Ethnic Differences in the Prevalence of Diabetic Retinopathy in Persons With Diabetes When First Presenting at a Diabetes Clinic in South Africa

The CDE is a private multispecialist center based in Johannesburg, South Africa, established in 1994. It serves as the principal center in a network of 262 smaller urban and rural centers providing diabetes care services in underdeveloped communities in South Africa. Details of the diabetes management program of the CDE have been described in detail previously (13).

All persons with diabetes attending the CDE undergo routine digital retinal photography performed at the time of their first visit and annually thereafter. Digital retinal photography was conducted in a darkened room using a nonmydriatic digital camera (Canon CR6–45NM; Canon) capturing one macular centered image per eye without the use of mydriasis by one of two trained technicians, one of whom is a diabetes nurse educator. All retinal images from the patient’s first screening event obtained from 2001–2010 were independently reviewed and graded by one of three senior retinal graders according to a modified U.K. standard DR-grading protocol used by the DR Screening Service for Wales (14). Levels of DR were classified as no DR (NDR) if no lesions were detected, any DR when at least one microaneurysm and/or a blot hemorrhage were detected, and referable DR (RDR), which included preproliferative and proliferative lesions of DR as well as exudative maculopathy. RDR is the level at which further assessment by an ophthalmologist is deemed necessary.

Subjects included in this analysis were classified as having type 1 or type 2 diabetes on clinical assessment according to the American Diabetes Association classification of diabetes (9). Those individuals who did not clinically clearly fit into this classification were excluded from the analysis.

At the time of initial presentation, when the first retinal photographs were taken, blood was obtained for baseline laboratory investigations including HbA_1c, lipid analyses, and serum creatinine, and urine was collected for the assessment of the microalbumin/creatinine ratio. This initial HbA_1c was regarded as the baseline value and used in the analysis. The HbA_1c was initially analyzed as Diabetes Control and Complications Trial (DCCT) percent values and then converted to International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) mmol/mol units using the formula: [DCCT percent − 2.15] × 10.929. Both DCCT and IFCC units are dually reported throughout. Subjects were considered to have hypertension if their blood pressure was found to be >140/90 mmHg taken in the right arm seated after 5-min rest and/or if they were already on antihypertensive therapy. Other variables such as BMI were not available for analysis in this study.

Statistical analysis was conducted using SPSS version 16 (SPSS) and the population characteristics described using means and SDs for normally distributed and medians and interquartile ranges (IQ) for nonnormally distributed continuous variables and percentages for categorical variables. Significance testing used: t tests and Mann-Whitney U tests for continuous variables and χ² for categorical variables. Stepwise logistic regression analysis was used to assess the association of clinical risk factors and the presence of any DR and RDR for persons with type 1 and type 2 diabetes separately. Odds ratios (OR) and 95% CI were calculated for each. The continuous variables (HbA_1c and duration of diabetes) were stratified to avoid assumptions of linearity (i.e., for type 1 diabetes): HbA_1c <7.0 (<53), 7.0–7.9% (53–63), 8.0–8.9% (64–74), and ≥9.0% (≥75 mmol/mol) and duration of diabetes <7, 7–15, and >15 years; and for type 2 diabetes: HbA_1c <6.6 (<49), 6.6–7.4 (49–57), 7.5–8.9 (58–74) and ≥9.0% (≥75 mmol/mol) and known duration of diabetes <3, 3–8, and >8 years. Different stratifications were used in those with type 1 and type 2 diabetes for HbA_1c and duration of diabetes to ensure an equal distribution among the strata for both diabetes types. Associations were considered significant if the P value was <0.05.

A total of 5,565 subjects were seen at the CDE in Johannesburg during 2001 and 2010. The majority of people had type 2 diabetes (71.5%), with 27.6% having type 1 diabetes. The remaining 0.9% had other forms of diabetes and were excluded from the analysis. The baseline characteristics of the population studied are listed in Table 1.

In persons with type 1 diabetes, those of Caucasian origin had a longer duration of diabetes (P < 0.001), were younger at diagnosis (P < 0.001), and had a lower HbA_1c at presentation (P < 0.001) compared with the non-Caucasian population. Among the non-Caucasians, indigenous Africans had a shorter duration of diabetes compared with those of mixed race (P = 0.026) and were older at diagnosis than either the Asian Indians (P = 0.004) or those of mixed race (P = 0.004) (Table 2).

There was no evidence of DR in 60.3% (95% CI 57.8–62.7; n = 927), background DR was detected in 26% (95% CI 23.8–28.2; n = 399), and RDR in 9.2% (95% CI 7.9–10.8; n = 142). The RDR category consisted of 1.2% (95% CI 0.7–1.8; n = 18) with preproliferative DR, 4.9% (95% CI 3.9–6.1; n = 75) had exudative maculopathy, 1.3% (95% CI 0.8–2.0; n = 20) preproliferative DR with exudative maculopathy, 1.0% (95% CI 0.6–1.6; n = 15) proliferative DR, and 0.9% (95% CI 0.5–1.5; n = 14) proliferative DR with exudative maculopathy. There were 4.5% (95% CI 3.6–5.6; n = 69) unassessable images mainly due to the presence of lens opacification.

Those who presented with any or RDR compared with those with NDR were older [mean (SD)]: 38.0 (13.6) and 38.6 (12.2) years versus 32.9 (15.6) years, respectively (P < 0.001); and were younger at diagnosis of diabetes: 19.6 (11.8) and 19.0 (11.4) years versus 23.5 (14.4) years, respectively (P < 0.001), with a longer duration of diabetes [median (IQ)]: 17.0 (12.0–23.0) and 18.0 (14.0–25.0) years versus 6.0 (3.0–12.0) years, respectively (P < 0.001). Those presenting with any or referable levels of DR compared with those without DR also had a higher HbA_1c level: 8.5 (7.6–9.9)% (69 [60–85] mmol/mol]), 8.7 (7.8–10.2)% (72 [62–88] mmol/mol) versus 8.3 (7.1–9.7)% (67 [54–83] mol/mol), respectively (any DR vs. NDR, P = 0.033; RDR vs. NDR, P = 0.013) and had a higher prevalence of hypertension (25.1 and 35.9% RDR versus 12.1%, respectively; P < 0.001). There were also more cigarette smokers with any DR or RDR compared with those without DR (22.9 and 22.5% versus 17.6%, respectively; P = 0.013). (Only one P value is shown in the text, as the P value is similar for the comparison between any DR versus NDR and RDR versus NDR.)

In logistic regression analyses, the presence of any DR and RDR was significantly associated with ethnicity (Table 3). Asian Indians were at an increased risk of any DR (OR 1.78) when compared with Caucasians and adjusted for age at diagnosis, sex, duration of diabetes, HbA_1c, hypertension, and smoking status. In comparison, indigenous Africans had an increased risk of RDR (OR 3.36). There was no significant difference for any DR or RDR comparing those of mixed race with Caucasians. Other risk factors independently associated with any DR (Table 3) were a longer duration of diabetes (OR 10.28, 7–15 years; 37.31, >15 years; reference group <7 years), an increased HbA_1c (OR 1.32, <7.0% [<53 mmol/mol]; 2.15, 8.0–8.9% [64–74 mmol/mol]; and 3.20, ≥9.0% [≥75 mmol/mol]; reference group >7.0% [>53 mmol/mol]), the presence of hypertension (OR 1.44), and the habit of smoking (OR 1.78). The presence of RDR was also significantly associated with duration of diabetes, HbA_1c, hypertension, and smoking. There was a weak but significant association between albumin/creatinine ratio and any DR and RDR in univariate analysis; however, these data were missing for many of the persons with diabetes and were therefore removed from the stepwise multivariate analyses.

Caucasian subjects with type 2 diabetes were older at baseline and at the time of diagnosis (P < 0.001) and had a lower HbA_1c (P < 0.001) than non-Caucasians (Table 2). The known duration of diabetes was similar across all ethnicities. There were differences in sex distribution, with more males of Caucasian ethnicity compared with non-Caucasians (P = 0.008) and more females of mixed race compared with indigenous Africans and Asian Indians (P < 0.013). There was also a difference in age and HbA_1c levels between those of non-Caucasian ethnicity with those indigenous Africans who were older (P = 0.037) and with higher HbA_1c levels (P = 0.003) compared with Asian Indians and mixed races (Table 2).

In the entire cohort of subjects with type 2 diabetes, the prevalence of any DR was 20.5% (95% CI 19.3–21.8), with NDR detected in 74.6% (95% CI 73.2–75.9; n = 2,968). The majority of DR seen was background DR: 14.1% (95% CI 13.1–15.2; n = 561), with 6.4% (95% CI 5.7–7.2; n = 255) having RDR. The category of RDR consisted of 0.7% (95% CI 0.5–1.0; n = 28) preproliferative DR, 3.5% (95% CI 3.0–4.2, n = 141) exudative maculopathy, 1.4% (95% CI 1.0–1.8; n = 54) preproliferative DR with exudative maculopathy, 0.2% (95% CI 0.1–0.4; n = 8) proliferative DR, and 0.6% (95% CI 0.4–0.9; n = 24) proliferative DR with exudative maculopathy. There was a similar proportion with unassessable images (4.8%; 95% CI 4.3–5.6) with type 2 diabetes as seen in the subjects with type 1 diabetes.

Those presenting with any or RDR compared with those without DR were younger at diagnosis of diabetes (mean [SD]) (45.8 [11.4] years for any DR or 43.9 [10.2] years RDR vs. 51.1 [11.7] years NDR; P < 0.001), had a longer known duration of diabetes (median [IQ]) (10.0 [6.0–16.0] years for any DR or 12.0 [8.0–17.0] years RDR vs. 3.0 [1.0–7.0] years NDR; P < 0.001), and had a higher HbA_1c level (8.2 [7.1–9.7]%; 66 [54–83] mmol/mol for any DR; or 8.7 [7.6–8.7]%; 72 [60–72] mmol/mol RDR vs. 7.3 [6.5–8.6]%; 56 [48–70] NDR; P < 0.001). There was also a higher proportion of persons with hypertension (59.2% any DR or 62.4% RDR vs. 51.9% NDR; P < 0.001). There was a (P < 0.001) lower proportion of Caucasians than all other ethnic groups in those with any or RDR (16.8 and 4.4% Caucasians, 27.2 and 13.1% indigenous Africans, 27.0 and 7.5% Asian Indians, and 34.6 and 11.9% mixed race, respectively) compared with those without DR (78.9% Caucasian, 63.8% indigenous Africans, 69.4% Asian Indians, and 61.0% mixed race; Fig. 1). (Only one P value is shown in the text, as the P value is similar for the comparison between any DR versus NDR and RDR versus NDR.)

When compared with the Caucasian population, non-Caucasians had an increased risk of any DR (indigenous Africans 1.90, Asian 1.74, and mixed race 2.95) when adjusted for age at diagnosis of diabetes, sex, known duration of diabetes, and HbA_1c (Table 3). Other risk factors independently associated with any DR included a longer known duration of diabetes (OR 2.33, 3–8 years; 9.59, >8 years; reference group <3 years) and an increased HbA_1c (OR 1.32 6.6–7.4% [49–57 mmol/mol]; 1.84, 7.5–8.9% [58–74 mmol/mol]; and 2.25, ≥9.0% [≥75 mmol/mol]; reference group <6.6% [49 mmol/mol]). Non-Caucasian ethnicity increased known duration of diabetes and increased HbA_1c were also associated with RDR. Smoking was not associated with an increased risk of any DR or RDR. Although hypertension was associated with any DR and RDR in univariate analysis, it was not included in the results of stepwise multivariate analyses. Although there was a weak significant association between albumin/creatinine ratio and any DR and RDR in univariate analysis, these data were missing for the majority of patients and therefore removed from the stepwise analysis.

In a large cohort of persons with type 1 and type 2 diabetes undergoing retinal photography when first presenting at the CDE, the overall prevalence of DR and RDR at the time of the first visit was 24.6 and 7.2%, respectively. The prevalence of any DR was lower than in the recent survey of the global prevalence of any DR of 34.6% (9). This is the first study to be conducted in the privately funded sector in South Africa, which may account for this difference.

Ethnic differences in the prevalence and associations with the presence of any DR (5,7,8) and also severe/referable stages of DR (6,7) have previously been reported to be higher in non-Caucasian persons when compared with Caucasians or Europeans. Only two previous studies, in relative numbers of persons with diabetes, have examined differences between ethnic groups in South Africa (4,10). One study did not report any significant associations between ethnicity and DR (4), and the other found that those of an African and Indian origin had a significantly higher prevalence of severe DR than Europeans (10). In contrast, we observed clear differences in the risk of DR between the ethnic groups studied. Although the risk of any DR was increased in Asian Indians, RDR was increased for indigenous Africans with type 1 diabetes when compared with Caucasians, and the risk of both any DR and RDR was increased for all non-Caucasian populations compared with Caucasians with type 2 diabetes. These differences remained after correction for other risk factors, which include the fact that those of African and mixed-race origin had higher HbA_1c levels at baseline and type 2 diabetes starting at a younger age than in the non-Caucasian population. Kalk et al. (10) also reported a higher prevalence of microalbuminuria in patients of indigenous African decent when compared with a Caucasian population. This suggests that ethnic differences exist in the propensity to develop microvascular complications such as DR.

Our study demonstrated that the presence of any DR and RDR were strongly associated with increasing duration of diabetes and a higher HbA_1c level in persons with both type 1 and type 2 diabetes, thus confirming earlier epidemiological studies (15,16). Previous epidemiological studies and clinical trials also indicate that hypertension is an important modifiable risk factor for DR (17–19). In this study however, hypertension was shown to be a significant risk factor only in persons with type 1 diabetes. A reason for the lack of an association for persons with type 2 diabetes may be due to the more aggressive treatment of hypertension and lower blood pressure targets in this patient group. The importance of smoking as a risk factor for DR is inconclusive, with some studies showing either an association (20–23) or even suggesting that smoking may be protective against the development of DR (24). In our study, smoking had a clear association with the development of referable retinopathy in patients with type 1 diabetes, but not those with type 2 diabetes.

Compared with previous population-based studies in Africa and South Africa, the strength of this study was the larger sample size and that all the data (i.e., retinal images and the putative risk factors) were derived from a single center (CDE). However, as most of the population in South Africa with diabetes uses the public health system, the population studied in this paper may not therefore be representative of this majority. Although the use of standardized digital retinal photography and grading protocol are strengths within this study, the lack of dilation may have led to a higher proportion of ungradeable images, and the availability of only one 45° field per eye may have resulted in underreporting of DR.

Long-term follow-up of the majority of the participants in this study, involving intensive diabetes and risk factor management, is underway at the CDE. This will allow assessment of the incidence of both newly developing DR and the progression of DR in this cohort of subjects with both type 1 and type 2 diabetes and will be the subject of a future report.

In this large population sample of individuals with diabetes entering a diabetes management program, there was a low prevalence of DR. Ethnicity was independently associated with the presence of DR and RDR in both type 1 and type 2 diabetes. Increasing duration of diabetes and poor glycemic control were the strongest risk factors associated with the presence of any and RDR in persons with both type 1 and type 2 diabetes. In type 1 diabetes, hypertension and smoking were additional risk factors for the presence of any and RDR.

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

R.L.T., L.D., S.D.L., S.R.C., V.J.M., B.K., and D.R.O. contributed to the writing of this article and approved the final version of the manuscript. R.L.T. processed and analyzed data. L.D. was the lead for this study with the support of B.K. and D.R.O. L.D., S.D.L., B.K., and D.R.O. contributed to the concept, study design, and interpretation of data. S.R.C. contributed to the processing of data. V.J.M. collated all data. D.R.O. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Parts of this study were presented in abstract form at the 71st Scientific Sessions of the American Diabetes Association, San Diego, California, 24–28 June 2011.