however, may be related to the fact that patients were selected based on their hospital discharge records and, thus, may be sicker that those in the general overall diabetic population. After adjusting for confounding variables, renal disease, poor glycemic control, and longer duration of diabetes were independently associated with the frequency of any diabetic retinopathy; renal disease, the highest levels of systolic blood pressure, and longer duration of diabetes were independently associated with the frequency of proliferative diabetic retinopathy. Longitudinal studies are needed to document the natural history of diabetic retinopathy in this population and to ascertain further the relationship of these risk factors, particularly renal disease, to the incidence and progression of retinopathy.

Accepted for publication May 21, 1999.

This research was supported by grant RO1 EY 09860 from the National Eye Institute, Bethesda, Md.

I thank Michael Borenstein, PhD, statistical consultant, Director of Biostatistics, Hillside Hospital, Long Island Jewish Medical Center, Glen Oaks, NY, for conducting the primary analyses; Ronald Klein, MD, MPH, for advice regarding recruitment of patients, data collection, and analysis, and for valuable suggestions on the article; Deborah Nuber, study coordinator, for technical assistance; Lisa Schoenherr, social worker, for patient recruitment; Jim Besra, MPH, research assistant; the Fundus Photograph Reading Center for the grading of the photographs; Maxine Wanner and Richard Press for the fundus photography; Clara Baker for secretarial help; and the New Jersey Department of Health and the 31 New Jersey hospitals that participated in the study.

Reprints: Monique S. Roy, MD, University of Medicine and Dentistry, New Jersey Medical School, Department of Ophthalmology, 90 Bergen St, Room 6164, Newark, NJ 07103.