Article Text

Download PDFPDF

Prevalence and determinants of undiagnosed diabetic retinopathy and vision-threatening retinopathy in a multiethnic Asian cohort: the Singapore Epidemiology of Eye Diseases (SEED) study
  1. Olivia S Huang1,
  2. Wan Ting Tay1,
  3. Peng Guan Ong1,
  4. Charumathi Sabanayagam1,
  5. Ching-Yu Cheng1,
  6. Gavin S Tan1,
  7. Gemmy C M Cheung1,
  8. Ecosse L Lamoureux1,2,
  9. Tien Y Wong1,2,3
  1. 1Singapore Eye Research Institute, Singapore National Eye Centre, Singapore, Singapore
  2. 2Center for Eye Research Australia, University of Melbourne, Melbourne, Royal Victorian Eye and Ear Hospital, Victoria, Australia
  3. 3Department of Ophthalmology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore
  1. Correspondence to Professor Tien Y Wong, Singapore Eye Research Institute, Singapore National Eye Centre, 11 Third Hospital Avenue, Singapore 168751, Singapore; ophwty{at}


Objective To determine the prevalence and risk factors of undiagnosed diabetic retinopathy (DR), in particular vision-threatening DR (VTDR) in a multiethnic Asian cohort.

Design A population-based survey of 3353 Chinese, 3280 Malays and 3400 Indians (73.6% response) aged 40–80 years residing in Singapore. Diabetes mellitus (DM) was defined as random glucose ≥11.1 mmol/L, use of diabetic medication or a previous physician diagnosis. DR severity was graded from retinal photographs following the modified Airlie House classification. VTDR was defined as the presence of severe non-proliferative DR (NPDR), proliferative DR (PDR) or clinically significant macular oedema (CSMO), using the Eye Diseases Prevalence Research Group definition. Participants were deemed ‘undiagnosed’ if they reported no prior physician diagnosis in structured interviews, in those with the condition.

Results Of 10 033 participants, 2376 had DM (23.7%), of which 805 (33.9%) had DR. Among 2376 with DM, 11.1% (n=263) were undiagnosed. Among 805 with DR, 671 (83.3%) were undiagnosed. Among 212 with VTDR, 59 (27.3%) were undiagnosed. In multivariate models, factors associated with undiagnosed VTDR were higher low-density lipoprotein (LDL) cholesterol (OR=1.53, 95% CI 0.99 to 2.35, p=0.05) and absence of visual impairment or blindness in any eye in terms of best-corrected vision OR=3.00, 95% CI 1.47 to 6.11, p=0.003).

Conclusions In this community, a quarter with VTDR is undiagnosed, and 8 in 10 with any DR are undiagnosed, compared with only 1 in 10 with DM undiagnosed. These findings suggest that screening for diabetes is successful, while screening for DR is currently inadequate in our population. Public health strategies to aid early diagnosis of DR in Singapore are urgently warranted to reduce blindness due to diabetes.

  • Epidemiology
  • Retina
  • Public health
View Full Text

Statistics from


Diabetic retinopathy (DR) is the leading cause of preventable blindness in working-age adults.1 Of 246 million worldwide with diabetes mellitus (DM), one-third has DR, and further one-third of these has vision-threatening DR (VTDR), defined as severe non-proliferative DR (NPDR), proliferative DR (PDR) or clinically significant macular oedema (CSMO).1 Early treatment modalities can, however, decrease the risk of blindness.1 Timely laser treatment reduces (a) the risk of severe visual loss in PDR by 50% over 5 years, (b) risk of progression of severe NPDR to high-risk PDR by half and (c) risk of moderate visual loss from CSMO by half.1 Furthermore, delayed treatment with intravitreal ranibizumab for CSMO results in lesser visual improvement compared with timely treatment.2

While the implications of undiagnosed DR are profound, data on the prevalence of undiagnosed DR and VTDR remain scarce.3 ,4 We aimed to elucidate the prevalence and determinants of undiagnosed DR, and in particular VTDR, in a multiethnic community in Singapore.


The Singapore Epidemiology of Eye Diseases Study is a population-based cross-sectional study of 3353 Chinese, 3280 Malays and 3400 Indians aged 40–80 years, conducted in Singapore from 2004 to 2011. Study methodology has been described elsewhere.5 In brief, an age-stratified random sampling strategy was used to select 6752 Chinese, 5600 Malay and 6350 Indian names from the Ministry of Home Affairs. Of which, 4605 Chinese, 4168 Malays and 4497 Indians were deemed eligible to participate. ‘Ineligible’ persons were those who had moved from the residential address, had not lived there in the past 6 months or were deceased or terminally ill. A total of 3353 Chinese, 3280 Malays and 3400 Indians participated, giving response rates of 72.8%, 78.7% and 75.6%, respectively (total response rate: 73.6%). Non-participants were slightly older than participants (p<0.001). The study adhered to the Declaration of Helsinki, and ethics approval was obtained from the Singapore Eye Research Institute (SERI) Institutional Review Board. Ocular examinations, systemic investigations and interviews were performed at SERI. Written consent was obtained from all participants.

Assessment of DM

We included all diabetic individuals, defined as having random glucose ≥11.1 mmol/L, use of diabetic medication or a participant-reported physician diagnosis of DM.

Assessment of DR

DR was assessed through standardised retinal photography, using a digital non-mydriatic retinal camera (Canon CR-DGi with 10D SLR back, Japan). After pupil dilation, two retinal photographs, centred at the optic disc and macula, were taken from both eyes. Photographs were graded at the University of Sydney by one certified ophthalmic grader, with adjudication by a senior retinal specialist. DR severity was graded based on the modified Airlie House classification system, using the Blue Mountains Eye Study protocol.6

DR was considered present if characteristic lesions (microaneurysms, haemorrhages, cotton wool spots, intraretinal microvascular abnormalities, hard exudates, venous beading, new vessels) as defined by the Early Treatment Diabetic Retinopathy Study (ETDRS)7 were found. Disease severity was based on the eye with more severe DR changes, and categorised as minimal NPDR, mild NPDR, moderate NPDR, severe NPDR and PDR. Diabetic macular oedema (DMO) was defined as exudates in the presence of microaneurysms and haemorrhages within one disc diameter of the fovea or presence of focal photocoagulation scars in the macular area.7 CSMO was defined as macular oedema within 500 µm of the fovea or presence of focal photocoagulation scars in the macular area. VTDR was defined as presence of severe NPDR, PDR or CSMO, using the Eye Diseases Prevalence Research Group definition.8


A standardised questionnaire was administered in English or Malay/Tamil/Mandarin, depending on participants' preferences, by trained interviewers at the same sitting as the ocular and systemic examination.

Participants were asked:

  1. ‘Has a doctor advised you that you have diabetes (high sugar in the blood/urine)?’

  2. ‘Have you ever been told by a doctor that you have eye disease or eye damage related to your diabetes (DR)?’

  3. ‘Have you ever had laser treatment for your diabetic eye disease?’

A participant was deemed undiagnosed for DM and DR if he/she did not answer ‘yes’, among those with the respective condition. A participant was deemed undiagnosed for VTDR if he/she did not answer ‘yes’ to question 2, and if there were no laser scars in retinal photography, in those with VTDR.

With regard to eye symptoms, participants were asked: ‘Are you aware of a deterioration of vision in one or both eyes?’

Participants who answered ‘yes’ were asked for the duration of visual deterioration in their right and/or left eye.

Other questionnaire data included ethnicity, education, income, literacy, occupation, housing, family history, smoking history, DM medication and duration of DM. Housing data were classified using Singapore's public housing authority, the Housing Development Board (HDB). A low socioeconomic status (SES) was defined as meeting criteria of having primary or lower education, and individual monthly income <S$2000. Participants using visual aids were asked how often they visited any optometrists, opticians or ophthalmologists to check their visual aids.

Assessment of systemic parameters and blood investigations

Height was measured using a wall-mounted measuring tape, weight was measured using a digital scale (SECA, model 782 2321009; Vogel & Halke, Germany) and body mass index (BMI) was calculated accordingly. Blood pressure (BP) was measured with an automatic BP monitor according to the protocol used in the multiethnic study of atherosclerosis.5 Blood samples were collected to assess non-fasting glucose, glycosylated haemoglobin (HbA1c), creatinine and cholesterol. Suboptimal HbA1c was defined as HbA1c ≥7%, based on the American Diabetes Association Guidelines.9

Assessment of visual impairment

Visual acuity (VA) was measured using a logarithm of the minimum angle of resolution (LogMAR) number chart (Lighthouse International, New York, USA) at 4 m, in both eyes. Best-corrected VA (BVA) was defined as best obtainable refraction. We used the modified WHO definition of visual impairment10 as LogMAR >0.48 (20/60) to ≤1.30 (20/400) and blindness as LogMAR >1.30 (20/400).

Statistical analysis

Analyses were performed in SPSS V.16 (SPSS, Chicago, Illinois, USA). Baseline characteristics of participants were evaluated. Age-standardised prevalence was calculated based on Singapore Census of Population 2010 figures. Age–gender and multivariable-adjusted ORs and 95% CIs were calculated for factors associated with undiagnosed VTDR and DR using logistic regression models.


Among 2385 diabetic individuals, 2376 (99.6%) provided data on a prior diagnosis of DM or DR. A total of 805 (33.9%) had DR, of which 212 (26.3%) had VTDR. Those with DR included minimal to moderate NPDR in 657 (81.6%), severe NPDR or PDR in 148 (18.4%) and CSMO in 103 (12.8%). Among 805 with DR, 43 (5.3%) had undiagnosed DM. Among 770 (94.7%) with diagnosed DM, 626 (81.3%) had undiagnosed DR.

The age-standardised prevalence of DM was highest in Indian participants (26.69%), followed by Malay (19.53%) and Chinese participants (12.89%), p<0.001. There was no significant difference in age-standardised prevalence of DR and VTDR across the ethnic groups, and this ranged from 30.30% to 36.88% and from 25.82% to 27.72%, respectively (table 1).

Table 1

Prevalence of DM, DR and VTDR by cohort study

The numbers of diabetic individuals were highest in the Indian cohort (47.3%), followed by the Malay (32.3%) and Chinese cohorts (20.5%). Correspondingly, the numbers of those with DR and VTDR were also highest in the Indian cohort (49.1% and 47.7%, respectively), compared with the Malay (33.1% and 34.6%, respectively) and Chinese cohorts (17.8% and 17.8%, respectively) (table 2). The Indian cohort had a longer mean duration of diabetes compared with the Chinese and Malay cohorts and also higher levels of diastolic BP, HbA1c, total cholesterol and BMI compared with the Chinese cohort (table 3).

Table 2

Baseline characteristics of participants

Table 3

Characteristics, by ethnicity, in SEED diabetic cohort

Among individuals with DR and VTDR, the mean (SD) age was 62.0 (9.2) and 61.7 (8.3) years, respectively; slightly less than half (49.1% and 43.9%, respectively) were of male gender, and more than three-quarters (72.1% and 78.4%, respectively) were of low SES. The mean (SD) duration of DM in those with DR and VTDR was 13.5 (9.7) and 16.8 (10) years, respectively, and 48.8% and 61.2% respectively, were aware of a deterioration in vision. Only 28.0% and 30.9% of participants with DR and VTDR, respectively, reported having annual eye examinations to check their glasses. Among diabetic individuals with visual aids, only 24.7% attended yearly eye checks (table 2).

Among those with VTDR, 27.3% (n=59) were undiagnosed (figure 1). Among individuals with any DR, 83.3% (n=671) were undiagnosed. This was similar across the three ethnic groups (p=0.77). Among individuals with DR who were aware of a visual deterioration, 78.1% (n=303) were undiagnosed. Among individuals with DM, 11.1% (n=263) were undiagnosed (figure 2). The leading causes of visual impairment in diabetic individuals were cataract and DR (figure 3).

Figure 1

Proportions (%) with undiagnosed DR. DR, diabetic retinopathy; BVA, best-corrected visual acuity; NPDR, non-proliferative DR; PDR, proliferative DR; VA, visual acuity.

Figure 2

Proportions (%) with undiagnosed diabetes mellitus (DM). BVA, best-corrected visual acuity; DR, diabetic retinopathy; NPDR, non-proliferative DR; PDR, proliferative DR; PVA, presenting visual acuity; VA, visual acuity.

Figure 3

Breakdown of cause of visual impairment (BVA in better eye) in diabetic individuals. AMD, age related macular degeneration; BVA, best-corrected visual acuity.

Participants with undiagnosed VTDR had higher low-density lipoprotein (LDL) cholesterol and less visual impairment or blindness (BVA) than those who were diagnosed (table 4). Participants with undiagnosed DR were older, had a shorter duration of DM, higher LDL cholesterol, less VTDR and less visual impairment or blindness (BVA) than those who were diagnosed (table 5).

Table 4

Factors related to undiagnosed VTDR (age–gender and multivariable-adjusted analyses)

Table 5

Factors related to undiagnosed DR (multivariable-adjusted analysis)


Our study provides population-based data on the prevalence and determinants of undiagnosed DR and VTDR in a multiethnic cohort in Singapore. Eight in 10 with DR were undiagnosed, and importantly, over a quarter with VTDR were undiagnosed. These data suggest that diabetic individuals are not receiving timely management of their DR, even at advanced stages of disease.11 Public health efforts towards achieving an early diagnosis of DR, such as improving awareness and eye-care usage, should be a key target.

Our study is the first to report on the prevalence of undiagnosed VTDR. Only one other study has published rates of those unaware of their DR—the 2005–2008 National Health and Nutrition Examination Survey (NHANES) in the USA (figure 3), but it did not include data on VTDR. In the NHANES study, 73% of 345 individuals with DR12 and 55.3% of 48 with DMO13 were unaware of their condition, which is lower than the 83.3% and 67.1% with undiagnosed DR and DMO, respectively, in our study.

The reasons underpinning undiagnosed DR are multifactorial. Intuitively, less visual impairment or blindness was correlated with undiagnosed DR and VTDR in our study. This likely indicates that patients become aware of their DR only when they develop visual impairment at advanced stages of their DR. To support this, in the NHANES study12 and ours, a lesser severity of DR and a shorter duration of DM were associated with unawareness of DR and undiagnosed DR, respectively.

In our study, higher levels of LDL cholesterol were also associated with undiagnosed VTDR. This result may indicate that patients who manage their general health and have well-controlled cholesterol levels may have more regular eye examinations, compared with those with poorer control of their cholesterol levels. The Victorian Population Health Survey in Australia similarly showed that diabetic individuals were more likely to have eye checks within the last 2 years if they had health checks, including checks not related to DM, within the same time.14

The prevalence of undiagnosed DM was low (11.1%) in our study, whereas the prevalence of undiagnosed DR (83.3%) and VTDR (27.3%) was high. This suggests that while screening for diabetes is successful, screening for DR is currently inadequate in our population. Eye-care usage rates were low in our study, with only 24.7% of diabetic individuals with visual aids attending yearly eye checks. This is much lower than that of the USA, where a meta-analysis15 in 2013 showed that annual eye-care rates ranged from 64% to 72%. Another surprising finding in our study was that 78.1% of those who were aware of their visual deterioration had undiagnosed DR, and were likely not accessing eye-care services.

While our study did not examine reasons for low rates of adherence to screening guidelines, other studies have shed light on this area. A study in Kuala Lumpur, Malaysia, in 2011 showed that 43.8% were unaware how frequently they should check their eyes.16 This emphasises the need for improved awareness on screening guidelines. Another study in Myanmar17 showed that despite 92% of diabetic individuals being aware that they should visit an ophthalmologist, 34% felt this need only when they developed trouble with their eyesight, and only 57% had seen an ophthalmologist. Another study in Egypt11 on patients presenting with advanced DR showed that 60.2% were not aware that DM could be sight-threatening until they developed sight-threatening complications. This represents a lack of understanding of the nature of DR. Patient education plays a major role in adherence to screening guidelines, and the media and government have key roles in this.

Other studies in the USA showed that reasons for non-adherence to screening guidelines included patients’ practical knowledge of diabetes18 and their apathy,19 limited mobility,19 lack of insurance,19 shorter diabetes duration,18 younger age,18 last eye examination performed by a non-ophthalmologist18 and availability of ophthalmologists.12 In Singapore, there are 26 ophthalmologists per million population,20 and DR screening services are available in at least 50 primary healthcare facilities since 2010, a number expected to increase by another 90 by 2015.21 In light of relatively high numbers of eye-care services available in Singapore, further studies are needed to investigate reasons for non-adherence to screening guidelines.

In our study, the numbers of individuals with DR/VTDR were higher in the Indian cohort, compared with the Chinese and Malay cohorts. This is likely due to higher numbers of diabetic individuals and a longer mean duration of DM in the Indian cohort compared with the Chinese and Malay cohorts and also higher levels of diastolic BP, HbA1c, total cholesterol and BMI in the Indian compared with the Chinese cohort. Other studies worldwide have similarly found that the prevalence of DR/VTDR is higher in South Asians compared with the white population,22 and attributed this to a younger age of onset of DM, obesity, insulin resistance and genetic predisposition.22

The strengths of our study include its large population-based sample, masked external grading of retinal photographs and detailed investigation of risk factors. Limitations include a single measurement of random glucose being used to define DM, which may have resulted in misclassification of DM status.23 Assessment of DR was based on two-field retinal photography, while ETDRS grading of DR is based on seven-field photography. However, two-field retinal photography has a high sensitivity and specificity for screening for any retinopathy (95.7% and 78.1%, respectively)24 and for VTDR (95% and 99%, respectively).25 CSMO was defined based on retinal photographs, without optical coherence tomography, which may have underdiagnosed the cases. We had no access to medical records, and may have overestimated the numbers of undiagnosed individuals, since this was determined based on participants’ self-report. In addition, only participants who used visual aids were asked how often they visited an eye-care provider to check their glasses. We did not have data on rates of eye-care usage in participants not using visual aids. We also did not have data on previous intravitreal injections for VTDR.

In summary, a timely diagnosis of VTDR is necessary to institute sight-saving treatments, and an early diagnosis of DR is a key starting point for patients in receiving surveillance and treatment. More than a quarter of those with VTDR were undiagnosed in Singapore, and 8 in 10 with DR were undiagnosed, compared with only 1 in 10 with DM who was undiagnosed. In light of effective treatment for DR being well-established for more than a quarter of a century, diabetic individuals are going blind not due to a lack of treatment options, but due to a lack of translation of clinical services. Our efforts should be intensified towards increasing awareness and improving rates of eye-care usage.


All authors named in the title page have provided permission to be named and have met authorship criteria. The authors also acknowledge The Singapore Eye Epidemiology Diseases group, which has made substantial contributions to the work reported in this manuscript (eg, data collection), but does not fulfil authorship criteria.


View Abstract


  • Contributors Each author’s contributions are listed as follows: OSH: substantial contribution to analysis and interpretation of data, drafting the article, final approval of version to be published. WTT, OPG, CS, C-YC, GST, GCMC, ELL: substantial contribution to analysis and interpretation of data, revising the article critically for important intellectual content, final approval of version to be published. TYW: substantial contributions to conception and design of study, analysis and interpretation of data, revising the article critically for important intellectual content, final approval of version to be published, full responsibility for the conduct of the study and guarantee of the integrity of the data.

  • Funding This study was funded by Biomedical Research Council (BMRC) 08/1/35/19/550, Singapore Bio Imaging Consortium (C-011/2006) and National Medical Research Council (NMRC), StaR/0003/2008, Singapore.

  • Competing interests None declared.

  • Ethics approval Singapore Eye Research Institute Institutional Review Board.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

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.

Linked Articles

  • At a glance
    Keith Barton James Chodosh Jost Jonas