Risk Factors for Proliferative Diabetic Retinopathy in African Americans with Type 2 Diabetes

Abstract

Purpose

To assess personal and demographic risk factors for proliferative diabetic retinopathy in African Americans with type 2 diabetes.

Methods

In this prospective, non-interventional, cross-sectional case-control study, 380 African Americans with type 2 diabetes were enrolled. Participants were recruited prospectively and had to have either; 1) absence of diabetic retinopathy after ≥10 years of type 2 diabetes or, 2) presence of proliferative diabetic retinopathy when enrolled. Dilated, 7-field fundus photographs were graded using the Early Treatment Diabetic Retinopathy Study scale. Covariates including hemoglobin A_1C (HbA_1C), blood pressure, height, weight and waist circumference were collected prospectively. Multivariate regression models adjusted for age, sex and site were constructed to assess associations between risk factors and proliferative diabetic retinopathy.

Results

Proliferative diabetic retinopathy was associated with longer duration of diabetes (odds ratio, OR, 1.62, P<0.001), higher systolic blood pressure (OR 1.65, P<0.001) and insulin use (OR 6.65, P<0.001) in the multivariate regression analysis. HbA_1C was associated with proliferative diabetic retinopathy in the univariate analysis (OR 1.31, P=0.002) but was no longer significant in the multivariate analysis.

Conclusions

In this case-control study of African Americans with type 2 diabetes, duration of diabetes, systolic hypertension and insulin use were strong risk factors for the development of proliferative diabetic retinopathy. Interestingly, HbA_1C did not confer additional risk in this cohort.

Introduction

Proliferative diabetic retinopathy (PDR) is a leading cause of new cases of blindness among adults aged 20–74 years in the United States.¹ In the US, African Americans have a high prevalence of type 2 diabetes and seem to be at high risk for microvascular complications, including diabetic retinopathy (DR).² Some studies have found that African Americans have a higher risk of developing DR compared with Caucasians after adjusting for clinical risk factors.^3–7 For example, the prevalence of moderate non-proliferative DR or worse was higher for African American veterans than for Caucasian veterans in the Veterans Affairs Diabetes Trial of type 2 diabetes, and could not be accounted for by traditional risk factors.⁵

Risk factors for developing any DR have previously been well delineated in many studies.^{3, 5, 8–17} Longer diabetes duration, hyperglycemia and hypertension are the most consistent risk factors for DR.¹⁸ Of studies that have reported DR risk factors in various ethnicities, few studies have reported risk factors for PDR in selected subsets of these large cohorts.^{14, 16, 19} Risk factors for progression to PDR in particular are of importance given the high risk of profound vision loss in these patients, and there is evidence to suggest that risk factors for advanced DR stages may differ in some respects from risk factors for any DR.²⁰

The number of studies that have examined DR risk factors in African Americans is more limited (Table 1). Among these studies, most have not reported risk factors for PDR specifically. Those that have were studies performed in either exclusive type 1 diabetes populations or mixed type 1 and type 2 diabetes populations.^{17, 19, 21} To the best of our knowledge, risk factors for developing PDR specifically (as opposed to any retinopathy) in an exclusively African American population with type 2 diabetes have not been described. In this report, we explore this issue in a case-control study to determine the personal and demographic risk factors for PDR in African Americans with type 2 diabetes.

Table 1.

Previous studies examining risk factors for diabetic retinopathy in African Americans

Study	Diabetes type	AA with PDR, n	Risk factors for PDR reported	Risk factors for PDR reported in AA separately
Davis et al 199819	T1D, T2D	176*	Yes	No
Arfken et al 199821	T1D	17	Yes	No
Harris et al 19988	T2D	5*	No	No
Roy 200017, 30	T1D	137	Yes	Yes
Klein et al 200211	T1D, T2D	16	No	No
Klein et al 20023	T2D	2*	No	No
Emanuele et al 20055	T2D	<74†	No	No
Wong et al 200610	T2D	26‡	No	No
Zhang et al 20109	T1D, T2D	28‡	No	No
Current study	T2D	290	Yes	Yes

^*The exact number of patients with PDR was not reported; this approximation is based on percentage of AA patients and PDR participants

^†74 was the number of AA participants with moderate non-proliferative diabetic retinopathy or worse; this study did not report the exact numbers or percentages of patients with PDR

^‡Number of patients with vision-threatening retinopathy which includes severe non-proliferative diabetic retinopathy, clinically significant macular edema and PDR.

AA, African American; PDR, proliferative diabetic retinopathy; T1D, type 1 diabetes; T2D, type 2 diabetes

Materials and Methods

Participants and retinopathy assessment

The Institutional Review Boards of the University of Mississippi Medical Center, Massachusetts Eye and Ear Infirmary, and Boston Medical Center approved this study, and all participants gave written informed consent after explanation of the nature and possible consequences of the study. All procedures conformed to the tenets of the Declaration of Helsinki.

Participants for the African American Proliferative Diabetic Retinopathy Study were recruited between 2011 and 2013 from 4 clinical sites: University of Mississippi Medical Center, Massachusetts Eye and Ear Infirmary, Boston Medical Center and Harvard Vanguard Medical Associates. All participants self-identified as African American and had a known diagnosis of type 2 diabetes by 2003 American Diabetes Association criteria²² and/or by being on anti-diabetic medication. To be enrolled in the study, participants had to meet either the case or control definition. Cases were patients with PDR either in 1 or both eyes. Controls were patients with no DR in either eye and at least 10 years of diagnosed diabetes. Dilated, digital 7-standard field fundus photography of both eyes was obtained using a Topcon TRC 50 DX camera (Topcon, Tokyo, Japan). Photographs were graded for degree of DR by 2 independent, masked ophthalmologist-investigators. Level of retinopathy was scored using the Early Treatment Diabetic Retinopathy Study (ETDRS) adaptation of the modified Airlie House classification.²³ Intergrader reliability was determined by percent agreement and weighted kappa. For ETDRS level, the percent agreement and weighted kappa were 96.5% and 0.90, respectively. This indicated substantial intergrader reliability.²⁴ Disagreements were arbitrated by a third masked ophthalmologist-investigator. If there was still disagreement after arbitration, the final grade was determined by joint review of the ophthalmologist-investigators.

Covariate data

The covariates examined in this study were age, sex, site of recruitment, duration of diabetes, hemoglobin A_1C (HbA_1C), systolic blood pressure (BP), diastolic BP, body mass index (BMI), waist circumference, total cholesterol, low density lipoprotein (LDL) cholesterol, high density lipoprotein (HDL) cholesterol, triglycerides, smoking status, and insulin use. These covariates have been implicated as risk factors for PDR in studies of type 2 diabetes patients of varied ethnicities.^{14, 16, 19, 25} Covariate data were collected by uniform methods at the study visit or from medical records. Using a standardized questionnaire, duration of diabetes, history of smoking, insulin use, and oral hypoglycemic agent use were recorded. Duration of diabetes was verified by review of medical records. For each participant, a blood sample was sent for HbA_1c measurement at a centralized laboratory. Blood was also collected for genetic analyses to be completed at a later date. Resting, seated BP was measured 3 times and the mean was used in the analyses. Standing height, weight, and waist circumference were recorded at the study visit. BMI was calculated as weight in kilograms divided by height in meters squared. 1 set of fasting lipid levels (total, LDL, and HDL cholesterol and triglycerides) closest to the study visit but within 1 year of the study visit were recorded from medical records.

Statistical analyses

Sample size calculations

Sample size for this study was determined a priori based on preliminary data regarding the number of patients seen at the participating clinics who would meet the case and control definitions and review of the medical literature regarding the expected means and standard deviations for the risk factors examined. With 70 controls and 300 cases, we would have between 93% and 100% power to detect the expected differences for the most consistent risk factors; duration of diabetes, HbA_1c, and BP.

Outcome and covariate definitions

Retinopathy status was determined based on the eye with the higher ETDRS level. Absence of retinopathy was defined as ETDRS level <14. PDR was determined as ETDRS level ≥60. If 1 eye was ungradable, the score for the other eye was used.

For statistical analyses, duration of diabetes, HbA_1c, mean systolic BP, mean diastolic BP, BMI, waist circumference, serum fasting total cholesterol, HDL cholesterol, LDL cholesterol, and triglycerides were evaluated as continuous variables for maximal power. Tobacco use was evaluated as a dichotomous variable; ever (past or current) smoker vs never smoker. Insulin use was also evaluated as a dichotomous variable; taking insulin vs not taking insulin.

Univariate and multivariate analyses

To examine the variables that have previously been associated with PDR and determine their association with PDR in the African American Proliferative Diabetic Retinopathy Study sample, we used the t-test for continuous variables and chi-square test for dichotomous variables. For all analyses, we used the subset of participants with complete information for the covariate of interest in that particular analysis to maximize the generalizability and power of the analysis.

Variables that differed significantly in the t-test and chi-square analyses (P<0.05) were considered for inclusion in the logistic regression models. All logistic regression models included age, sex and study site a priori. Variables were considered sequentially with measures of diabetic severity considered first. Collinearity and confounding were assessed by changes in the effect estimates and standard errors of the existing variables in the model. Collinear variables were excluded from the final models. Confounders were retained if they changed the effect estimate for an existing variable by 10% or more. All analyses were performed using Stata/IC 12.1 (StataCorp, College Station, TX, USA).

Results

There were 358 participants enrolled in this study; 68 participants had no diabetic retinopathy (ETDRS grade <14) despite at least 10 years of type 2 diabetes and qualified as controls, and 290 participants had PDR (ETDRS grade ≥60) and were designated cases. These numbers were very close to the a priori enrollment goals for the study. Of the 290 participants with PDR, 212 (73%) had PDR in both eyes, 43 (15%) had PDR in 1 eye and the contralateral eye was ungradable, and 35 (12%) had PDR in 1 eye with a lesser degree of retinopathy in the contralateral eye.

Table 2 shows characteristics of participants with PDR compared to those with no DR. Participants with PDR were more likely to have a longer duration of diabetes (P<0.001), higher HbA_1c (P=0.001), higher systolic BP (P<0.001), and be using insulin (P<0.001). After adjusting for age, sex and site, these associations remained significant for the individual risk factors (Table 3). In the full multivariate logistic regression model, duration of diabetes, systolic BP, and insulin use remained independent risk factors for PDR (all P<0.001), while HbA_1c was no longer statistically significant (Table 3). The odds of having PDR were 62% higher for every 5 years of diabetes duration and 65% higher for every 10mmHg systolic BP. Insulin use was associated with a >6-fold risk of PDR.

Table 2.

Risk factors in type 2 diabetes African American participants with and without proliferative diabetic retinopathy, USA

Variable	No DR		PDR		P value*
	n	% or mean (SD)	n	% or mean (SD)
Age, years	68	61.3 (9.3)	290	59.5 (11.2)	0.21
Female	68	67.6	290	60.3	0.27
Duration of diabetes, years	68	15.3 (5.4)	290	21.2 (10.3)	<0.001
Hemoglobin A1C, %	68	7.6 (1.9)	289	8.5 (2.1)	0.001
Body mass index, kg/m2	68	34.2 (8.4)	288	33.6 (7.4)	0.55
Waist circumference, cm	68	104.6 (18.1)	276	106.4 (17.0)	0.46
Systolic BP, mmHg	68	136.1 (17.7)	290	152.1 (22.9)	<0.001
Diastolic BP, mmHg	68	81.0 (11.0)	290	82.1 (13.9)	0.54
Total cholesterol, mg/dL	52	173.7 (35.8)	215	179.8 (52.0)	0.42
Triglycerides, mg/dL	51	140.3 (70.6)	211	131.4 (85.5)	0.49
LDL cholesterol, mg/dL	52	97.0 (29.2)	211	102.6 (44.6)	0.39
HDL cholesterol, mg/dL	52	50.0 (19.0)	214	51.9 (18.3)	0.52
Ever smoked	68	36.8	290	35.2	0.81
Use of insulin medication	68	32.4	290	76.2	<0.001

DR, diabetic retinopathy; PDR, proliferative diabetic retinopathy; BP, blood pressure; SD, standard deviation; LDL, low density lipoprotein; HDL, high density lipoprotein

^*Chi-square test used for dichotomous variables, t-test used for continuous variables.

Table 3.

Logistic regression analysis to study the effect of risk factors for proliferative diabetic retinopathy controlling for age, sex and site in African American participants, USA

Variable	Univariate analysis		Multivariate analysis
	OR (95% CI)	P value	OR (95% CI)	P value
Duration of diabetes*	1.58 (1.31–1.91)	<0.001	1.62 (1.28–2.03)	<0.001
Hemoglobin A1C	1.31 (1.10–1.55)	0.002	1.04 (0.88–1.23)	0.68
Systolic blood pressure†	1.47 (1.27–1.70)	<0.001	1.65 (1.37–1.97)	<0.001
Insulin treatment	6.63 (3.71–11.8)	<0.001	6.65 (3.33–13.3)	<0.001

OR, odds ratio; CI, confidence interval

^*OR calculated per 5 year increase in duration of diabetes

^†OR calculated per 10mmHg increase in systolic blood pressure

Discussion

In this study of African Americans with type 2 diabetes, insulin use, longer duration of diabetes, and systolic hypertension were found to be associated with the development of PDR. Participants using insulin had much higher odds of having PDR compared with those not on insulin. This association has also been found in other ethnicities and may be due to the fact that participants using insulin have more severe diabetes and poorer glycemic control.^{16, 25} Although the association with insulin use did remain significant after adjustment for HbA_1c, it should be noted that HbA_1c was based on a 1-time draw at the time of fundus photography, and average levels over an extended period of the participants’ diabetes were not available.

The findings that longer duration of diabetes and higher systolic BP are associated with PDR are consistent with previous studies in various ethnicities and of both diabetes types.^{14, 16–19, 25–27} African Americans in our cohort had a 65% increased risk of developing PDR for every 10mmHg increase in systolic BP. These findings underscore the need for patient education among African Americans with type 2 diabetes about the importance of BP control in preventing blindness from advanced DR. Prospective randomized controlled trials in patients with type 2 diabetes have shown the benefit of BP control on progression of retinopathy in general.^{28, 29} Although these trials have been executed in predominantly Caucasian populations, 1 study enrolled 30% non-white patients which included African Americans along with other non-white ethnic minorities.²⁹ A previous study conducted in African Americans with type 1 diabetes also found elevated BP, along with glycemic control and renal disease, to be associated with PDR.³⁰ We did not assess renal disease in the current study. We did not find age, sex, diastolic BP, serum lipid levels, BMI, waist circumference or smoking status to be independently associated with PDR in African Americans with type 2 diabetes. While there are some reports of these factors being associated with PDR in other populations,^{16, 19, 25, 31, 32} these have not been consistently found to impact development of PDR overall.

Randomized clinical trials in patients with type 2 diabetes have consistently shown that glycemic control is crucial to reducing the risk of DR progression.^{28, 29} Poorer glycemic control, as measured by a higher HbA_1c level, was associated with PDR in our univariate analyses but did not reach statistical significance in the multivariate model. This same finding has been observed in a similar PDR study in a Latino population.²⁵ The association between HbA_1c and DR may weaken with more advanced forms of retinopathy. One explanation for this phenomenon is that the diagnosis of PDR can be the impetus for tighter regulation of blood sugar control. Patients with PDR may be motivated to achieve better HbA_1c levels after receiving this diagnosis and their HbA_1c after PDR diagnosis may not be reflective of their glycemic control for most of the duration of their diabetes. There is conflicting evidence as to whether retinopathy begins to appear at lower HbA_1c levels in African Americans.^{33, 34} These population-based studies on ethnic differences in HbA_1c have focused on prevalence of any DR and not on advanced DR. It is not known whether these potential ethnic-specific differences might affect the relationship between HbA_1c and PDR in African Americans.

This investigation’s strengths include DR phenotyping with dilated, 7-field photography. The case and control criteria were strictly defined to enrich the study for participants at the 2 extremes of the disease spectrum and minimize misclassification bias. Most covariates were measured using standardized protocols at the same time point as fundus photography. The study fell only slightly short of its recruitment goals, and the power to detect the associations for the most consistent risk factors (diabetes duration, HbA_1c, and systolic BP) still ranged between 93% and 100% for the recruited cohort. The case to control ratio was 4:1. The power of this study is equivalent to that of a 220-participant study with a 1:1 case to control ratio.

There are some limitations to our study. Although most covariates were collected prospectively at the study visit and had few missing data points, we collected serum lipid measurements retrospectively and there was a higher rate of missing data for these variables. This could limit the power to detect associations with serum lipids and creates a potential selection bias, although comparison of participants missing lipid measurements compared with those with lipid measurements did not reveal any significant differences in duration of diabetes, HbA_1c or systolic BP. We did not have data on time to requiring insulin which might be an important covariate with shorter time to requiring insulin indicating longer, undiagnosed type 2 diabetes duration. The cross-sectional design of our study does not allow us to judge the causal or temporal relationships of the associations found. Caution must be taken in comparing risk factor results in our study to previous studies as these differences might be accounted for in part by differences in retinal photography protocols and/or definitions of retinopathy. Our sample is clinic-based rather than population-based and this limits the generalizability of the results to the overall African American population.

In summary, our study shows that longer duration of diabetes, increased systolic BP and insulin use are strong risk factors for PDR in an African American population with type 2 diabetes. To the best of our knowledge, this is the first study reporting the risk factors specifically for the development of PDR in an exclusively African American population with type 2 diabetes. These data may be useful to future investigations, particularly because African Americans have the highest prevalence of diabetes as a percentage of the adult population in the US and this prevalence is projected to increase through to 2050.³⁵