Aims: To estimate the prevalence of diabetic retinopathy (DR) and the possible risk factors associated with DR, in a population of south India.
Methods: A cross-sectional sample of subjects aged 30 years and older was selected using a cluster sampling technique from Theni district of Tamilnadu state. Eligible subjects were identified through a door-to-door survey and fasting blood glucose estimation. History of diabetes was elicited, and height, weight and blood pressure were measured for all subjects. Ocular examinations including visual acuity and anterior and posterior segment examinations were performed at preselected sites within clusters.
Results: Among the 25 969 persons screened for diabetes mellitus (DM), 2802 (10.8%) (95% CI 9.3 to 12.2%) were found to have DM. DR was detected in 298 (1.2%) of 25 969 subjects. The age–gender-adjusted prevalence of DR is 0.05% (95% CI 0.04 to 0.06%) for rural and 1.03% (95% CI 0.89 to 1.12%) for urban areas. The overall age–gender-cluster adjusted prevalence of DR was 0.74% (95% CI 0.66 to 0.83%). Diabetic retinopathy was present in 12.2% (95% CI 10.4 to 14.1%) of the DM population.
Conclusion: Adequate training of ophthalmologists in treating DR and improvement in eye-care infrastructure are needed to tackle this major public health problem in India.
Statistics from Altmetric.com
It is estimated that 79.4 million people in India will have diabetes by the year 2030.1 The rapid increase in the number of persons with diabetes is expected to lead to an increase in the number of persons with complications from diabetes. Few studies have reported the magnitude and distribution of diabetic retinopathy (DR), in India.2–5 Accurate estimates of the magnitude and distribution of diabetes and complications from diabetes are essential to plan for appropriate healthcare infrastructure development in India. This study aimed at estimating the population prevalence of DR and the possible risk factors associated with DR.
METHODS AND MATERIALS
A population-based cross-sectional study was done from August 2005 to March 2006 in the district of Theni in Tamil Nadu, south India (a geographical area of 3244 km2) covering individuals aged 30 years and older and resident of the district. Theni district is a semirural agrarian district in the southern part of Tamilnadu. It has five subdistricts: Theni, Andipatti, Bodinayakanur, Uthamapalayam, Periyakulam. We chose this district for the study, since it has a semirural structure (55% rural). Assuming a prevalence of 1% for DR (based on previous studies that reported prevalence of DR from 0.5 to 1.8% in India2–5), a relative precision of 15% with 95% confidence, a design effect of 1.5 and a dropout rate of 10%, the sample size for the study was estimated at 28 172 individuals. Twenty-four rural and 29 urban clusters (based on the proportion of individuals aged 30 years and above estimated at 35–40% of the total population6) were randomly selected from a sample frame that included all villages of the district. All full-time residents of the sampled clusters irrespective of gender who had completed 30 years of age as of the day of enumeration were considered eligible for inclusion in the study. Persons living in the same household, at least for the past 6 months, were considered as full-time residents.
Trained field staff, through a door-to-door survey, identified eligible subjects for the study. Subjects were enrolled into the study after explaining the study and obtaining informed consent. Trained field staff administered a standardised set of questionnaires that included socio-demographic details, a diabetes screening form, a risk-factor assessment form that collected details of physical activity, smoking and dietary habits, and clinical examination forms (measurement on blood pressure, anthropometry, visual acuity and cardiovascular disease (CVS)) to all enrolled subjects. The diabetes screening form includes the personal details of the subject, the details of whether a person was a known diabetic (determined by the subject’s previous medical history) and if they were diabetic, the duration of diabetes and details of the current medication they were using. These details were collected by the field worker and cross-checked by the field coordinator. Regarding diet, the data collected focused on whether the subject is a vegetarian or a non-vegetarian. Regarding smoking, the data collected focused on if the subject was a current smoker or not. We defined a person as having cardiovascular disease based on the medical history and medical records, and prescriptions.
The study protocol was approved by the Institutional Review Board, and the research adhered to the tenets of the Declaration of Helsinki. A schematic representation of the study design is shown in fig 1.
The process of fasting for blood glucose estimation was explained to the enrolled subjects. Fasting blood glucose was estimated on capillary blood samples using a glucometer (MediSence Optimum, Abbott Laboratories, Bedford, Massachusetts) and test strips (Optimum Point-of-Care Blood Glucose Test Strips). Individuals were categorised as “eligible subjects with diabetes” if the fasting blood glucose level was ⩾126 mg/dl.7 Previously diagnosed (based on medical history and prescriptions) persons with diabetes (known diabetes mellitus (KDM)) were categorised as “eligible subjects with diabetes,” irrespective of the fasting blood sugar levels.
A standard measuring tape was used to measure the height of subjects up to one decimal place. A standard platform weighing scale was used to measure weight (kg) to the nearest one decimal place correcting for the zero error. Body mass indices (BMI) were derived, and individuals were categorised as underweight (<18.5 kg/m2), normal (18.5–24.9 kg/m2), overweight (25.0–29.9 kg/m2) and obese (⩾30 kg/m2).8
A trained nurse measured blood pressure on the left arm of each individual using a sphygmomanometer (Diamond Co industrial Electronics and Allied Products, Pune, India) by a standardised technique.9 Hypertension was defined as a systolic blood pressure of >140 mm Hg and/or diastolic pressure of >90 mm Hg.
Trained ophthalmic nurses used E logMAR charts to measure presenting distance visual acuity. Fellowship trained retina specialists evaluated the anterior segment of enrolled subjects using a portable hand-held slit lamp (Heine HSL 100). Subjects suspected of having narrow angles (by van Hericks method) had their dilatation deferred and were referred to the base hospital for gonioscopy and further evaluation.
Pupillary dilatation was achieved by instilling a combination of tropicamide (0.8%) and phenylephrine hydrochloride (5%) (Tropicamet Plus sterile eye-drops, Sun Pharmaceuticals, India). The study ophthalmologists used both direct and indirect ophthalmoscopy to look for signs of DR including the presence of even one microaneurysm within the arcades; dot or blot haemorrhages, hard exudates, cotton-wool spots, intraretinal microvascular anomalies (IRMA), venous beading, neovascularisation of the disc (NVD) or elsewhere (NVE), vitreous haemorrhages with or without fibrovascular proliferation and evidence of tractional retinal detachment. DR was classified for each eye as no retinopathy (level 1), mild–moderate non-proliferative DR (NPDR, levels 1.5–3), severe NPDR (levels 4–5) and proliferative DR (PDR, levels 6–7) based on the modified classification method as described by Klein et al.10 The level of retinopathy in the worst eye was used to classify the retinopathy status of each person. Persons for whom retinal status could not be assessed due to media opacities or other reasons were referred to the base hospital for further examination.
The study questionnaire collected details on activity habits (h/day) of diabetic subjects. Activity levels at work and in household duties were considered per day, and the scores were multiplied by the number of days per week engaged in similar activity. The score varied from 1 to 70 and classified as sedentary, light, moderate and strenuous.11
Study supervisors checked the forms for completeness and internal consistency using a predetermined list of random numbers.
Between groups, comparisons for continuous and categorical variables were done using an independent t test and Pearson χ2 test respectively. A cluster-adjusted logistic regression analysis was carried out to determine the factors associated with DM and DR. Factors that had a p value <0.25 in a univariate logistic regression analysis were included in a multivariate model. A p value <0.05 was considered statistically significant in the multivariate model. All analyses were carried out using STATA version 8. (StataCorp LP, College Station, Texas). All analyses took account of the cluster design through the use of survey commands in STATA.
The study covered 28 039 persons (99.5%) of the estimated sample size (n = 28 172), and included 15 362 (54.8%) urban and 12 677 (45.2%) rural residents. The distribution did not differ significantly by gender (males = 13 887 (49.5%), females = 14 152 (50.5%)). The mean age was 47.0 (SD 12.7) years. Screening for DM was done on 25 969 (92.6%) of 28 039 enrolled subjects. The information on non-participation rate is provided in fig 1. The participation rate was similar in both urban and rural areas (92.4% vs 93.0%, p<0.06) but was significantly different between males and females (89.8% vs 95.4%, p<0.000).
DM was detected in 2802 (10.8%) of 25 969 people. This included 1324 (5.1%) individuals previously diagnosed as having DM (KDM) and 1478 (5.7%) individuals newly detected with diabetes by the study. Table 1 shows the prevalence of DM and DR and 95% CI by gender and urban/rural population. DR was present in 12.2% (95% CI 10.4 to 14.1%) of the DM population.
Table 2 shows the distribution of socio-demographic characteristics and the risk-factor analysis of DM. A cluster-adjusted analysis of the association between socio-demographic characteristics and DM status showed that the prevalence of DM was more among people aged above 45 years (adjusted OR 2.3, 95% CI 2.1 to 2.6), urban population (adjusted OR 1.4, 95% CI 1.1 to 1.8), female gender (adjusted OR 1.4, 95% CI 1.2 to 1.5), Muslim (adjusted OR 2.1, 95% CI 1.2 to 3.7), business people (adjusted OR 1.8, 95% CI 1.4 to 2.3) and graduates (adjusted OR 2.0, 95% CI 1.5 to 2.6). The age–gender-adjusted prevalence of DM was 5.6% (95% CI 5.3 to 5.9%) for rural clusters and 8.7% (95% CI 8.4 to 9.1%) for urban clusters. The overall age–gender-adjusted prevalence of DM was 6.98% (95% CI 6.73 to 7.23%).
A risk-factor assessment was done for 2531 (90.3%) of the 2802 eligible diabetic subjects. A clinical examination was performed on 2509 (99.1%) of 2531 subjects, and dilated fundus examination of at least one eye was performed on 2448 subjects, as 61 subjects were primary angle closure (PAC) suspects who could not be dilated. The retinal status of both eyes could not be determined in 12 patients due to dense cataracts.
DR was detected in 298 (1.2%) of 25 969 subjects. The age–gender-adjusted prevalence of DR is 0.05% (95% CI 0.04 to 0.06%) for rural and 1.03% (95% CI 0.89 to 1.12%) for urban areas. The overall age–gender-adjusted prevalence of DR was 0.74% (95% CI 0.66 to 0.83%). Higher rates of DR were found among older age, male, urban, graduates, KDM, duration of DM>5 years, sedentary–light physical activity vegetarians and normal BMI category. The mean blood glucose level (217.7 (79.8) vs 178.2 (66.0), p<0.001) and mean systolic blood pressure (SBP) (138.0 (21.3) vs 130.4 (18.5), p<0.001) were significantly high among DR subjects. The risk factors for DR are presented in table 3.
Since there was a significant difference observed between KDM and NDM subjects (21.7% vs 3.1%, p<0.000), a subset analysis was constructed to identify independent factors associated with DR among KDM (Table 4). Among the 1242 KDM subjects, the information on duration of DM (mean duration 5.1 (5.0) years) was available only for 1186 subjects. The classifications of duration of DM ⩽5 and >5 years and blood glucose level ⩽160 and >160 were derived based on the median. Prediction was done using the duration of DM>5 years, SBP>140 and blood glucose level>160. A multivariate logistic regression model was used to estimate the quantitative effect of each significant risk factor with DR and to determine for each of them their relative weight (β coefficient).
The probability of developing DR will be estimated from the equation,
where Z = β0+β1X1+β2X2+…βnXn; β0—constant (−2.5); β1 = 1.31—duration of DM>5 years (1, yes; 0, no); β2 = 0.74—blood glucose level>160 (1, yes; 0, no); β3 = 0.84 SBP>140 (1, yes; 0, no); Z = −2.5+(1.31×1)+(0.74×1)+(0.84×1) = 0.39. Prob(DR) = 0.6, that is, the probability of developing DR for a person with duration of diabetes >5 years, fasting blood glucose level >160 and systolic blood pressure >140 was estimated at 60%. Keeping the blood glucose level (>160) and duration (>5 years) constant, the probability of DR was 49% when SBP was >120 and 68% when SBP was >160. Keeping the blood glucose (>160) and SBP (>140) constant, the probability of DR was 74% when the duration was >10 years.
The grading of DR was done on 298 subjects who had retinopathy in at least one eye. Of the 596 eyes of 298 subjects, DR was present in 514 (86.2%) eyes (mild–moderate NPDR, 80.0%; severe NPDR, 11.1%; PDR, 8.9%). One hundred and sixty-three (54.7%) subjects had mild–moderate NPDR, 16 (5.4%) had severe NPDR, and 14 (4.7%) had PDR in both their eyes. The remaining 105 subjects had different levels of retinopathy in both eyes.
The visual acuity results are shown in table 5. About 90.5% of mild/moderate NPDR subjects had a visual acuity of ⩾6/18 as compared with 68.9% of subjects with PDR. Only 8.7% of subjects with mild/moderate NPDR had a visual acuity in the range of 6/18 to 3/60, whereas 26.7% of PDR subjects had visual acuity in that range. Three subjects with mild/moderate NPDR had a visual acuity of <3/60, due to significant cataract. One subject who had DR but whose grade of DR could not be determined due to cataract had a visual acuity of 6/60.
The age–gender-cluster adjusted prevalence of DR was 0.74% (95% CI 0.66 to 0.83%) in our study. The prevalence of DR in the diabetic population was 12.2%, which was higher than a previous study conducted in Andhra Pradesh (7.8%)2 and was lower than the Chennai-based study (17.6%).3 These differences may be due to differences in the characteristics of the populations studied. The prevalence of DR among KDM subjects in our study (21.7%) is comparable with other population-based studies done in India2 3 but lower than the 32.4% reported in the Blue Mountains Eye Study,12 36.8% in the Beaver Dam Eye Study13 and 52% in the Melton Study.14 The prevalence of DR was 3.1% among newly detected diabetic subjects, which is slightly lower than the 5.1% reported in a recent study.2 Studies from other countries have reported a higher prevalence of DR (20–35%) among persons detected newly with diabetes.15 16 Gene–environment interactions may possibly have a role in the different risk for DR among different populations.
The urban–rural difference and the association of higher systolic blood pressure, duration of diabetes and poor glycaemic control with DR are consistent with previous reports.17–21 The association of overweight–obese diabetic subjects and DR has been reported previously22 but is contrary to the findings of the Hoorn Study.23 A previous study24 had reported that diabetic subjects have a tendency towards weight loss after the diagnosis of diabetes, and this may be the reason for a lower BMI being associated with increased prevalence of retinopathy (a longer duration of diabetes may lead to more weight loss).
The population-based design, large sample size and coverage, and evaluation by trained retina specialists are strengths of the study. The prevalence of DR reported from our study is an underestimate, since we could not ascertain the status of DR in PAC suspects at the field level. It is possible that we may have missed mild DR, since we did not use photographs for grading.25 Even though the association of smoking with DR was not statistically significant in our study, the other forms of tobacco use were not assessed, so the full effect of tobacco on DM and DR could not be investigated.
The population structure of Theni district is 55% rural and 45% urban compared with 72.2% rural and 27.8% urban for the whole of India. However, if, we extrapolate our results to India, approximately 5.8 million people may have DR in India by 2030. This extrapolation has to be viewed with caution, as our study population is not representative of the population of India; however, it indicates the need for a large number of ophthalmologists trained to provide medical and surgical retina services. Currently, few centres provide fellowship training in vitreoretinal diseases in India, and medical or surgical retina training is not part of the regular ophthalmology residency training. Eye-care infrastructure will have to develop at a rapid pace to deliver appropriate services to such a large population with diabetes and complications from diabetes.
We acknowledge the support of TIFAC-CORE in diabetic retinopathy, a project of the Department of Science and Technology, the government of India, and P Nirmalan, of Prashasa Health Consultants Pvt, Hyderabad, India in conducting the study.
Funding: This study was funded by World Diabetes Foundation (WDF), Denmark.
Competing interests: None.
Ethics approval: Ethics approval was provided by institutional review board 00002999-Aravind eye care system.
Patient consent: Obtained.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.