Global Prevalence of Diabetic Retinopathy

and Projection of Burden through 2045

Systematic Review and Meta-analysis
Zhen Ling Teo, MBBS, MRCS (Edin),1,* Yih-Chung Tham, PhD,1,2,* Marco Yu, PhD,1,2,* Miao Li Chee, BSc,1
Tyler Hyungtaek Rim, MD, PhD,1,2 Ning Cheung, MD,1,2 Mukharram M. Bikbov, MD, PhD,3
Ya Xing Wang, MD,4 Yating Tang, MD, PhD,5 Yi Lu, MD, PhD,5 Ian Y. Wong, MD, FRCOphth,6,9
Daniel Shu Wei Ting, MD, PhD,1,2 Gavin Siew Wei Tan, MD, PhD,1,2 Jost B. Jonas, MD,7
Charumathi Sabanayagam, MD, PhD,1,2 Tien Yin Wong, MD, PhD,1,2,8,z Ching-Yu Cheng, MD, PhD1,2,8,z

Topic: To provide updated estimates on the global prevalence and number of people with diabetic reti-
nopathy (DR) through 2045.
Clinical Relevance: The International Diabetes Federation (IDF) estimated the global population with dia-
betes mellitus (DM) to be 463 million in 2019 and 700 million in 2045. Diabetic retinopathy remains a common
complication of DM and a leading cause of preventable blindness in the adult working population.
Methods: We conducted a systematic review using PubMed, Medline, Web of Science, and Scopus for
population-based studies published up to March 2020. Random effect meta-analysis with logit transformation
was performed to estimate global and regional prevalence of DR, vision-threatening DR (VTDR), and clinically
significant macular edema (CSME). Projections of DR, VTDR, and CSME burden were based on population data
from the IDF Atlas 2019.
Results: We included 59 population-based studies. Among individuals with diabetes, global prevalence was
22.27% (95% confidence interval [CI], 19.73%e25.03%) for DR, 6.17% (95% CI, 5.43%e6.98%) for VTDR, and
4.07% (95% CI, 3.42%e4.82%) for CSME. In 2020, the number of adults worldwide with DR, VTDR, and CSME
was estimated to be 103.12 million, 28.54 million, and 18.83 million, respectively; by 2045, the numbers are
projected to increase to 160.50 million, 44.82 million, and 28.61 million, respectively. Diabetic retinopathy
prevalence was highest in Africa (35.90%) and North American and the Caribbean (33.30%) and was lowest in
South and Central America (13.37%). In meta-regression models adjusting for habitation type, response rate,
study year, and DR diagnostic method, Hispanics (odds ratio [OR], 2.92; 95% CI, 1.22e6.98) and Middle East-
erners (OR, 2.44; 95% CI, 1.51e3.94) with diabetes were more likely to have DR compared with Asians.
Discussion: The global DR burden is expected to remain high through 2045, disproportionately affecting
countries in the Middle East and North Africa and the Western Pacific. These updated estimates may guide DR
screening, treatment, and public health care strategies. Ophthalmology 2021;-:1e12 ª 2021 by the American
Academy of Ophthalmology

Supplemental material available at www.aaojournal.org.

The International Diabetes Federation (IDF) estimated the lifespan of people living with DM, and lifestyle changes
global population with diabetes mellitus (DM) to be 463 leading to an increased risk for DM, a higher burden of
million in 2019 and projected it to be 700 million by 2045.1 DR and demand for eye care and treatment are expected.
As the most common and specific complication of DM,2 Thus, up-to-date and accurate estimation of the prevalence
diabetic retinopathy (DR) also is one of the leading causes of DR is critical in the formulation of health policies and for
of preventable blindness in the adult working allocation of adequate resources to address this global
population.3e6 The Global Burden of Disease Study found problem.
that in adults 50 years of age and older, DR was the fifth A previous meta-analysis on the global prevalence of DR
leading cause of blindness and of moderate and severe was conducted more than a decade ago using data up to
vision impairment.7 In particular, the age-standardized 2008 from 35 population-based studies.8 A need exists for
global prevalence for blindness resulting from diabetic eye contemporary data because several important changes
disease has increased by 14.9% to 18.5% from 1990 to regarding the epidemiologic features of DR have emerged
2020.7 With a rapidly aging global population, increasing in recent years. First, a declining trend for DR prevalence

ª 2021 by the American Academy of Ophthalmology https://doi.org/10.1016/j.ophtha.2021.04.027 1

Published by Elsevier Inc. ISSN 0161-6420/21
 Ophthalmology Volume -, Number -, Month 2021

has been suggested,9e11 especially in developed countries. Study,22 the Hong Kong Eye Study,23 and the Ural Eye and
This is likely a result of increased awareness and improved Medical Study.24 The IRB/ethics committee at each
systemic control for patients with DM.11 Second, most institution approved this study. All research adhered to the
studies included in the last meta-analysis were derived tenets of the Declaration of Helsinki. The requirement for
informed consent was waived because of the retrospective
from populations of European ancestry.8 Since 2008, a
nature of the study.
substantial increase has occurred in the number of
population-based studies in other regions, particularly in
Asia, which accounts for approximately half of the global Inclusion and Exclusion Criteria for Literature
DM population.1,12 The top 2 countries with the highest Search
number of people with DM are both in AsiadChina (116
million) and India (77 million)1dreflecting the rapid We included studies with the following criteria: (1) population-
economic growth and urbanization in Asia over the past based study; (2) clear definition on random or clustered sampling
decade with significant lifestyle and dietary changes.12,13 procedure; (3) 60% or more participation rate of the eligible pop-
ulation; (4) provided DR, VTDR, or CSME prevalence, or a
Thus, these recent data from Asia should be included to combination thereof, amongst the DM group(s); (5) provided a
provide better and contemporary estimates of the global clear definition of DM with at least 1 of the following used for DM
prevalence and burden of DR. Third, diabetic macular diagnosis: fasting blood glucose  7 mmol/l, random blood
edema is the most common form of DR causing moderate glucose of more than 11.1 mmol/l, oral glucose tolerance test re-
vision loss.10 However, the global prevalence of diabetic sults of 11.1 mmol/l or more, glycated hemoglobin findings of
macular edema has not been described previously. Data 6.5% (48 mmol/mol) or more, self-reporting of physician-
on diabetic macular edema is important for global health diagnosed DM, existing DM treatment, and medical records; and
care guidelines and resource planning, particularly in the (6) DR defined by the presence of retinal hemorrhages, micro-
context of increasing use of intraocular antievascular aneurysms, cotton-wool spots, panretinal photocoagulation laser
endothelial growth factor therapy, which may not be avail- scars, or a combination thereof found on color fundus photographs,
dilated slit-lamp examination by an ophthalmologist, or a combi-
able or accessible to all countries.14,15 nation thereof. We excluded studies that (1) were clinical trials or
To address these important gaps, we aimed to re-evaluate hospital-based or clinic-based studies, (2) were duplicates, (3) did
and re-estimate the global prevalence of DR and to provide not have full-text articles, (4) solely reported on type 1 DM in
future projections of the number of people with DR, vision- pediatric populations, and (5) had a response rate of less than 60%.
threatening DR (VTDR), and clinically significant macular Based on the above criteria, 2 reviewers (Z.L.T., Y.-C.T.) inde-
edema (CSME) through 2045. These findings are important pendently selected the studies for final inclusion. Disagreements
in the planning of DR public health policies and screening between the 2 were resolved and adjudicated by the senior author
and management strategies for DR worldwide. (C.-Y.C.).

Methods Quality Assessment

Funnel plots and Begg and Mazumdar tests (for DR, VTDR, and
Systematic Review Process CSME) were performed to determine potential publication bias
We conducted a systematic review and meta-analysis to deter- among included studies. Overall, no significant publication bias
mine the prevalence of DR, VTDR, and CSME, in accordance was found for DR (P ¼ 0.633; Fig S1, available at
with the Preferred Reporting Items for Systematic Reviews and www.aaojournal.org), but publication bias was significant for
Meta-analyses guidelines16 (Appendix A, available at VTDR (P ¼ 0.022; Fig S2A, available at www.aaojournal.org)
www.aaojournal.org). A systematic search using PubMed, and CSME (P ¼ 0.007; Fig S3A, available at
Medline, Web of Science, and Scopus was conducted to www.aaojournal.org). The 2 studies identified with potential
identify studies on DR, VTDR, and CSME prevalence. We publication bias in VTDR were the Bhaktapur Glaucoma Study
included a combination of key words such as diabetic and the Central India Eye and Medical Study; 2 other studies
retinopathy, prevalence, global prevalence, and population and identified with publication bias in CSME were the Beijing Eye
specific region and country names used in all fields including Study and the Tromso Eye Study. These studies were removed
title, abstract, and medical subject headings (Appendix B, in the VTDR and CSME analyses, respectively. After removing
available at www.aaojournal.org). We included publications in these studies, the revised funnel plots for VTDR and CSME
the English language published up to March 20, 2020. Further showed no significant publication bias (P  0.073; Figs S2B
literature search consisted of reviewing the reference lists of and S3B, respectively). The quality of included studies was
relevant articles such as previous country or region-based sys- assessed using a modified version of the risk of bias tool
tematic reviews and meta-analyses providing DR prevalence. modified from the Grades of Recommendation, Assessment,
This adopted strategy identified all articles used in previous re- Development and Evaluation approach developed by Hoy
views.8,17-19 et al25 (Appendix C, available at www.aaojournal.org). The
In addition, we further obtained original unpublished DR, tool assessed both external and internal validity via the
VTDR, and CSME data from Asian population studies via the following domains: representation of the national population,
Asian Eye Epidemiology Consortium. The Asian Eye Epide- nonresponse bias, distinction between type 1 and type 2
miology Consortium is a collaborative network of population- diabetes, DR diagnosis method, consistency in data collection,
based studies across Asia.20,21 Within this network, and year of data collection. Studies were assessed and given a
unpublished DR data were obtained from 3 epidemiologic total risk score from 0 to 6, with 0 representing good quality.
studies that fulfilled our study selection criteria and Studies with a score of 4 or more were classified as having
previously published their methodologies: the Tai Zhou Eye higher risk of bias.

2
 Teo et al 
Global Prevalence and Projections of Diabetic Retinopathy

Data Extraction Pooling of Diabetic Retinopathy Prevalence. To estimate the

regional pooled prevalence, random-effects meta-analysis was
A data extraction spreadsheet was used to collect information on performed by binomial likelihood maximization under a general-
study and participant characteristics. For each included study, we ized linear mixed model and with logit transformation.31 Pooled
extracted the following data: country of studied population, year prevalence was estimated using back transformation. Test for
that the study was conducted, habitation type (urban, rural, or heterogeneity (I2 index) was performed to determine significant
mixed), participation response rate, ethnicity of study participants differences in prevalence estimates between studies.
(European ancestry, Asian, Middle Easterner, Hispanic, or African Future Projection Estimates of Diabetic Retinopathy, Vision-
ancestry), number of patients with DM, number of patients with Threatening Diabetic Retinopathy, and Clinically Significant
DR, number of patients with VTDR, number of patients with Macular Edema in 2030 and 2045. The IDF Atlas 2019 provides
CSME, DR grading system, DM diagnostic method, and DR up-to-date estimates on current DM population data (20e79 years
diagnostic method. Current and projected global and regional DM of age) and projection of global and regional DM numbers until
numbers (up to 2045) were extracted from the IDF Atlas 2019.26 2045.1 For the projection estimates of the number of people with
The IDF performed an extensive systematic review up to 2018, DR, VTDR, and CSME in 2030 and 2045, we incorporated the
including publications of multiple languages and data from IDF DM population projection data into our pooled prevalence
national health databases. The Analytical Hierarchy Process estimate of DR, VTDR, and CSME for each region. In brief, this
scoring system was used to evaluate the data, and population was modeled by performing the binomial-normal hierarchical
data from the United Nations was used. Details of the IDF model fitting with Markov chain Monte Carlo simulations (using a
methodology have been reported previously1 and are described sample size of 10 000 based on WinBUGS software version 14.3;
further in Appendix D (available at www.aaojournal.org). We Appendix E, available at www.aaojournal.org).
classified countries based on the Ninth IDF Atlas into 7 world In this projection estimate, region-specific DR, VTDR, and
regions: Africa, Europe, Middle East and North Africa (MENA), CSME prevalence rates within people with DM were assumed to
North America and the Caribbean (NAC), South and Central be constant over the next 25 years through year 2045. This was
America (SACA), South East Asia (SEA), and Western Pacific because, based on meta-regression analysis, we observed that the
(WP; Appendix D). The IDF regional classification was used in year of study was not associated significantly with DR prevalence
this study to allow for DR projection to be in concordance with for each region (Appendix F, available at www.aaojournal.org).
the IDF projection of DM. This allows for better regional and This observation indicated a constant trend of prevalence across
country-specific policy planning. In addition, the IDF regions all regions from 1980 through 2017 among the eligible studies
took into account the World Bank-defined income group charac- included in this review.
teristics to classify countries further based on epidemiologic and Meta-Regression Modelling. To evaluate factors associated
socioeconomic similarities,1 which are important factors for DM with the prevalence of DR and VTDR, the meta-regression model
burden and management. We did not use the 7 super regions as was used to model the logit of prevalence for DR and VTDR. We
used in the Global Burden of Disease Study because no data first performed a nonadjusted model followed by a multiple
were available on region or country-specific DM projections (in adjusted model, adjusting for region, habitation type, response rate,
terms of forecasted DM numbers, which are crucial for this study’s year of study, and DR diagnostic method. We did not concurrently
purpose). include ethnicity and world region in the same model as covariates
because of high collinearity between the two. We did not adjust for
Diabetic Retinopathy Definition and Assessment age and gender in the main multivariate model because only 27
studies reported complete age data and only 49 studies reported
We found that most studies defined DR according to either the complete gender data. Random effects were incorporated in the
Early Treatment Diabetic Retinopathy Study classification,27 models to account for between-study variability.
the American Association of Ophthalmology International
Clinical Diabetic Retinopathy Disease Severity Scale,28 or
the Scottish Diabetic Retinopathy Grading Scheme.29 In this Results
study, DR represents any DR, including VTDR and CSME.
Vision-threatening DR was defined as the presence of severe Figure S4 (available at www.aaojournal.org) shows the article
nonproliferative DR, proliferative DR, CSME, or a combina- selection process for studies included in the final meta-analysis.
tion thereof according to the Eye Diseases Prevalence In brief, a total of 3433 individual records were identified during
Research Group definition.30 Clinically significant macular the initial search, of which 81 relevant articles were selected. After
edema was defined as (1) thickening of the retina at or further reviewing the full text of these selected articles, 25 articles
within 500 mm of the center of the macula, (2) hard exudate were excluded. Meanwhile, unpublished DR data was obtained
at or within 500 mm of the center of the macula associated from a further 3 studies that were affiliated with the Asian Eye
with thickening of adjacent retina, or (3) a zone of retinal Epidemiology Consortium and met our selection criteria. Ulti-
thickening 1 disc area or larger, any part of which is within mately, 59 studies were included in the final analysis (Table S1,
1 disc diameter of the center of the macula according to the available at www.aaojournal.org).
Early Treatment Diabetic Retinopathy Study definition.27
Studies that provided only CSME data without a breakdown
of severe nonproliferative DR and proliferative DR data Summary of Included Studies
were not included in VTDR analysis and were included only The final included data (59 studies from 27 countries) consisted of
in the CSME analysis. 9685 patients with DR among 40 857 individuals with DM (age
range, 20e87 years). We further extracted VTDR data from 51 of
Statistical Analysis the 59 studies comprising 1789 patients with VTDR (of 36 091
individuals with DM) and CSME data from 41 studies comprising
All statistical analyses were performed using R statistical software 1145 patients with CSME (of 27 125 individuals with DM). By
version 3.4.4 (R Foundation for Statistical Computing). A P value IDF world regions, 17 study populations were from the WP, 14
of less than 0.05 was considered statistically significant. were from SEA, 12 were from NAC, 7 were from the MENA, 6

3
 Ophthalmology Volume -, Number -, Month 2021

were from Europe, 2 were from SACA, and 1 was from Africa. odds of having VTDR compared with those in SEA: Africa (OR
Detailed characteristics of the studies are described in Table S1. Of 4.32; 95% CI, 1.35e13.79), NAC (OR 2.94; 95% CI, 1.73e4.98),
the included studies, 56 reported complete data on region, and the MENA (OR 2.34; 95% CI, 1.36e4.01; Table 2). The
habitation type, response rate, year of study, and DR diagnosis VTDR prevalences for the remaining regions are as follows:
method and were used as adjusted covariates in the meta- SACA, 5.83% (95% CI, 4.15%e8.13%); the WP, 5.54% (95% CI,
regression modeling for DR. Of the 56 studies, a subset of 50 4.53%e6.76%); Europe, 5.49% (95% CI, 4.63%e6.51%); and
provided ethnicity information, 49 reported gender proportion data, SEA, 3.53% (95% CI, 2.45%e5.05%; Fig S7).
and 27 provided mean age data. Meanwhile, 49 studies were used For CSME (41 studies), the MENA showed the highest CSME
for VTDR-related meta-regression analysis and 39 studies were prevalence at 6.06% (95% CI, 3.59%e10.06%; Table 1; Fig S8,
used for CSME-related meta-regression analysis (results described available at www.aaojournal.org). Meta-regression analysis
below). showed that individuals with diabetes residing in the MENA were
significantly more likely to have CSME (OR, 2.48; 95% CI,
Global Prevalence and Numbers of Diabetic 1.33e4.61) compared with those residing in SEA (Table 2). North
Retinopathy, Vision-Threatening Diabetic America and the Caribbean had an estimated CSME prevalence of
Retinopathy, and Clinically Significant Macular 4.89% (95% CI, 2.92%e8.08%) and was also found to have
significantly higher odds of CSME (OR, 2.82; 95% CI,
Edema in 2020 1.48e5.39) compared with SEA (Table 2). The CSME
Figure S5 (available at www.aaojournal.org) shows the pooled prevalences for the remaining regions are as follows: Europe,
prevalence of DR, VTDR, and CSME globally and by region. 5.29% (95% CI, 4.18%e6.68%); SACA, 4.92% (95% CI,
The prevalence of DR was estimated to be 22.27% (95% 3.39%e7.08%); Africa, 4.10% (95% CI, 2.06%e7.99%); the
confidence interval [CI], 19.73%e25.03%) globally within the WP, 3.23%; 95% CI, 2.26%e4.59%); and SEA, 2.30% (95% CI,
DM population. The global number of adults with DR in 1.44%e3.67%; Fig S8).
2020 was estimated to be 103.12 million (95% CI,
91.34e115.90 million; Table 1; Fig 1). Meanwhile, the
prevalence of VTDR was estimated to be 6.17% (95% CI, Variations in Diabetic Retinopathy, Vision-
5.43%e6.98%) within the DM population, and the number of Threatening Diabetic Retinopathy, and Clinically
adults with VTDR was estimated to be 28.54 million (95% Significant Macular Edema Prevalences across
CI, 25.12e32.34 million) in 2020 globally (Table 1). The Ethnicities
global prevalence of CSME was estimated to be 4.07% (95%
CI, 3.42%e4.82%) within the DM population, with a global Figure S9 (available at www.aaojournal.org) illustrates the
CSME population of 18.83 million (95% CI, 15.82e22.32 variation in the prevalence of DR, VTDR, and CSME across
million; Table 1). ethnic groups. Hispanics showed the highest DR prevalence
at 47.40% (95% CI, 45.29%e49.52%) followed by Middle
Regional Variations in Diabetic Retinopathy, Easterners (32.90%; 95% CI, 26.06%e40.55%), people of
Vision-Threatening Diabetic Retinopathy, and African ancestry (31.01%; 95% CI, 26.10%e36.38%), people
Clinically Significant Macular Edema Prevalence of European ancestry (23.71%; 95% CI, 17.13%e31.84%),
and Asians at 17.94% (95% CI, 14.77%e21.61%; Fig S10,
Analysis of the 59 included studies showed that NAC (33.30%; available at www.aaojournal.org). For VTDR, people of
95% CI, 25.29%e42.40%) and MENA (32.90%; 95% CI, African ancestry showed the highest VTDR prevalence at
26.06e40.55%) regions showed significantly higher DR preva- 10.90% (95% CI, 7.87%e14.91), followed by Hispanics
lence than other regions (Table 1; Fig S6, available at (8.26%; 95% CI, 5.77- 11.71%), Middle Easterners (8.19%;
www.aaojournal.org). In the meta-regression analysis adjusting 95% CI, 5.11%e12.87%), people of European ancestry
for response rate, habitation type, year of study, and DR diagnostic (5.87%; 95% CI, 4.44%e7.72%), and Asians (4.06%; 95%
method, individuals with DM residing in NAC (odds ratio [OR], CI, 3.22%e5.11%). For CSME, Middle Easterners showed
2.33; 95% CI, 1.39e3.92) and the MENA (OR, 2.72; 95% CI, the highest CSME prevalence at 6.06% (95% CI,
1.58e4.68) showed significantly higher odds of DR compared with 3.59%e10.06%), followed by Hispanics (5.71%; 95% CI,
those residing in the SEA region (Table 2). 4.81%e6.78%), people of European ancestry (4.65%; 95%
Although pooled DR prevalence also was high in Africa at CI, 3.60%e5.99%), people of African ancestry (4.10%; 95%
35.90% (95% CI, 29.48%e42.87%), meta-regression analysis CI, 2.06%e7.99%), and Asians (2.67%; 95% CI,
showed only marginally significantly higher odds of DR in Africa 2.01%e3.54%).
compared with SEA (P ¼ 0.055; Table 2). Diabetic retinopathy In meta-regression analysis adjusting for response rate, habita-
prevalence for the remaining regions were as follows: the WP, tion type, year of study, and DR diagnostic method, compared with
19.20% (95% CI, 14.16%e25.50%); Europe, 18.75% (95% CI, Asians, Hispanics with diabetes were 2.92 times (OR, 2.92; 95%
13.69%e25.12%); SEA, 16.99% (95% CI, 14.13%e20.28%); CI, 1.22e6.98) more likely to have DR, and Middle Easterners
and SACA, 13.37% (95% CI, 6.13%e26.74%; Table 1; Fig 1; were 2.44 times (OR, 2.44; 95% CI, 1.51e3.94) more likely to
Fig S6). have DR (Table 2). Similarly, compared with Asians, Hispanics
For VTDR (51 studies), the top 3 regions were Africa (14.36%; (OR, 2.71; 95% CI, 1.30e5.67), people of African ancestry (OR,
95% CI, 10.10%e20.01%), the MENA (8.19%; 95% CI, 5.11%e 2.58; 95% CI, 1.24e5.38), and Middle Easterners (OR, 1.84;
12.87%), and NAC (7.82%; 95% CI, 5.34%e11.31%; Table 1; Fig 95% CI, 1.20e2.82) were more likely to have VTDR (Table 2).
S7, available at www.aaojournal.org). Meta-regression analysis We found that Hispanics (OR, 3.93; 95% CI, 1.74e8.88) and
adjusting for response rate, habitation type, year of study, and DR Middle Easterners (OR, 2.27; 95% CI, 1.43e3.60) also were
diagnostic method (Table 2) showed that individuals with DM observed to have higher odds of CSME compared with Asians
residing in these top 3 regions demonstrated significantly higher with diabetes (Table 2).

4
 Teo et al 
Global Prevalence and Projections of Diabetic Retinopathy

Table 1. Prevalence and Number of Adults with Diabetic Retinopathy, Vision-Threatening Diabetic Retinopathy, and Clinically Sig-
nificant Macular Edema in 2020

Diabetic Retinopathy Vision-Threatening Diabetic Retinopathy Clinically Significant Macular Edema

World Region Prevalence (%) No. (in Millions) Prevalence (%) No. (in Millions) Prevalence (%) No. (in Millions)
SEA 16.99 (14.13e20.28) 14.95 (12.42e17.81) 3.53 (2.45e5.05) 3.15 (2.15e4.44) 2.30 (1.44e3.67) 2.08 (1.26e3.22)
Africa 35.90 (29.48e42.87) 6.99 (5.73e8.33) 14.36 (10.10e20.01) 2.83 (1.97e3.90) 4.10 (2.06e7.99) 0.85 (0.40e1.56)
Europe 18.75 (13.69e25.12) 11.25 (8.12e14.93) 5.49 (4.63e6.51) 3.28 (2.74e3.88) 5.29 (4.18e6.68) 3.16 (2.47e3.98)
MENA 32.90 (26.06e40.55) 18.07 (14.28e22.28) 8.19 (5.11e12.87) 4.59 (2.80e7.09) 6.06 (3.59e10.06) 3.43 (1.96e5.54)
NAC 33.30 (25.29e42.40) 15.89 (12.03e20.16) 7.82 (5.34e11.31) 3.78 (2.54e5.37) 4.89 (2.92e8.08) 2.40 (1.38e3.83)
SACA 13.37 (6.13e26.74) 4.47 (1.93e8.51) 5.83 (4.15e8.13) 1.87 (1.31e2.58) 4.92 (3.39e7.08) 1.58 (1.07e2.25)
WP 19.20 (14.16e25.50) 31.50 (22.97e41.56) 5.54 (4.53e6.76) 9.06 (7.36e11.03) 3.23 (2.26e4.59) 5.34 (3.68e7.47)
Global 22.27 (19.73e25.03) 103.12 (91.34e115.90) 6.17 (5.43e6.98) 28.54 (25.12e32.34) 4.07 (3.42e4.82) 18.83 (15.82e22.32)

MENA ¼ Middle East and North Africa; NAC ¼ North America and Caribbean; SACA ¼ South and Central America; SEA ¼ South East Asia; WP ¼
Western Pacific.
Data are presented as percentage or number (95% confidence interval).

Figure 1. Global map showing diabetic retinopathy (DR) prevalence and numbers by International Diabetes Foundation world regions in 2020. AFR ¼
Africa; EUR ¼ Europe; MENA ¼ Middle East and North Africa; NAC ¼ North America and Caribbean; SACA ¼ South and Central America; SEA ¼
South East Asia; WP ¼ Western Pacific.

5
 Ophthalmology Volume -, Number -, Month 2021

Table 2. Factors Associated with Diabetic Retinopathy, Vision-Threatening Diabetic Retinopathy, and Clinically Significant Macular
Edema (Multivariate Analysis)

Diabetic Retinopathy* Vision-Threatening Diabetic Retinopathyy Clinically Significant Macular Edemaz

Adjusted Odds Adjusted Odds Adjusted Odds
Ratiox (95% CI) P Value Ratiox (95% CI) P Value Ratiox (95% CI) P Value
Region
SEA Reference
Africa 3.32 (0.97e11.33) 0.055 4.32 (1.35e13.79) 0.013 1.23 (0.30e5.14) 0.773
Europe 1.45 (0.76e2.75) 0.258 1.63 (0.76e3.48) 0.209 1.59 (0.60e4.18) 0.349
MENA 2.72 (1.58e4.68) < 0.001 2.34 (1.36e4.01) 0.002 2.48 (1.33e4.61) 0.004
NAC 2.33 (1.39e3.92) 0.001 2.94 (1.73e4.98) < 0.001 2.82 (1.48e5.39) 0.002
SACA 0.93 (0.38e2.30) 0.880 1.43 (0.44e4.67) 0.557 1.93 (0.52e7.13) 0.326
WP 1.40 (0.84e2.31) 0.196 1.77 (1.03e3.05) 0.039 1.22 (0.60e2.50) 0.586
Ethnicity||
Asian Reference
African ancestry 1.94 (0.83e4.57) 0.127 2.58 (1.24e5.38) 0.011 1.20 (0.37e3.92) 0.762
European ancestry 1.35 (0.89e2.04) 0.162 1.50 (0.99e2.27) 0.054 1.72 (0.98e3.01) 0.058
Hispanic 2.92 (1.22e6.98) 0.016 2.71 (1.30e5.67) 0.008 3.93 (1.74e8.88) 0.001
Middle Eastern 2.44 (1.51e3.94) < 0.001 1.84 (1.20e2.82) 0.005 2.27 (1.43e3.60) < 0.001
Habitation type
Rural Reference
Urban 1.05 (0.68e1.62) 0.837 0.95 (0.62e1.47) 0.822 1.35 (0.78e2.34) 0.291
Mixed 0.85 (0.54e1.34) 0.495 0.85 (0.55e1.32) 0.463 1.05 (0.57e1.92) 0.878
Response rate 1.00 (0.98e1.02) 0.818 1.01 (0.99e1.03) 0.569 1.01 (0.99e1.04) 0.271
Year of study conducted 0.99 (0.97e1.01) 0.398 1.01 (0.99e1.03) 0.413 1.01 (0.98e1.04) 0.386
Diagnostic method
Clinical fundus Reference
examination only
CFP (1e2 fields) 0.98 (0.63e1.52) 0.924 0.90 (0.56e1.43) 0.644 1.18 (0.63e2.21) 0.597
CFP (3e7 fields) 1.38 (0.88e2.17) 0.158 0.68 (0.43e1.06) 0.088 0.55 (0.31e0.98) 0.044
Age, per decade{ 2.41 (1.20e4.82) 0.013 0.83 (0.32e2.16) 0.704 1.71 (0.42e7.04) 0.455
Gender#
Male Reference
Female 0.10 (0.00e2.41) 0.158 0.05 (0.00e0.99) 0.049 0.17 (0.00e5.83) 0.324
Mydriatic CFP**
No Reference
Yes 1.10 (0.65e1.85) 0.721 1.21 (0.68e2.13) 0.514 1.57 (0.66e3.74) 0.307

CFP ¼ color fundus photography; CI ¼ confidence interval; CSME ¼ clinically significant macular edema; DR ¼ diabetic retinopathy; MENA ¼ Middle
East and North Africa; NAC ¼ North America and Caribbean; SACA ¼ South and Central America; SEA ¼ South East Asia; VTDR ¼ vision-
threatening diabetic retinopathy; WP ¼ Western Pacific.
Boldface indicates statistical significance of P < 0.05.
*Analysis performed on only 56 studies with DR and adjustment covariates data available.
y
Analysis performed on only 49 studies with VTDR data available.
z
Analysis performed on only 39 studies with CSME data available.
x
Meta-regression models were adjusted by region (but excluded from model when evaluating ethnicity as exposure), habitation type, response rate, year study
conducted, and DR diagnosis method.
||
Analysis performed on 50 studies, 44 studies, and 34 studies with ethnicity information available for DR, VTDR, and CSME analyses, respectively.
{
Analysis performed on 27 studies, 23 studies, and 21 studies with mean age information available for DR, VTDR, and CSME analyses, respectively.
#
Analysis performed on 49 studies, 43 studies, and 39 studies with gender proportion information available for DR, VTDR, and CSME analyses, respectively.
**Analysis performed on 39 studies, 35 studies, and 26 studies that used CFP for DR diagnosis for DR, VTDR, and CSME analyses, respectively.

Effect of Diabetic Retinopathy Diagnostic subset of 39 studies that used color fundus photography for
Method and Dilated Fundus Photography on diagnosis, multivariate logistic regression showed that dilated
fundus photography had no significant effect on the odds of DR,
Diabetic Retinopathy, Vision-Threatening VTDR, or CSME (Table 2).
Diabetic Retinopathy, and Clinically Significant
Macular Edema
Effect of Age and Gender on Diabetic
We further evaluated the effects of diagnostic method and dilated Retinopathy, Vision-Threatening Diabetic
fundus photography on the estimates of DR, VTDR, and CSME Retinopathy, and Clinically Significant Macular
prevalence. Multivariate logistic regression showed that the use of
Edema
different diagnostic methods generally did not have significant
effects on the odds of DR, VTDR, or CSME (except for the use of Multivariate meta-regression subgroup analysis revealed that the
3e7 fields of color fundus photographs on CSME, which was of OR of DR was 2.41 (95% CI, 1.20e4.82; P ¼ 0.013) with each
borderline significance [P ¼ 0.044]; Table 2). In addition, in a decade increase in age, after adjusting for world region, habitation

6
 Teo et al 
Global Prevalence and Projections of Diabetic Retinopathy

type, response rate, year of study, and DR diagnostic method

(23.85e34.29)
Table 3. Projection of the Number of People with Diabetic Retinopathy, Vision-Threatening Diabetic Retinopathy, and Clinically Significant Macular Edema in 2030 and 2045

(3.84e10.85)

CI ¼ confidence interval; MENA ¼ Middle East and North Africa; NAC ¼ North America and Caribbean; SACA ¼ South and Central America; SEA ¼ South East Asia; WP ¼ Western Pacific.
(2.22e5.64)
(0.98e3.77)
(2.81e4.56)

(1.83e5.12)
(1.67e3.46)
(4.78e9.84)
(Table 2). However, the effect of age on VTDR and CSME
Clinically Significant Macular Edema, prevalence was not statistically significant (P  0.704; Table 2).

2045
Subgroup multivariate meta-regression analysis showed that
gender had no significant effect on DR or CSME prevalence (P 
No. in Millions (95% CI)

0.158), whereas the effect of gender on VTDR was marginally

3.62
2.04
3.63
6.72
3.18
2.46
6.96
28.61
significant (P ¼ 0.049; Table 2). Habitation type, response rate,
and year of study did not have significant effects on DR, VTDR,
or CSME prevalence (P  0.271; for univariate analyses,
Table 2 and Table S2 [available at www.aaojournal.org]).

(19.69e27.97)
(1.64e4.23)
(0.58e2.30)
(2.74e4.40)
(2.70e7.62)
(1.63e4.55)
(1.36e2.87)
(4.43e9.03) Quality Assessment: Sensitivity Analysis
2030

Excluding Studies with Higher Risk of Bias

2.73
1.24
3.51
4.74
2.84
2.01
6.43
23.50

Quality assessment using the risk-of-bias tool showed that 3 studies

(the Hoorn Study,32 the Tehran Eye Study,33 and the Taipei Ta An
Study34), had a score of 4 or more and were deemed to be of higher
risk of bias (Table S3, available at www.aaojournal.org).
Sensitivity analysis excluding these 3 studies showed that
(39.20e51.33)

findings remained largely similar (Table S4A, B, available at

(5.48e13.89)

(9.58e14.43)
(3.76e7.77)
(4.78e9.44)
(3.13e4.45)

(3.36e7.16)
(2.04e3.97)

www.aaojournal.org).
2045
No. in Millions (95% CI)

Projection of Number of Individuals with

Diabetic Retinopathy,
Vision-Threatening

5.48
6.84
3.76
9.01
5.02
2.90
11.81
44.82

Diabetic Retinopathy, Vision-Threatening

Diabetic Retinopathy, and Clinically Significant
Macular Edema in 2030 and 2040
In 2020, the total number of adults with DR was estimated to be
(31.63e41.15)
(8.89e13.32)

103.12 million globally (95% CI, 91.34e115.90 million), with the

(2.79e5.82)
(2.86e5.75)
(3.04e4.30)
(3.86e9.77)
(2.98e6.36)
(1.66e3.30)

highest numbers in the WP region (31.50 million), followed by the

2030

MENA (18.07 million) and NAC (15.89 million; Table 1; Fig 1).
For VTDR, the total number was estimated to be 28.54 million
(95% CI, 25.12e32.34 million) in 2020, and similarly with the
4.13
4.16
3.64
6.36
4.46
2.37
10.93
36.05

largest numbers in the WP (9.06 million), followed by the

MENA (4.59 million) and NAC (3.78 million; Table 1). For
CSME, the total number was estimated to be 18.83 million (95%
CI, 15.82e22.32 million) in 2020, with corresponding highest
numbers in the WP (5.34 million) followed by the MENA (3.43
(143.70e178.60)
(21.64e31.08)
(13.91e20.19)

(27.98e43.66)
(15.95e26.83)

(30.03e54.40)

million) and Europe (3.16 million; Table 1). In further meta-

(9.23e17.17)

(3.00e12.99)

regression analysis by region subgroups, we found no statisti-

cally significant trend in DR prevalence over time (based on study
2045

year) across all regions (P  0.084; results not shown in tables; Fig
S11, available at www.aaojournal.org).
No. in Millions (95% CI)
Diabetic Retinopathy,

6.96
26.06
16.93
12.89
35.47
21.11

41.08
160.50

Hence, while assuming the rate of DR prevalence to be constant

over the next 25 years, we estimated that the global number of
adults with DR would increase by 25.9% to 129.84 million (95%
CI, 115.30e145.60 million) in 2030 and by 55.6% to 160.50
million (95% CI, 143.70e178.60 million) in 2045 (Table 3). Much
(115.30e145.60)

of the projected increase in DR numbers is attributable to

(16.18e23.37)

(19.75e30.79)
(14.15e23.83)

(27.76e50.13)

significant increases in the MENA, SEA, and WP regions. The

(8.36e12.28)
(9.03e16.50)

(2.44e10.82)

WP will remain as the region with the greatest number of people

with DR in 2045, with an increment of 9.58 million from 2020.
2030

Across world regions, the MENA is projected to have the most

drastic increase in DR cases by 96.3% (17.4 million) from 2020
5.69
19.62
10.29
12.46
25.05
18.75

37.98
129.84

to 2045. However, it is projected that the NAC, Europe, and

SACA regions will show small increments in DR cases from
2020 to 2045.
Projection of the number of people with VTDR and CSME also
was performed while assuming the rate of prevalence to be con-
World Region

stant over the next 25 years. We estimated that the global number
of adults with VTDR will increase by 26.3% to 36.05 million (95%
CI, 31.63e41.15 million) in 2030 and by 57.0% to 44.82 million
MENA
Europe

SACA

Global
Africa

NAC

(95% CI, 39.20e51.33 million) in 2045. The global number of

SEA

WP

adults with CSME will increase by 24.8% to 23.50 million (95%

7
 Ophthalmology Volume -, Number -, Month 2021

CI, 19.69e27.97 million) in 2030 and by 51.9% to 28.61 million base from Asia, including many new studies from the WP
(95% CI, 23.85e34.29 million) in 2045. and SEA, which have lower DR prevalence than other
regions, as shown in our results. Second, also changes
Discussion have been made in the definition of DM over time. For
example, DM now includes the use of hemoglobin A1c of
Our study provides comprehensive and up-to-date evalua- more than 6.5% (48 mmol/mol)36,37 as a diagnostic
tions of the current global DR prevalence with the largest criterion. The updated DM diagnostic criteria and
meta-analysis to date. Our study provides novel estimates on improvement in standards of care in DM38 allow for
global and regional CSME prevalence and future projection earlier DM diagnosis, stricter glycemic control, and
of the number of people with DR, VTDR, and CSME consequently better prevention of complications that have
globally and regionally. From a global prevalence of resulted in lower rates of other diabetes-related complica-
22.27% for DR, 6.17% for VTDR, and 4.07% for CSME, tions, including microangiopathy and nephropathy.39 This
we estimated that there will be 103.12 million people with could explain the lower DR prevalence found in our study
DR, 28.54 million people with VTDR, and 18.83 million compared with previous estimates by Yau et al.8 Third,
people with CSME in 2020. The number of people with DR, significant public interest exists regarding DM in Asia,12
VTDR, and CSME is projected to rise to 160.50 million, which has led to national policies for primary prevention
44.82 million, and 28.61 million, respectively, in 2045, of DM and screening for high-risk populations in many
disproportionately affecting individuals with DM residing in Asian countries (e.g., Singapore,40 India,41 and China42),
the MENA and WP regions. The demand for DR and CSME potentially leading to earlier diagnosis of DM and a
treatment will continue to rise significantly in the future. corresponding lower prevalence of DM-related complica-
tions. Finally, some studies with low response rates included
Key Strengths and Findings in the previous analysis8 were excluded in the current
review; these studies have a relatively high DR prevalence
A key strength of this current systematic review lies in the perhaps because of a selection artifact.43-45 Hence, the
significantly more comprehensive and up-to-date estimates current study estimates are likely to represent more accurate
as compared with the last review conducted a decade ago.8 and up-to-date prevalence estimates.
This was coupled with critical appraisal of study quality Our study further provides regional and ethnic variations
including only population-based studies with response in DR, VTDR, and CSME prevalence estimates that are not
rates of 60% or more and strict application of inclusion and currently available. We observed significant regional varia-
exclusion criteria. Substantial improvement occurred in the tion, with people with DM living in NAC and the MENA
number of included studies from 35 to 59, with a substantial having higher odds of DR and CSME and those in Africa, the
increment of data from Asian populations, specifically the MENA, NAC, and the WP having higher odds of VTDR.
WP (from 9 to 17 studies) and SEA (from 2 to 14 studies). Similar regional variation was seen in DM prevalence where
With the inclusion of 40 new studies and better Asian rep- prevalence is estimated to be the highest in the MENA
resentation compared with the previous review, our findings (13.9%) followed by NAC (13.0%) in 2045.1 These 2 regions
provide more up-to-date estimates. Importantly, we included correspondingly showed higher DR prevalence in the current
a notably higher number of studies from China and India, findings (Fig S12, available at www.aaojournal.org). These
which have the highest numbers of people with DM (China, regional estimates can aid further in region-specific health
116 million; India, 77 million).1 In this review, we included care policy planning. Similarly, we reported a significantly
7 studies from China and 11 from India, a substantial higher DR, VTDR, and CSME prevalence among Hispanics
improvement compared with the previous review, which and Middle Easterners compared with Asians. The previous
consisted of 2 studies from China and 3 from India.8 review by Yau et al8 similarly reported a lower DR
Further novelties of our study are the inclusion of studies prevalence among Asians. In addition, both our review and
from all regions including the MENA, SACA, and Africa, the study of Yau et al8 found that people of African
areas from which previously no studies were included, ancestry have the highest prevalence of VTDR. Similarly,
and the inclusion of a recent study from Russia (which is the National Health and Nutrition Examination Survey
the first Russian population-based study reporting on DR found higher VTDR prevalence among Hispanic and Black
prevalence).35 We classified studies according to IDF individuals. The top 3 ethnic groups (Hispanics, Middle
regions, with most regions well represented by a sufficient Easterners, and people of African ancestry) correspondingly
number of studies with large sample sizes. The most up- were from the 3 regions with the highest DR prevalence
to-date DM data from the IDF Atlas 2019 also was used (Africa, NAC, and the MENA). It is interesting that these
to provide robust DR population estimates. ethnic groups showed significantly higher odds of DR and
Our study estimated the pooled global prevalence of DR VTDR, despite being from different countries and regions
and VTDR to be 22.27% and 6.17%, respectively, lower (e.g., Middle Easterners were from Egypt, Iran, Jordan, and
than the previous estimates of 34.6% and 10.2% by Yau Saudi Arabia; people of African ancestry were from Africa
et al.8 Differences in estimates may be the result of a and NAC). This potentially suggests that the effect of
combination of factors. First, as discussed earlier, our ethnicity as a risk factor for DR and VTDR may transcend
analysis consisted of a more extensive and recent evidence geographical regions.

8
 Teo et al 
Global Prevalence and Projections of Diabetic Retinopathy

Projection of Global Number of People with covariate in our analysis because of limited age-related data
Diabetic Retinopathy and Clinically Significant provided by studies. Third, most studies defined CSME by
Macular Edema stereoscopic fundus photography, before OCT (a sensitive
and more accurate technique to detect CSME47) was used
We have attempted to project the number of people with DR widely; thus, this may underestimate our CSME estimates.
over time globally and by regions. Across all regions, IDF In addition, we were unable to provide diabetic macular
estimated that both DM prevalence and numbers will rise edema estimates that include nonclinically significant
steadily from 2019 to 20451 (Appendix G, available at macular edema because of limited data from studies.
www.aaojournal.org). Substantial regional variation was Fourth, the change in prevalence over time is difficult to
found, with the MENA projected to have the greatest quantify, especially given the nature of a disease in which
increase in DM prevalence (by 2.9%) and SEA projected environmental and behavioral factors play a significant
to have the greatest increase in absolute DM population role. Nonetheless, our meta-regression analysis by region
size (by 65.2 million) by 2045.26 However, the WP will showed that the year of study had no significant effect on
remain as the region with the largest DM population DR prevalence (Fig S11; P  0.084 for all regions); thus,
(212.2 million) by 2045.26 constant prevalence rates were used for projection of
By 2045, the MENA is expected to have the greatest numbers. Fifth, the duration and systemic control of DM
increase in DR population by 96.3% (17.4 million). This is are important risk factors for DR. However, among the
because its DM population is estimated to increase drasti- included studies, few provided data in this regard, and
cally from 54.8 million in 2019 to 107.6 million in 2045,1 when reported, these may not be completely reliable
with DM prevalence in the MENA rising the most across because methods to obtain these measurements vary
all regions, from 12.8% to 15.7% in 2045.26 greatly. Despite insufficient data to analyze this in detail,
Meanwhile, although the estimated DR prevalence in the we cannot entirely rule out the impact of duration and
WP (19.20%) is lower than in some regions, the WP control of DM on the current estimates. Longer duration
currently has the largest DR population in absolute numbers of DM likely is associated with higher prevalence of DR,
and is projected to continue to do so with a DR population which may explain in part the higher prevalence in NAC
size of 41.08 million in 2045, an increase of 9.58 million where DM has been a consistent top chronic disease and
patients with DR from 2020. This is because of the sheer individuals with DM are living longer.48 However, the
number of people with DM residing in the WP, which is higher prevalence of DR in Africa and the MENA may be
expected to grow to 212.2 million in 2045,1,26 leading to the attributed to poor control of DM, because as previous
largest absolute number of patients with DR in a world reports indicate high proportions of untreated DM in
region, currently and in the future. Africa (69.2%)1 and high rates of poorly controlled DM in
We estimated that the global DR population will increase the MENA countries (approximately 50%).49,50 Hence,
by 55.6% (57.4 million) from 2020 to 2045. This is mainly future analysis that can incorporate information on duration
attributed to the rapidly growing global DM population and control status of DM further would help to improve the
especially in Africa, the MENA, and the WP.1 Our findings accuracy of these estimates.
show that DR population size is tightly correlated with DM Sixth, 34 of 59 studies provided data on the proportion of
population size and suggest the need for more resources for ungradable fundus, which ranged between 0.4% and 22%.
DM and DR management, particularly in these regions. Diabetes mellitus, especially when poorly controlled, in-
With the rising DM population, attention should be paid creases the risk of cataract formation,51 which may lead to
to prevent complications such as DR. significant cataract that may obscure the fundal view. This
Diabetic macular edema is now known to be the main could represent an underestimation of DR prevalence in
cause of moderate vision loss among individuals with DM regions with poor access to cataract surgery. However, only
globally46 but estimates of CSME prevalence have not been 3 studies had more than 10% of ungradable fundus, and
available previously. Our study provides novel estimates of thus this is unlikely to affect our estimates significantly.
CSME prevalence globally and projections of the number of Finally, we acknowledge that a significant difference exists
adults with CSME. We estimated that the global number of between the DR prevalence among individuals with type 1
adults with CSME will rise by 51.9% to 28.61 million in DM (T1DM) and type 2 DM (T2DM), but most included
2045. This suggests the need to improve access to CSME studies did not provide data on DR prevalence by diabetes
treatment such as intravitreal antievascular endothelial types. This is because accurate differentiation between the 2
growth factor therapy or laser treatment. types requires sophisticated laboratory tests, which
generally is not feasible in large-scale population studies.
Study Limitations To our knowledge, for the same reason, separate global es-
timates of diabetes prevalence for T1DM and T2DM, in
Our review has some limitations. First, in Africa and SACA, particular in adults, do not exist.47,52 Nevertheless, it is
the limited studies may be insufficient to represent the re- important to note that although complete data on DR
gion entirely. In addition, the limited studies in Africa likely prevalence by diabetes types are not available, this would
resulted in insufficient statistical power in analysis, leading not have affected the overall DR prevalence estimates
to statistically insignificant higher odds of DR, despite Af- substantially in the study population. Sensitivity analysis
rica having the highest DR prevalence estimate. Second, in excluding studies that did not provide information on the
the projection of DR, we were unable to include age as a proportion of T1DM and T2DM showed similar prevalence

9
 Ophthalmology Volume -, Number -, Month 2021

results (Table S5, available at www.aaojournal.org). To In conclusion, our study provides more precise and
evaluate formally how the prevalence of DR changes with contemporary estimates of the global prevalence of DR,
varying proportions of T1DM and T2DM, we performed a VTDR, and CSME, with projections of the present and
simulation analysis (Appendix H, available at future burden up to 2045. Our findings suggest that
www.aaojournal.org) that shows that the estimate on global approximately 1 in 5 persons with diabetes worldwide
prevalence of DR would still fall within the 95% CI of our have DR. Although the current prevalence estimates for
original estimate when 93% or more of DM cases are VTDR are lower than earlier estimates, the total number of
T2DM in populations. Because T2DM accounts for more people losing vision as a result of DR may continue to
than 90% to 95% for all diabetes cases, in the population rise. Our findings also suggest the continual need for high-
older than 20 years (age range of this study), the proportion quality population-based studies of DR, especially in Af-
of T2DM cases likely would be even higher than 90% to rica and SACA. Findings and estimates from this study
95% (thus fulfilling the cutoff of 93%). Therefore, this may aid in the planning of global, regional, and country-
limitation would not have affected our DR prevalence specific health care strategies to prevent diabetes-related
estimates substantially in people older than 20 years. vision loss.

Footnotes and Disclosures

Originally received: January 10, 2021. HUMAN SUBJECTS: No human subjects were included in this study. The
Final revision: April 20, 2021. IRB/ethics committee at each institution approved this study. All research
Accepted: April 23, 2021. adhered to the tenets of the Declaration of Helsinki. The requirement for
Available online: ---. Manuscript no. D-21-00059. informed consent was waived because of the retrospective nature of the
1
Singapore National Eye Centre, Singapore Eye Research Institute, study.
Singapore, Republic of Singapore. No animal subjects were included in this study.
2
Duke-NUS Medical School, Singapore, Republic of Singapore. Author Contributions:
3
Ufa Eye Research Institute, Ufa, Bashkortostan, Russia. Conception and design: T.Y.Wong, Cheng
4
Beijing Institute of Ophthalmology, Beijing Key Laboratory of Ophthal- Analysis and interpretation: Sabanayagam, T.Y.Wong, Cheng
mology and Visual Sciences, Beijing Tongren Eye Center, Beijing Tongren Data collection: Teo, Tham, Yu, Chee, Rim, Cheung, Bikbov, Wang, Tang,
Hospital, Capital Medical University, Beijing, China. Lu, I.Y.H.Wong, Ting, Tan, Jonas, Sabanayagam, T.Y.Wong, Cheng
5
Department of Ophthalmology, Eye and ENT Hospital of Fudan Uni- Obtained funding: N/A
versity, Shanghai, China.
Overall responsibility: Teo, Tham, Yu, Chee, Rim, Cheung, Bikbov, Wang,
6
Department of Ophthalmology, University of Hong Kong Li Ka Shing Tang, Lu, I.Y.H.Wong, Ting, Tan, Jonas, Sabanayagam, T.Y.Wong, Cheng
Faculty of Medicine, Hong Kong, China.
7 Abbreviations and Acronyms:
Department of Ophthalmology, Medical Faculty Mannheim, Heidelberg CI ¼ confidence interval; CSME ¼ clinically significant macular edema;
University, Mannheim, Germany. DM ¼ diabetes mellitus; DR ¼ diabetic retinopathy;
8
Yong Loo Lin School of Medicine, National University of Singapore, IDF ¼ International Diabetes Federation; MENA ¼ Middle East and
Singapore, Republic of Singapore. North Africa; NAC ¼ North America and the Caribbean; OR ¼ odds
9
Department of Ophthalmology, Hong Kong Sanatorium and Hospital, ratio; SACA ¼ South and Central America; SEA ¼ South East Asia;
Hong Kong SAR, China. T1DM ¼ type 1 diabetes mellitus; T2DM ¼ type 2 diabetes mellitus;
*These authors contributed equally as first authors. VTDR ¼ vision-threatening diabetic retinopathy; WP ¼ Western
z
Both authors contributed equally as senior authors. Pacific.
Disclosure(s): Keywords:
All authors have completed and submitted the ICMJE disclosures form. Diabetes mellitus, Diabetic retinopathy, Population, Prevalence, Systematic
Daniel Shu Wei Ting, an editorial board member of this journal, was review.
recused from the peer-review process of this article and had no access to Correspondence:
information regarding its peer-review. Ching-Yu Cheng, MD, PhD, Singapore Eye Research Institute, The
The author(s) have no proprietary or commercial interest in any materials Academia, 20 College Road, Discovery Tower Level 6, Singapore, 169856,
discussed in this article. Republic of Singapore. E-mail: [email protected].
Supported by the National Medical Research Council, Singapore, Republic
of Singapore (grant nos.: NMRC/CIRG/1417/2015 and NMRC/CIRG/
1488/2018).

References

1. Saeedi P, Petersohn I, Salpea P, et al. Global and regional 2. Khalil H. Diabetes microvascular complications: a clinical
diabetes prevalence estimates for 2019 and projections for update. Diabetes Metab Syndr. 2017;11(Suppl 1):
2030 and 2045: results from the International Diabetes S133eS139.
Federation Diabetes Atlas, 9(th) edition. Diabetes Res Clin 3. Cheung N, Mitchell P, Wong TY. Diabetic retinopathy. Lan-
Pract. 2019;157:107843. cet. 2010;376(9735):124e136.

10
 Teo et al 
Global Prevalence and Projections of Diabetic Retinopathy

4. Fong DS, Aiello L, Gardner TW, et al. Retinopathy in dia- 22. Tang Y, Wang X, Wang J, et al. Prevalence and causes of
betes. Diabetes Care. 2004;27(Suppl 1):S84eS87. visual impairment in a Chinese adult population: the Taizhou
5. Sivaprasad S, Gupta B, Crosby-Nwaobi R, Evans J. Prevalence Eye Study. Ophthalmology. 2015;122(7):1480e1488.
of diabetic retinopathy in various ethnic groups: a worldwide 23. You QS, Choy BKN, Chan JCH, et al. Prevalence and causes
perspective. Surv Ophthalmol. 2012;57(4):347e370. of visual impairment and blindness among adult Chinese in
6. Leasher JL, Bourne RR, Flaxman SR, et al. Global estimates Hong Kongdthe Hong Kong Eye Study. Ophthalmic Epi-
on the number of people blind or visually impaired by diabetic demiol. 2020;27(5):354e363.
retinopathy: a meta-analysis from 1990 to 2010. Diabetes 24. Bikbov M, Fayzrakhmanov RR, Kazakbaeva G, Jonas JB.
Care. 2016;39(9):1643e1649. Ural Eye and Medical Study: description of study design and
7. Global Burden of Disease 2019 Blindness and Vision methodology. Ophthalmic Epidemiol. 2018;25(3):187e198.
Impairment Collaborators, Vision Loss Expert Group of 25. Hoy D, Brooks P, Woolf A, et al. Assessing risk of bias in
the Global Burden of Disease Study. Causes of blindness prevalence studies: modification of an existing tool and evidence
and vision impairment in 2020 and trends over 30 years, of interrater agreement. J Clin Epidemiol. 2012;65(9):934e939.
and prevalence of avoidable blindness in relation to 26. International Diabetes Federation. International Diabetes
VISION 2020: the Right to Sight: an analysis for the Federation Diabetes Atlas. 9th Edition. Brussels, Belgium:
Global Burden of Disease Study. Lancet Glob Health. International Diabetes Federation; 2019.
2021;9(2):e144ee160. 27. Early Treatment Diabetic Retinopathy Study Research Group.
8. Yau JW, Rogers SL, Kawasaki R, et al. Global prevalence and Grading diabetic retinopathy from stereoscopic color fundus
major risk factors of diabetic retinopathy. Diabetes Care. photographsdan extension of the modified Airlie House
2012;35(3):556e564. classification. ETDRS report number 10. Ophthalmology.
9. Claessen H, Kvitkina T, Narres M, et al. Markedly decreasing 1991;98(5 Suppl):786e806.
incidence of blindness in people with and without diabetes in 28. Wilkinson CP, Ferris 3rd FL, Klein RE, et al. Proposed interna-
southern Germany. Diabetes Care. 2018;41(3):478e484. tional clinical diabetic retinopathy and diabetic macular edema
10. Zhang X, Saaddine JB, Chou CF, et al. Prevalence of diabetic disease severity scales. Ophthalmology. 2003;110(9):1677e1682.
retinopathy in the United States, 2005e2008. JAMA. 29. Zachariah S, Wykes W, Yorston D. Grading diabetic reti-
2010;304(6):649e656. nopathy (DR) using the Scottish grading protocol. Community
11. Cheung N, Wong TY. Changing trends of blindness: the initial Eye Health. 2015;28(92):72e73.
harvest from translational public health and clinical research in 30. Kempen JH, O’Colmain BJ, Leske MC, et al. The prevalence
ophthalmology. Am J Ophthalmol. 2012;153(2):193e195. of diabetic retinopathy among adults in the United States. Arch
12. Ramachandran A, Ma RC, Snehalatha C. Diabetes in Asia. Ophthalmol. 2004;122(4):552e563.
Lancet. 2010;375(9712):408e418. 31. Stijnen T, Hamza TH, Ozdemir P. Random effects meta-
13. India State-Level Disease Burden Initiative Diabetes Col- analysis of event outcome in the framework of the general-
laborators. The increasing burden of diabetes and variations ized linear mixed model with applications in sparse data. Stat
among the states of India: the Global Burden of Disease Med. 2010;29(29):3046e3067.
Study 1990e2016. Lancet Glob Health. 2018;6(12): 32. van Leiden HA, Dekker JM, Moll AC, et al. Risk factors for
E1352eE1362. incident retinopathy in a diabetic and nondiabetic popula-
14. Tan GS, Cheung N, Simo R, et al. Diabetic macular oedema. tion: the Hoorn study. Arch Ophthalmol. 2003;121(2):
Lancet Diabetes Endocrinol. 2017;5(2):143e155. 245e251.
15. Wong TY, Sun J, Kawasaki R, et al. Guidelines on diabetic 33. Hatef E, Fotouhi A, Hashemi H, et al. Prevalence of retinal
eye care: the International Council of Ophthalmology recom- diseases and their pattern in Tehran: the Tehran eye study.
mendations for screening, follow-up, referral, and treatment Retina. 2008;28(5):755e762.
based on resource settings. Ophthalmology. 2018;125(10): 34. Chen MS, Kao CS, Chang CJ, et al. Prevalence and risk
1608e1622. factors of diabetic retinopathy among noninsulin-depen-
16. Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items dent diabetic subjects. Am J Ophthalmol. 1992;114(6):
for systematic reviews and meta-analyses: the PRISMA 723e730.
statement. BMJ. 2009;339:b2535. 35. Bikbov MM, Gilmanshin TR, Kazakbaeva GM, et al. Preva-
17. Yang QH, Zhang Y, Zhang XM, Li XR. Prevalence of diabetic lence of myopic maculopathy among adults in a Russian
retinopathy, proliferative diabetic retinopathy and non- population. JAMA Netw Open. 2020;3(3):e200567.
proliferative diabetic retinopathy in Asian T2DM patients: a 36. Tsugawa Y, Takahashi O, Meigs JB, et al. New diabetes
systematic review and meta-analysis. Int J Ophthalmol. diagnostic threshold of hemoglobin A(1c) and the 3-year
2019;12(2):302e311. incidence of retinopathy. Diabetes. 2012;61(12):
18. Ding J, Wong TY. Current epidemiology of diabetic retinop- 3280e3284.
athy and diabetic macular edema. Curr Diab Rep. 2012;12(4): 37. Sabanayagam C, Khoo EY, Lye WK, et al. Diagnosis of
346e354. diabetes mellitus using HbA1c in Asians: relationship between
19. Liu L, Wu X, Liu L, et al. Prevalence of diabetic retinopathy in HbA1c and retinopathy in a multiethnic Asian population.
mainland China: a meta-analysis. PLoS One. 2012;7(9): J Clin Endocrinol Metab. 2015;100(2):689e696.
e45264. 38. American Diabetes Association. Executive summary: stan-
20. Hyungtaek Rim T, Ryo K, Tham YC, et al. Prevalence and dards of medical care in diabetesd2014. Diabetes Care.
Pattern of Geographic Atrophy in Asia: the Asian Eye 2014;37(Suppl 1):S5eS13.
Epidemiology Consortium. Ophthalmology. 2020;127(10): 39. Gregg EW, Li Y, Wang J, et al. Changes in diabetes-related
1371e1381. complications in the United States, 1990e2010. N Engl J
21. Tham YC, Tao Y, Zhang L, et al. Is kidney function associated Med. 2014;370(16):1514e1523.
with primary open-angle glaucoma? Findings from the Asian 40. Nanditha A, Ma RC, Ramachandran A, et al. Diabetes in Asia
Eye Epidemiology Consortium. Br J Ophthalmol. and the Pacific: implications for the global epidemic. Diabetes
2020;104(9):1298e1303. Care. 2016;39(3):472e485.

11
 Ophthalmology Volume -, Number -, Month 2021

41. Verma R, Khanna P, Mehta B. National programme on pre- 47. Chhablani J, Mathai A, Rani P, et al. Comparison of con-
vention and control of diabetes in India: need to focus. Aus- ventional pattern and novel navigated panretinal photocoagu-
tralas Med J. 2012;5(6):310e315. lation in proliferative diabetic retinopathy. Invest Ophthalmol
42. Yin J, Kong AP, Chan JC. Prevention and care programs Vis Sci. 2014;55(6):3432e3438.
addressing the growing prevalence of diabetes in China. Curr 48. Zhang N, Yang X, Zhu X, et al. Type 2 diabetes mellitus
Diab Rep. 2016;16(12):130. unawareness, prevalence, trends and risk factors: Na-
43. Hove MN, Kristensen JK, Lauritzen T, Bek T. The prevalence tional Health and Nutrition Examination Survey
of retinopathy in an unselected population of type 2 diabetes (NHANES) 1999e2010. J Int Med Res. 2017;45(2):
patients from Arhus County, Denmark. Acta Ophthalmol 594e609.
Scand. 2004;82(4):443e448. 49. Majeed A, El-Sayed AA, Khoja T, et al. Diabetes in the
44. Raymond NT, Varadhan L, Reynold DR, et al. Higher prev- Middle-East and North Africa: an update. Diabetes Res Clin
alence of retinopathy in diabetic patients of South Asian Pract. 2014;103(2):218e222.
ethnicity compared with white Europeans in the community: a 50. Alhyas L, McKay A, Balasanthiran A, Majeed A. Quality of
cross-sectional study. Diabetes Care. 2009;32(3):410e415. type 2 diabetes management in the states of the Co-operation
45. Wong TY, Klein R, Islam FM, et al. Diabetic retinopathy in a Council for the Arab States of the Gulf: a systematic review.
multi-ethnic cohort in the United States. Am J Ophthalmol. PLoS One. 2011;6(8):e22186.
2006;141(3):446e455. 51. Javadi MA, Zarei-Ghanavati S. Cataracts in diabetic patients: a
46. Romero-Aroca P. Managing diabetic macular edema: the review article. J Ophthalmic Vis Res. 2008;3(1):52e65.
leading cause of diabetes blindness. World J Diabetes. 52. World Health Organization. Global Report on Diabetes.
2011;2(6):98e104. Geneva: World Health Organization; 2016.

12