American Journal of Epidemiology Vol. 147, No.

5
Copyright O 1998 by The Johns Hopkins University School of Hygiene and Public Health Printed In U.S.A
All rights reserved

Prevalence and Risk Factors for Diabetic Retinopathy in the Multiethnic

Population of Mauritius

G. K. Dowse,1 A. R. G. Humphrey,2 V. R. Collins,3 W. Plehwe,3 H. Gareeboo,4 D. Fareed,4 F. Hemraj,4

H. R. Taylor,5 J. Tuomilehto,8 K. G. M. M. Alberti,7 and P. Z. Zimmet3

This study examines the prevalence of, and risk factors for, diabetic retinopathy in Asian Indian, Chinese,
and Creole Mauritians in whom there is an increasing prevalence of non-insulin-dependent diabetes mellitus
(NIDDM). As part of a population-based survey on the Indian Ocean island of Mauritius in 1992, glucose
tolerance was classified using a 75-g oral glucose tolerance test on 6,553 persons. Subjects with newly
diagnosed (n = 358) or known diabetes (n = 388), and a random sample of one in four subjects with impaired
glucose tolerance (n = 165), had stereoscopic 45 retinal photographs taken of three fields in the right eye after

Downloaded from http://aje.oxfordjournals.org/ by guest on March 19, 2016

mydriasis. Photographs were graded according to a modified version of the Airiie House criteria. The
prevalence of nonproliferative and proliferate retinopathy was: 14.5% and 0.3%, respectively, in newly
diagnosed diabetic subjects; 42.0% and 2.3%, respectively, in known diabetic subjects; and 9.1% and 0%,
respectively, in persons with impaired glucose tolerance. Muslim Indians had the lowest prevalence of
retinopathy (10.8% and 34.0% for new and known diabetes, respectively), but after adjusting for other factors,
this was significantly different only to Creoles (18.8% and 53.8%, respectively). Univariate analysis revealed
significant differences between diabetic subjects with and without retinopathy in mean age, body mass index,
fasting and 2-hour plasma glucose levels, systolic and diastolic blood pressure, fasting triglycerides, serum
creatinine, and urinary albumin levels. For known diabetes, mean duration of diabetes and the proportion using
insulin were also greater in those with retinopathy. Multivariate analysis using logistic regression confirmed
that increasing duration of diabetes, fasting plasma glucose, systolic blood pressure, and urinary albumin
concentration, and decreasing body mass index, were independently associated with retinopathy. The high
prevalence of diabetic retinopathy observed in all major ethnic groups in Mauritius portends a serious public
health problem, given the relative recency of the NIDDM epidemic in that country and the limited resources for
laser photocoagulation. Strategies to minimize this problem among those already known to have diabetes
should include strict control of plasma glucose and blood pressure. Aw J Epidemiol 1998;147:448-57.

diabetes mellitus, non-insulin-dependent; diabetic retinopathy; ethnic groups; prevalence; risk factors

Diabetic retinopathy is the leading cause of blind-
ness in developed countries (1). In many developing
countries the incidence and prevalence of non-insulin-
dependent diabetes mellitus (NIDDM) far exceed rates
in the developed world (2), but facilities for the de-
tection and treatment of retinopathy are limited. Con-
sequently, diabetic retinopathy has the potential to
become a serious public health problem in these pop-
ulations. Few studies of diabetic retinopathy have
Received for publication November 5, 1996, and accepted for
publication Jury 16, 1997.
G. K. Dowse and A. R. G. Humphrey are joint first authors.
Abbreviations: Cl, confidence interval; NIDDM, non-insulln-
dependent diabetes mellitus.
1
Public Hearth Division, Hearth Department Perth, Western Aus-
tralia.
2
Redbridge and Wattham Forest Health Authority, Essex, United
Kingdom.
used standardized criteria for case definition, and
population-based studies have been confined mainly to
developed countries. The multiethnic and rapidly de-
veloping population of Mauritius has a high preva-
lence of NIDDM, even in young adults (3). In this
paper, we report the results of a population-based
study of diabetic retinopathy using standardized pho-
tographic methods (4) in Indian, Creole, and Chinese
Mauritians.
3
lntemational Diabetes Institute, Melbourne, Australia.
Ministry of Health, Port Louis, Mauritius.
'Department of Ophthalmology, University of Melbourne, Mel-
bourne, Australia.
"Diabetes and Genetic Epidemiology Unit, National Public Hearth
Institute, Helsinki, Finland.
7
Human Diabetes and Metabolism Research Centre, Univer-
sity of Newcastle Upon Tyne, Newcastle Upon Tyne, United
Kingdom.

448

MATERIALS AND METHODS
Background and study population
Mauritius is an independent island nation situated in
the Indian Ocean 800 km east of Madagascar. Approx-
imately 70 percent of the population are of Indian
(Hindu and Muslim) ethnic origin, 2 percent are Chi-
nese, and the majority of the remaining 28 percent are
Creoles, being predominantly of African and Mala-
gasy ancestry with some European admixture. The
1992 Mauritius noncommunicable disease survey was
performed in 14 geographically defined population
clusters, 11 of which had been investigated in an
earlier survey in 1987 (3, 5). The overall response rate
in 1992 was 89.1 percent, with 6,553 attenders. A
major aim of the study was to investigate the compli-
cations of NIDDM in Mauritians, and retinal photo-
graphs were taken for the first time in 1992.
During the baseline (1987) survey, 10 population
clusters were randomly selected and one area was
purposively selected for its high concentration of Chi-
nese residents, as described previously (3). In 1992, an
additional three clusters were selected for their pre-
dominant ethnic group, making a total of 10 randomly
chosen areas and four ethnically selected areas (Chi-
nese plus Hindu, Muslim, and Creole). In the original
11 clusters, all resident adults aged 2574 years were
invited to attend. In the three new clusters, stratified
sampling was used to select equivalent numbers of
men and women in three age strata (35-44,45-54, and
55-64 years) (5).
Survey procedure
The following procedures were carried out on all
participants as described previously (3, 5-7): measure-
ment of height and weight, and waist and hip circum-
ferences; systolic and fifth phase diastolic blood pres-
sure measurements in duplicate using a standard
mercury sphygmomanometer; and an interviewer-
administered medical history questionnaire in Creole
(the lingua franca). All subjects not on diabetic med-
ication had a 2-hour 75-g (dextrose monohydrate) oral
glucose tolerance test (8). Fasting and 2-hour venous
blood samples were centrifuged and separated imme-
diately, and plasma glucose was measured on site
using YSI glucose analyzers (Yellow Springs Instru-
ment Company, Yellow Springs, Ohio) within 3 hours
of collection. Fasting triglycerides, total cholesterol,
and creatinine levels were measured in serum at the
Central Laboratory, Mauritius, using an automated
Chemistry Profile Analyzer (Model LS) (Coultronics,
France).
Hypertension was defined according to World
Health Organization criteria (9) as a mean diastolic
Am J Epidemiol Vol. 147, No. 5, 1998
Retinopathy in Mauritius 449
blood pressure ^95 mmHg or a mean systolic blood
pressure s i 6 0 mmHg and/or treatment with hypoten-
sive drugs (6). Diabetes was defined according to
World Health Organization criteria (3, 5, 8). Briefly,
subjects reporting a history of diabetes had "known"
diabetes if they were currently taking oral agents or
insulin, or if they had a fasting plasma glucose S:7.8
mmol/liter or a 2-hour plasma glucose > 11.1 mmol/
liter. "New" diabetic subjects were those without a
diabetic history who had a 2-hour plasma glucose
a 11.1 mmol/liter, or in the absence of a 2-hour sam-
ple, a fasting plasma glucose >7.8 mmol/liter. Im-
paired glucose tolerance was defined by 2-hour plasma
glucose 5:7.8 mmol/liter and <11.1 mmol/liter, and
subjects with normal glucose tolerance had a 2-hour
plasma glucose concentration <7.8 mmol/liter.
Downloaded from http://aje.oxfordjournals.org/ by guest on March 19, 2016
Retinopathy subsample
All new or known diabetic subjects and every fourth
subject with impaired glucose tolerance from the orig-
inal 11 survey clusters were eligible for the retinopa-
thy study. These subjects were transported by bus
from each of the field survey sites to the National Eye
Hospital at Moka within a few days of their glucose
tolerance tests. Distant visual acuity was assessed be-
fore mydriasis using a Snellen chart. First morning
clean urine samples were also collected from subjects
eligible for the retinopathy substudy, and albumin
concentration was determined by an in-house radioim-
munoassay (10) (interassay coefficient of variation =
5.0 percent, intraassay coefficient of variation = 2.4
percent) on samples transported frozen to Newcastle
Upon Tyne, United Kingdom.
After excluding subjects who could not be photo-
graphed, or whose photographs were not gradable,
valid retinopathy data were available for 340 known
diabetic subjects, 324 newly diagnosed diabetic sub-
jects, and 165 subjects with impaired glucose toler-
ance, which represents 83, 81, and 74 percent, respec-
tively, of those eligible in each status of glucose
tolerance from the main survey sample. In addition,
retinal photographs were taken and visual acuity as-
sessed in 48 known and 34 newly diagnosed diabetic
subjects from the 12th survey area. For logistic rea-
sons, retinal photographs could not be taken in the
final two survey areas. Therefore, a total of 911 sub-
jects from 12 survey areas had valid retinal photo-
graphs, and are the focus of this report.
Stereoscopic 45 retinal photographs in three over-
lapping fields of the right eye only were taken using a
Topcon TRC-50VT camera (Topcon Corporation,
Tokyo, Japan). The left eye was used if it was not
possible to photograph the right retina. The three pho-
tographic fields (with overlap) were: centered on the
450 Dowse et al.

optic disc; macula (temporal to the optic disc); and

in in OJ in
nasal to the disc. Photographs were graded by a trained to a j

and certified assessor (W.P.) masked to all subject

information. Photographs were compared with Airlie
House reference photographs, and a level of severity
of diabetic retinopathy was assigned by a modification T-; CO CO OJ

of the 191 system (4). Photographs for every tenth ig

subject were later regraded by the same assessor to
ensure internal validity. The gradings were catego- co co co
o d oi ^
rized for the purposes of this study into: no diabetic
retinopathy (levels 10-15); nonproliferative retinop-
athy (levels 20-55); and proliferative diabetic retinop-
athy (levels 61-85).

Statistical analysis
Analyses were performed using the Statistical Pack-
age for the Social Sciences (SPSS PC+) (11). Preva- o

Downloaded from http://aje.oxfordjournals.org/ by guest on March 19, 2016

lence estimates were age-standardized by the direct
method (12) using the diabetic population from the
s
1987 survey in 10-year age-bands as standard. Anal-
yses of risk factor associations have been limited to g
newly diagnosed and known diabetic subjects. Differ-
ences in mean values were compared using t tests and in co
1
p
o 0 i-
proportions were compared with the x2 test. Measures
T^

of glucose, triglycerides, serum creatinine, and urine

albumin were loge-transformed to normalize distribu-
tions for analysis, and results were back-transformed
for presentation as geometric means in the tables.
Tertiles of body mass index were calculated from the
whole survey population. Trends in proportions were
assessed using the x2 test for linear trend (12, 13). c\| c\j m to

Multiple logistic regression was used to assess the o

independent effect of the following parameters on the CD

presence of retinopathy: age, sex, ethnic group, dura-

tion of diabetes, fasting plasma glucose, total serum 3 3"
cholesterol, fasting serum triglycerides, urinary albu-
min, body mass index, waist-hip ratio, age at diagno-
sis, systolic blood pressure, smoking status, and alco- I
hol consumption. Categories are further described in
the tables. A forward stepwise selection procedure was
used with a p value of 0.05 as the criterion for entry of z l a.
the variables. Regression analyses were performed for
all subjects with diabetes, and then separately for new
o
and known diabetes. z

RESULTS
The prevalence of diabetic retinopathy according to
glucose tolerance status is shown in table 1. There was
3 a
no proliferative disease among the impaired glucose
tolerance group, although 9.1 percent had nonprolif-
O5
nil
erative retinopathy. Overall, 28.8 percent of diabetic o
subjects had nonproliferative retinopathy and 1.3 per-
cent had proliferative disease. The prevalence of reti-

Aw J Epidemiol Vol. 147, No. 5, 1998

 Retinopathy in Mauritius 451

nopathy in known diabetic subjects was three times
higher than in newly diagnosed cases. Women tended
to have lower prevalences of total retinopathy than
men in each class of glucose tolerance, but the differ-
ences were not significant.
Table 2 illustrates characteristics of the diabetic
subsample according to retinopathy status. There were
significant differences in a number of variables: sub-
jects with retinopathy were older, had higher mean
fasting and 2-hour plasma glucose, systolic and dia-
stolic blood pressures, triglycerides, serum creatinine,
and urinary albumin, and had lower mean body mass
index. Among the known diabetic group, retinopathy
was associated with longer mean duration of diabetes
and with insulin treatment. Adjusting for duration and
age did not appreciably change these results.
Among known diabetic subjects, the age-standardized
newly diagnosed diabetic subjects, the prevalence of
retinopathy varied from 3.8 percent in Chinese to 18.9
percent in Creoles. When all diabetic subjects were
considered together, Muslim rates remained the lowest
overall (table 3). After stratification by fasting plasma
glucose levels and duration, Muslims had a lower
prevalence of retinopathy compared with other ethnic
groups (combined) in all but one category (data not
shown). Muslims and Chinese had higher ascertain-
ment of diabetes than the other ethnic groups, with
55.4 percent and 56.1 percent, respectively, of all
cases known, compared with 47.2 percent and 49.4
percent, respectively, among Hindus and Creoles.
Among known diabetic subjects: insulin treatment was
more common in Muslims (7.8 percent) than in Hin-
dus (2.9 percent), Creoles (6.0 percent), or Chinese
(6.3 percent); "poor control" (fasting plasma glucose

Downloaded from http://aje.oxfordjournals.org/ by guest on March 19, 2016

prevalence of retinopathy was over 45 percent in concentration >11.1 mmol/liter) was less frequent in
Hindu Indian, Creole, and Chinese subjects, with a Muslims (26 percent) and Chinese (22 percent) than in
lower rate (30.6 percent) in Muslim Indians. Among Hindus (39 percent) and Creoles (47 percent); but
mean duration and mean age at diagnosis varied little
TABLE 2. Characteristics of diabetic subjects with and
between ethnic groups (data not shown).
without retinopathy, Mauritius 1992 The prevalence of any retinopathy increased signif-
No
Retino- P
icantly with duration of diabetes (x 2 test for trend =
Characteristic retino-
pathy pathy valuet 127, p <0.0001) and with increasing fasting plasma
All diabetes
glucose (x* test for trend = 40.3, p <0.0001) (table 4).
No. 521 225 Over 50 percent of subjects with self-reported duration
Male (%) 43.6 48.9 0.18 of 5 or more years had retinopathy. Hypertensive
Age (years) 51.8 54.6 0.002 subjects had a generally higher prevalence of retinop-
Age at diagnosis (years) 49.7 48.1 0.08 athy than nonhypertensive subjects even after stratifi-
visual acuity <6/12 (%) 13.3 15.2 0.50
Current smoker (%) 19.2 16.2 0.34
cation by fasting plasma glucose and duration catego-
Alcohol 3 units/day 7.9 4.5 0.09 ries (table 5) or albuminuria status (figure 1). The
Body mass index (kg/m) 26.7 25.7 0.004 prevalence of retinopathy also increased with rising
Waist-hip ratio 0.91 0.92 0.09 levels of urinary albumin concentration irrespective of
Fasting plasma glucose
hypertension status (figure 1). Overall, 75 percent of
(mmol/liter)* 8.4 10.4 <0.001
2-hour plasma glucose subjects with macroalbuminuria had retinopathy.
(mmol/liter)* 15.0 16.6 0.002 Greater weight loss in more severe diabetes may
Systolic blood pressure have confounded the relation between body mass in-
(mmHg) 133 138 0.003
Diastollc blood pressure
dex and retinopathy, so the data were stratified by new
(mmHg) 80 83 0.007 and known diabetes (figure 2). The inverse relation
Total cholesterol (mmoi/lrter) 5.1 5.3 0.15 between retinopathy and body mass index was clearly
Fasting triglycerides stronger in known diabetic subjects (x2 for trend =
(mmol/llter)* 1.92 2.13 0.03 13.9, p <0.001) than in newly diagnosed subjects (jf
Serum creatinine Oimol/liter)* 90 101 <0.001
Urinary albumin (jig/ml)* 2.33 4.86 <0.001 for trend = 2.2, p = 0.14). By contrast, there was no
relation between waist-hip ratio and retinopathy (data
Known diabetes only not shown).
No. 216 172
For combined new and known diabetes, multiple
Duration (years) 4.9 8.5 <0.001
Oral hypoglycemic treatment
logistic regression revealed that duration of diabetes,
(%) 80.1 78.5 0.38 fasting plasma glucose, systolic blood pressure, uri-
Insulin treatment (%) 2.3 8.1 0.006 nary albumin, and body mass index (inversely) were
Visual acuity <6/12 (%) 7.1 7.1 0.99 independently associated with retinopathy (table 6). A
Education (years) 5.2 5.3 0.70
known duration of 10 or more years carried an inde-
* Geometric means. pendent odds ratio for any retinopathy of 9.0, relative
t From x 1 test for proportions and f test for means. to newly diagnosed diabetes. High fasting plasma glu-

Am J Epidemiol Vol. 147, No. 5, 1998

452 Dowse et al.

cose ("poor control") had an odds ratio of 2.9 relative

to "good control," and the odds of retinopathy were
1.2 for every 10 mmHg rise in systolic blood pressure.
o
Hi
For subjects with newly diagnosed diabetes only,
current age (equivalent to "age at diagnosis"), fasting cq r^ CD
plasma glucose, systolic blood pressure, and male sex cri in iri

conferred a significantly increased risk of retinopathy.

Among known diabetic subjects, duration, fasting
plasma glucose, systolic blood pressure, urinary albu-
min, and younger age at diagnosis were independently
associated with retinopathy.
When a variable comparing each ethnic group to S!
j co Ln
Muslims was forced into selected models, Creoles had
statistically significant odds ratios of 1.9 (95 percent
confidence interval (CI) 1.1-3.3) and 2.4 (95 percent o
OJCS ^
CI 1.2-4.8) for retinopathy in "all" and "known"
diabetes, respectively, relative to Muslims. Other eth-

Downloaded from http://aje.oxfordjournals.org/ by guest on March 19, 2016

nic comparisons were not significant: Hindus relative
to Muslims had an odds ratio of 1.5 (95 percent CI
0.9-2.5) for "all" and 1.7 (95 percent CI 0.9-3.1) for in m
o
"known" diabetes; and Chinese relative to Muslims o cJ T^

had an odds ratio of 1.0 (95 percent CI 0.4-2.2) for

"all" and 1.5 (95 percent CI 0.9-2.5) for "known"
diabetes.

DISCUSSION
o
Prevalence
We have shown previously that the prevalence of
NTDDM in Mauritius is high, and that it varies little
between ethnic groups (3). This study shows that di-
abetic retinopathy is also common in Mauritians, irre-
spective of ethnic background. To our knowledge, II?
there have been no other studies of diabetic retinop-
athy using standardized photographic methods in In-
Q.
dian, African-origin Creole, or Chinese populations.
Given that the NIDDM epidemic in Mauritius is rel- 6 S8S
atively recent, there is a high prevalence in young
adults, and there is a high proportion of newly diag- CO C\j CD

nosed cases (3), the prevalence of retinopathy is likely

to increase in the future, as average duration of dia-
betes increases.
The overall prevalence of retinopathy in diabetic
Mauritians (30 percent) was higher than that found in
Nauruans (24 percent) using direct and indirect oph- 09 U)

thalmoscopy (14), and similar to that determined using if* ONT-'

a
a non-mydriatic camera in Pima Indians (34 percent) in cq i^.
<o c\i cri
(15). However, caution must be exercised when com-
paring prevalence determined by different methods,
particularly studies which have relied on ophthalmos-
copy (16). Klein et al. (17) have demonstrated that
direct ophthalmoscopy through an undilated pupil has if,
an agreement of only 54 percent for grading of reti-
nopathy defined on the basis of stereoscopic photo-

Am J Epidemiol Vol. 147, No. 5, 1998

 Retinopathy in Mauritius 463

TABLE 4. Prevalence of retinopathy In diabetic subjects, in our ascertainment of retinal lesions should, therefore,
relation to duration of diabetes and fasting plasma glucose, be high.
Mauritius 1992
Because of the difficulties of direct comparison of
Retinopathy status (%) rates between studieseven those using stereoscopic
Non-
No
pro Bfera-
Proltfera-
All
retino-
photographs and modified Airlie House gradingit is
tlve
tive pathy instructive to compare rates between ethnic groups
Duration (years) determined within a single study by the same investi-
Newly diagnosed 358 14.5 0.3 14.8 gators. In Mauritius, Muslim Indians seemed to have
1-4 176 27.3 1.7 29.0 lower susceptibility to retinopathy than Hindu Indians
5-9 106 45.3 0.0 45.3 and Creoles, but after adjusting for other factors, only
210 105 63.8 5.7 69.5
the latter difference remained statistically significant.
Fasting plasma glucose However, the prevalence in the ethnically distinct
(mmot/liter) Hindu Indians and Creoles was similar. These results
<7.8 285 16.8 0.7 17.5 suggest that nongenetic factors were most likely to
7.8-11.0 260 27.7 1.5 29.2 explain the "ethnic" differences in prevalence. Fasting
201 47.3 2.0 49.3
glucose levels are reasonably reliable indicators of
long-term metabolic control in NIDDM (24). Muslim

Downloaded from http://aje.oxfordjournals.org/ by guest on March 19, 2016

graphs in three standard fields. However, after mydri-
asis and training of ophthalmoscopists agreement of
86 percent has been reported (18). Agreement of non-
mydriatic photography with three standard stereo-
scopic fields was 82.5 percent (17).
As demonstrated in table 7, when comparison is
limited to studies of known diabetes using retinal
photographs graded according to the modified Airlie
House classification (4), the prevalences of retinop-
athy in Indian, Creole, and Chinese Mauritians are
broadly similar to levels in other populations. This
includes Caucasians (19-21), Hispanic- (19) and
Mexican-Americans (20), and Polynesians (22). Sim-
ilarly, the prevalence determined using comparable
photographic methods in newly diagnosed NIDDM
detected on the basis of glucose tolerance tests in
population-based studies are in general agreement. For
example, our rates for diabetic retinopathy in the total
sample of new NIDDM (15 percent), and more spe-
cifically Hindu Indians (18 percent), Muslim Indians
(13 percent), Creoles (19 percent), and Chinese (4
percent), compare with 15 percent in Samoans (22), 16
percent in Mexican-Americans (20), and 14 percent in
Caucasians (20).
However, variability in the distribution of age and
disease duration, and remaining methodological dif-
ferences including the number of photographic fields,
the use of one versus two eyes, and whether levels 14
and 15 (exudates or hemorrhages without microaneu-
rysm) (4) are classified as diabetic retinopathy, may
still influence comparability of studies. Moss et al.
(23) have previously shown good agreement (92 per-
cent sensitivity) between classification of "any reti-
nopathy" based on three versus seven photographic
fields. Although their fields did not correspond with
those used in the Mauritius study, our use of a wider
angle lens ensured greater coverage of the retina, and
Indians had lower levels in this study than other ethnic
groups, and the likelihood of better control was also
supported by a higher ascertainment rate and more
frequent insulin treatment. Nevertheless, the differ-
ence between Muslims and Creoles remained after
controlling for other factors, including glucose levels
and duration. The reason for the lower rates of reti-
nopathy in newly diagnosed diabetic Chinese Mauri-
tians is unclear, but may reflect better diabetes case-
finding in Chinese, particularly given that rates in
Chinese with known NIDDM were similar to those of
other ethnic groups.
Evidence for ethnic differences in other studies is
contradictory, and on the whole, also unconvincing.
Harris et al. (25) found a marginally elevated risk of
retinopathy in African-American men, but not women,
compared with Caucasians. In the San Luis Valley
study, non-Hispanic white known diabetic subjects
had higher rates of duration-adjusted retinopathy
than Hispanic subjects (19). By contrast, in another
population-based study, retinopathy was more com-
mon in Mexican-Americans than in Caucasians (20).
Haffner et al. (26) hypothesized that greater insulin
resistance in Mexican-Americans might explain the
apparent excess risk of retinopathy compared with
Caucasians, but this cannot explain the Mauritian re-
sults, as Muslim Indians have the highest levels of
basal and stimulated insulin (27).
Risk factors
Duration of diabetes was the strongest risk factor for
retinopathy in diabetic Mauritians. Compared with
newly diagnosed subjects, those with a known dura-
tion of 10 or more years had a ninefold increased risk
of retinopathy after controlling for glycemia and other
risk factors. Duration has been consistently identified
as a strong risk factor for retinopathy in cross-sectional

Am J Epidemiol Vol. 147, No. 5, 1998

454 Dowse et al.

TABLE 5. Prevalence of retinopathy (%) In nonhypertensive and hypertensive diabetic subjects, stratified by duration of
diabetes and fasting plasma glucose concentration, Mauritius 1992
Nonhypertensive Hypertensive
Fasting plasma glucose (mmoMiter) Fasting plasma glucose (mmot/liter)
Duration <7.8 7.8-11.0 <7.8 7.8-11.0
(years)
Total Total Total Total Total Total
no. no. no. no. no. no.
Newly diagnosed 133 6.8 72 12.5 53 30.2 61 24.6 29 3.4 9 33.3
1-4 26 19.2 54 22.2 42 35.7 29 31.0 14 35.7 11 45.5
5-9 13 15.4 32 40.6 27 55.6 6 50.0 16 43.8 12 66.7
10 9 66.7 28 67.9 26 80.8 7 14.3 13 69.2 21 76.2

Prevalence (%)
100
I Not Hypertensive IS3 Hypertensive

80

Downloaded from http://aje.oxfordjournals.org/ by guest on March 19, 2016

10

60

40
177
453 f

20

Normal Mlcroalbuminuria Macroalbuminuria

FIGURE 1. Prevalence of retinopathy in hypertensive and nonhypertensive subjects with diabetes stratified by urinary albumin levels:
Normal = <30 mg/ml, Mlcroalbuminuria = 30-299 mg/ml, and Macroalbuminuria = ^300 mg/ml urinary albumin. Mauritius, 1992.

(14, 16, 20-22) and longitudinal (28-30) studies. Du-
ration most likely reflects a complex "summary" of
long-term exposure to other known and putative risk
factors, including glucose and blood pressure levels.
As discussed by Harris et al. (31), true onset of
NIDDM may commonly occur several years before
clinical diagnosis. This is reflected in the fact that
retinopathy is common in subjects at the time of di-
agnosis, whether the latter occurs because of symp-
tomatic presentation or through detection by screening
asymptomatic individuals, as in this study.
Nonproliferative retinopathy was found in 9.1
percent of Mauritians with impaired glucose toler-
ance. This is similar to findings in a study of Sa-
moans which used identical methods (22), and to
results using other photographic regimes (32-34).
Klein et al. (34) concluded that retinal blot hemor-
rhages or microaneurysms are observed at low fre-
quency in the general population (including those
with normal glucose tolerance) and may reflect
other factors including blood pressure, age, and
platelet count. It is also possible that some of the
retinopathy observed in Mauritian subjects with im-
paired glucose tolerance may have been due to
misclassification of, or improvement in, glucose
tolerance or misgrading of retinal photographs.
We found a higher prevalence of diabetic retinop-
athy in men, and after controlling for other factors, this
retained significance in those with newly diagnosed
diabetes. Other studies have reported small excesses of
retinopathy in Caucasian (26), Mexican-American
(26), Samoan (22), and African-American (25) men,
but in general, these differences appear to be explained
by other risk factors.
After controlling for fasting glucose and other vari-
ables, current age was an important risk factor for
retinopathy only in newly diagnosed Mauritian dia-
betic subjects, in whom it may have behaved as a
surrogate marker of duration. Among known diabetic
Mauritians, age at diagnosis was inversely and inde-
pendently associated with retinopathy. A similar rela-
tion has been observed in other populations (19, 26,

Am J Epidemiol Vol. 147, No. 5, 1998

 Retinopathy in Mauritius 455

Prevalence (%)
60

140 IB New Diabetes

50 ^ H Known Diabetes

40

30

20
- 1 i 116

10 123 ^^H

0
l l IBM I MB
Tertiles of Body Mass Index

Downloaded from http://aje.oxfordjournals.org/ by guest on March 19, 2016

FIGURE 2. Prevalence of retinopathy in subjects with newly diagnosed and known diabetes by tertiles of body mass index. Tertiles were
derived from the whole study group as: 1 = <24.4 kg/m 2 , 2 = 24.4-27.8 kg/m 2 , and 3 = 2:27.9 kg/m 2 . Mauritius, 1992.

TABLE 6. Logistic regression analysis* for factors associated with diabetic retinopathy, Mauritius 1992

Diabetes
All New Known
Variable (n = 727) (n = 347) (n = 380)
ORt 95%Clt OR 95% Cl OR 95% Cl
Duration of diabetes (years) (p< 0.0001) (p< 0.0001)
Newly diagnosed 1.00 Not applicable
1-4 2.05 1.28-3.27 1.00
5-9 3.77 2.23-6.37 1.96 1.14-3.35
10 8.95 5.17-15.50 4.03 2.28-7.13
Fasting plasma glucose (mmol/Dter) (p< 0.0001) (p = 0.0001) (p = 0.012)
Good control (<7.8) 1.00 1.00 1.00
Fair control (7.8-11.0) 1.23 0.0-1.96 0.77 0.34-1.71 1.45 0.79-2.70
Poor control (211.1) 2.87 1.79-4.61 4.26 2.03-B.96 2.51 1.33-4.76
Body mass Index (tertiles) (p= 0.017)
<24.4 1.00
24.4-27.9 0.69 0.44-1.06
>27.9 0.52 0.32-0.82
Systolic blood pressure (per 10 mmHg) 1.18 1.08-12B (p = 0.0001) 1.18 1.02-8.96 (p 0.026) 1.18 1.04-1.31 (p = 0.009)
Age (per 10 years) 1.40 1.04-1.86 (p = 0.019)
Age at diagnosis (per 10 years) 0.78 0.61-0.99 (p = 0.049)
Female sex 0.51 0.27-0.95 (p = 0.035)
Urinary albumin (per 50 mg/ml) 1.14 1.04-1.24 (p = 0.006) 1.20 1.04-1.37 (p = 0.009)
Forward stepwlse selection procedure. Odds ratios are shown onty for the variables selected ki the particular model. Other variables not entering any model
were: ethnic group; fasting total cholesterol; fasting trtgtycerfdes; smotdng status (present smoker/nonsmoker); alcohol consumption (S3 units per day vs. <3 units
per day).
t OR, odds ratio; Cl, confidence Interval.

28), although age-at-diagnosis is not as strong a risk
factor as disease duration.
The inverse relation between body mass index and
retinopathy in Mauritians remained significant even
after controlling for confounding by duration of dia-
betes, with which body mass index also has an inverse
association. An inverse relation between body mass
index and retinopathy has been found in several other
studies (14, 16, 21, 22). It is likely that leanness is a
marker for severity of diabetes, and, hence, exposure
to prolonged hyperglycemia.
Hyperglycemia is clearly a major risk factor in the
development of retinopathy in Mauritians, as demon-
strated in both cross-sectional and longitudinal studies
in several populations (14-16, 19-22, 28-30). In the
absence of measures of glycated hemoglobin we have
used a single fasting plasma glucose measurement as a
marker for control. Although this may not reflect long-

Am J Epidemiol Vol. 147, No. 5, 1998

456 Dowse et al.

TABLE 7. Crude prevalence of retlnopathy determined using Alrlle House criteria on retinal photographs In people with known
non-lnsulln-dependent diabetes mellitus from selected ethnic groups
Mean Retlnopathy prevalence (%)
Age
duration Sample
Reference Ethnictty range Non-
of diabetes size Proffierattve
(years) proflferattve
(years)

Klein eta)., 1984(21) Caucasian a30 8.8 696 35.5 2.9

Hammanetal., 1989 (19) Hispanic 20-74 10.5 166 36.2 6.6
Caucasian 20-74 4.7 85 47.0 4.7
Haffner et al., 1993(26) Mexican American 30-70 8.0 231 54.6
Caucasian 30-70 8.1 478 41.2
Collins et al., 1995(22) Polynesian 25-74 4.3 166 38.7 4.5
Present study Hindu Indian 25-79 6.3 174 42.8 2.9
Muslim Indian 25-79 6.0 103 33.0 1.0
Creole 25-79 6.6 83 51.3 2.5
Chinese 25-79 8.7 32 43.8 3.1

term control as well as glycated hemoglobin (35), urinary albumin concentration. These results are of
there is Little difference between these tests as predic- great significance given the extremely high and rising

Downloaded from http://aje.oxfordjournals.org/ by guest on March 19, 2016

tors of retinopathy in NIDDM (15, 24, 36). Tight
control has been shown to reduce retinopathy in
IDDM (37), and the randomized United Kingdom
Prospective Diabetes Study is currently attempting to
determine whether glycemic control can also reduce
retinopathy in NIDDM (38).
The relation between systolic blood pressure and
retinopathy is well recognized, and our results support
those of previous studies (16, 19, 21, 39, 40). Al-
though controlling for nephropathy (as urinary albu-
min concentration) weakened the association, systolic
blood pressure remained an important risk factor for
retinopathy in multivariate models. In IDDM, hyper-
tension is nearly always associated with nephropathy
(41), whereas in NIDDM hypertension is present in
many people at diagnosis in the absence of nephrop-
athy. The pathogenesis of renal microangiopathy may
be very similar to that of retinal microangiopathy, and
both may be accelerated by high blood pressure (42).
The strong association between retinopathy and uri-
nary albumin concentration in Mauritians has been
observed in other populations (22, 43) and reinforces
the importance of monitoring albumin excretion as a
marker of vascular damage in diabetes. As Knowler et
al. (39) have suggested, hypertension is an eminently
modifiable risk factor for retinopathy in NIDDM pa-
tients. Mean blood pressure level and the prevalence
of hypertension were lower in Muslim Indians than in
other ethnic groups in Mauritius (6), perhaps contrib-
uting to the lower prevalence of retinopathy observed
in Muslim Indians.
In conclusion, our study has shown relatively high
levels of diabetic retinopathy in all ethnic groups in
Mauritius, including asymptomatic subjects with dia-
betes diagnosed during screening. Independent risk
factors identified were duration of diabetes, fasting
plasma glucose, systolic blood pressure, leanness, and
prevalence of NTDDM in Mauritius (5) coupled with
the relatively early age of onset of diabetes and limited
resources for tertiary preventive therapy through laser
photocoagulation. As in other populations of the de-
veloping world with high rates of NIDDM, effective
primary prevention, early detection of diabetes, and
quality clinical care and education, focussing particu-
larly on blood pressure and glycemic control, provide
the key to averting an epidemic of blindness and other
diabetic complications.
ACKNOWLEDGMENTS
This study was undertaken with the support and collab-
oration of the Government of Mauritius and the World
Health Organization, and was supported by grant no. DK-
25446 from the National Institute of Diabetes and Digestive
and Kidney Diseases; and the British Diabetic Association
and NovoNordisk (United Kingdom) for laboratory assays
in Newcastle Upon Tyne.
We thank the many staff of the participating institutions
who worked in the survey, and especially acknowledge Mr.
J. Dabee, Dr. Gaya and colleagues at the Moka Eye Hospital
for their assistance, and Mr. Jim Rule for performing the
retina] photography.
REFERENCES
1. Operational Research Department, National Society for the
Prevention of Blindness. Vision problems in the United States:
a statistical analysis. New York, NY: National Society for the
Prevention of Blindness, 1980:1-46.
2. Zimmet P, Dowse G, Finch C, et al. The epidemiology and
natural history of NIDDMlessons from the South Pacific.
Diabetes Metab Rev 1990;6:91-124.
3. Dowse GK, Gareeboo H, Zimmet PZ, et al. High prevalence
of NIDDM and impaired glucose tolerance in Indian, Creole,
and Chinese Mauritians. Diabetes 1990;39:390-6.

Am J Epidemiol Vol. 147, No. 5, 1998


4. Diabetic retinopathy study. Report no. 7. A modification of
the Airlie House classification of diabetic retinopathy. Invest
Ophthalmol Vis Sci 1981;21:210-26.
5. Dowse GK, Gareeboo H, Alberti KGMM, et al. Changes in
population cholesterol concentrations and other cardiovascu-
lar risk factor levels after five years of the non-communicable
disease intervention programme in Mauritius. BMJ 1995 ;311:
1255-9.
6. Nan L, Tuomilehto J, Dowse G, et al. Prevalence and medical
care of hypertension in four ethnic groups in the newly-
industrialized nation of Mauritius. J Hypertens 1991;9:
859-66.
7. Dowse GK, Zimmet PZ, Gareeboo H, et al. Abdominal obe-
sity and physical inactivity as risk factors for NIDDM and
impaired glucose tolerance in Indian, Creole, and Chinese
Mauritians. Diabetes Care 1991 ;14:271-82.
8. Diabetes mellitus: report of a WHO study group. WHO tech-
nical report series no. 727. Geneva, Switzerland: World
Health Organization, 1985.
9. Arterial hypertension: report of a WHO expert committee.
WHO technical report series no. 628. Geneva, Switzerland:
World Health Organization, 1978.
10. Christensen C, 0rskov C. Rapid screening PEG radioimmu-
noassay for quantification of pathological microalbuminuria.
Diabet Nephrop 1984;3:92-97.
11. NoruSis MJ. Statistical package for the social sciences (SPSS/
PC+) statistics 4.0. Chicago, IL: SPSS Inc, 1990.
12. Armitage P, Berry G. Statistical methods in medical research.
3rd ed. Oxford, England: Blackwell Scientific, 1994.
13. Epi info version 5.0. Atlanta, GA: Centers for Disease Control
and Prevention, Epidemiology Program Office, 1990.
14. King H, Balkau B, Zimmet P, et al. Diabetic retinopathy in
Nauruans. Am J Epidemiol 1983;117:659-67.
15. Liu QZ, Pettitt DJ, Hanson RL, et al. Glycated haemoglobin,
plasma glucose and diabetic retinopathy: cross-sectional and
prospective analyses. Diabetologia 1993;36:428-32.
16. Diabetes Drafting group. Prevalence of small vessel and large
vessel disease in diabetic patients from 14 centres: the World
Health Organization Multinational Study of Vascular Disease
in diabetics. Diabetologia 1985;28:615-40.
17. Klein R, Klein BEK, Neider W, et al. Diabetic retinopathy as
detected using ophthalmoscopy, a non-mydriatic camera, and
a standard fundus camera. Ophthalmology 1985;92:485-91.
18. Moss SE, Klein R, Kessler SD, et al. Comparison between
ophthalmoscopy and fundus photography in determining se-
verity of diabetic retinopathy. Ophthalmology 1985;92:62-7.
19. Hamman RF, Mayer EJ, Moo-Young GA, et al. Prevalence
and risk factors of diabetic retinopathy in non-Hispanic whites
and Hispanics with NIDDM. Diabetes 1989;38:1231-7.
20. Haffner SM, Fong D, Stern MP, et al. Diabetic retinopathy in
Mexican Americans and non-Hispanic whites. Diabetes 1988;
37:878-84.
21. Klein R, Klein BEK, Moss SE, et al. The Wisconsin Epide-
miologic Study of Diabetic Retinopathy. III. Prevalence and
risk of diabetic retinopathy when age at diagnosis is 30 or
more years. Arch Ophthalmol 1984; 102:527-32.
22. Collins VR, Dowse GK, Plehwe WE, et al. High prevalence of
diabetic retinopathy and nephropathy in Polynesians of West-
em Samoa. Diabetes Care 1995; 18:1140-9.
23. Moss SE, Meuer SM, Klein R, et al. Are seven standard
photographic fields necessary for classification of diabetic
retinopathy? Invest Ophthalmol Vis Sci 1989;30:823-8.
24. Nathan DM, Singer DE, Godine JE, et al. Retinopathy in older
type II diabeticsassociation with glucose control. Diabetes
1986;35:797-801.
25. Harris EL, Feldman S, Rahj Robinson C, et al. Racial differ-
ences in the relationship between blood pressure and risk of
Am J Epidemiol Vol. 147, No. 5, 1998
Retinopathy in Mauritius 457
retinopathy among individuals with NIDDM. Diabetes Care
1993;16:748-54.
26. Haffner SM, Mitchell BD, Moss SE, et al. Is there an ethnic
difference in the effect of risk factors for diabetic retinopathy?
Ann Epidemiol 1993;3:2-8.
27. Dowse GK, Zimmet PZ, Alberti KGMM, et al. Serum insulin
distributions and reproducibility of the relationship between
2-hour insulin and plasma glucose levels in Asian Indian,
Creole, and Chinese Mauritians. Metabolism 1993;42:
1232-41.
28. Ballard DJ, Melton LJ, Dwyer MS, et al. Risk factors for
diabetic retinopathy: a population-based study in Rochester,
Minnesota. Diabetes Care 1986;9:334-42.
29. Nelson RG, Wolfe JA, Horton MB, et al. Proliferative reti-
nopathy in NIDDM: incidence and risk factors in Pima Indi-
ans. Diabetes 1989;38:435-40.
30. Klein R, Klein BEK, Moss SE, et al. The Wisconsin Epide-
miologic Study of Diabetic Retinopathy. X. Four-year inci-
dence and progression of diabetic retinopathy when age at
diagnosis is 30 years or more. Arch Ophthalmol 1989;107:
244-9.
31. Harris MI, Klein R, Welborn TA, et al. Onset of NIDDM
occurs at least 4-7 yr before clinical diagnosis. Diabetes Care
Downloaded from http://aje.oxfordjournals.org/ by guest on March 19, 2016
1992;15:815-19.
32. Klein R, Barrett-Connor EL, Blunt BA, et al. Visual impair-
ment and retinopathy in people with normal glucose tolerance,
impaired glucose tolerance, and newly diagnosed NIDDM.
Diabetes Care 1991;14:914-18.
33. Stolk RP, Vingerling JR, de Jong PTVM, et al. Retinopathy,
glucose, and insulin in an elderly population. The Rotterdam
Study. Diabetes 1995;44:11-15.
34. Klein R, Moss S, Klein BEK, et al. When is retinopathy diabetic
retinopathy? (Abstract). Am J Epidemiol 1991;134:740.
35. Duncan BB, Heiss G. Nonenzymatic glycosylation of
proteinsa new tool for assessment of cumulative hypergly-
cemia in epidemiologic studies, past and future. Am J Epide-
miol 1984;120:169-89.
36. McCance DR, Hanson RL, Charles MA, et al. Comparison of
tests for glycated haemoglobin and fasting and two hour
plasma glucose concentrations as diagnostic methods for dia-
betes. BMJ 1994;308:1323-8.
37. The Diabetes Control and Complications Trial Research
Group. The effect of intensive treatment of diabetes on the
development and progression of long-term complications in
insulin-dependent diabetes mellitus. N Engl J Med 1993,329:
977-86.
38. UK Prospective Diabetes Study Group. UK Prospective Dia-
betes Study (UKPDS). VIII. Study design, progress, and per-
formance. Diabetologia 1991;34:877-90.
39. Knowler WC, Bennett PH, Ballantine EJ. Increased incidence
of retinopathy in diabetics with elevated blood pressure: a
six-year follow-up study in Pima Indians. N Engl J Med
198O:3O2;645-5O.
40. Cignarelli M, De Cicco ML, Damato A, et al. High systolic
blood pressure increases prevalence and severity of retinop-
athy in NIDDM patients. Diabetes Care 1992;15:1002-8.
41. Simonson DC. Etiology and prevalence of hypertension in
diabetic patients. Diabetes Care 1988; 11:821-7.
42. Parving HH. Arterial hypertension in diabetes mellitus. In:
Alberti KGMM, Defronzo RA, Keen H, et al., eds. Interna-
tional textbook of diabetes mellitus. Chichester, England:
John Wiley and Sons, 1992:1521-34.
43. Gall MA, Rossing P, Sk0tt P, et al. Prevalence of micro- and
macroalbuminuria, arterial hypertension, retinopathy, and
large vessel disease in European type 2 (non-insulin-
dependent) diabetic patients. Diabetologia 1991 ;34:65561.