Systematic review or Meta-analysis Screening intervals for diabetic retinopathy and incidence of visual loss: a systematic review J. B. Echouffo–Tcheugui 1 , M. K. Ali 1 , G. Roglic 2 , R. A. Hayward 3,4,5 and K. M. Narayan 1 1 Hubert Department of Global Health, Rollins School of Public Health, Emory University, Atlanta, GA, USA, 2 Department of Chronic Diseases and Health Promotion, World Health Organization, Geneva, Switzerland, 3 Department of Internal Medicine, 4 Department of Health Management and Policy, University of Michigan, and 5 VA Center for Practice Management and Outcomes Research at the University of Michigan, Ann Arbor, MI, USA Accepted 18 June 2013 Abstract Screening for diabetic retinopathy can help to prevent this complication, but evidence regarding frequency of screening is uncertain. This paper systematically reviews the published literature on the relationship between screening intervals for diabetic retinopathy and the incidence of visual loss. The PubMed and EMBASE databases were searched until December 2012. Twenty five studies fulfilled the inclusion criteria, as these assessed the incidence/prevalence of sight-threatening diabetic retinopathy in relation to screening frequency. The included studies comprised 15 evaluations of real-world screening programmes, three studies modelling the natural history of diabetic retinopathy and seven cost-effectiveness studies. In evaluations of diabetic retinopathy screening programmes, the appropriate screening interval ranged from one to four years, in people with no retinopathy at baseline. Despite study heterogeneity, the overall tendency observed in these programmes was that 2-year screening intervals among people with no diabetic retinopathy at diagnosis were not associated with high incidence of sight-threatening diabetic retinopathy. The modelling studies (non-economic and economic) assessed a range of screening intervals (1–5 years). The aggregated evidence from both the natural history and cost-effectiveness models favors a screening interval >1 year, but ≤2 years. Such an interval would be appropriate, safe and cost-effective for people with no diabetic retinopathy at diagnosis, while screening intervals ≤1 year would be preferable for people with pre-existing diabetic retinopathy. A 2-year screening interval for people with no sight threatening diabetic retinopathy at diagnosis may be safely adopted. For patients with pre-existing diabetic retinopathy, a shorter interval ≤1 year is warranted. Diabet. Med. 30, 1272–1292 (2013) Introduction Diabetic retinopathy commonly complicates diabetes mell- itus [1] and meets the World Health Organization (WHO) criteria of suitability for screening [2]. It is a major cause of vision loss worldwide. Approximately one third of people with diabetes have diabetic retinopathy, and a third of those with diabetic retinopathy may have sight-threat- ening diabetic retinopathy, defined as clinically significant proliferative retinopathy or macula oedema [1]. The prevalence of diabetic retinopathy is projected to increase in the coming decades. The number of Americans aged 40 years or older, for example, with diabetic retinopathy and sight-threatening diabetic retinopathy is predicted to triple by 2050 [3]. In China, the prevalence of diabetic retinopathy among people with diabetes reaches 43% [1], with up to 9.2 million people in rural areas having diabetic retinopathy, including 1.3 million with sight-threatening diabetic retinopathy [1]. The natural history of diabetic retinopathy is relatively well understood, with recognizable stages. Major risk factors for developing diabetic retinopathy include duration of diabetes [4,5], severity of hyperglycaemia [6–8], hyperten- sion [9] and dyslipidaemia [10]. Once sight-threatening diabetic retinopathy is present, the progression is rapid and complications are unpredictable. Twenty years after diagno- sis, almost all people with Type 1 diabetes mellitus and 60% of people with Type 2 diabetes mellitus will have some degree of diabetic retinopathy [4,5]. There are precise, safe and accepted screening tests (oph- thalmoscopy and fundus photography) for diabetic retinopa- thy [11]. Glycaemic and blood pressure control may prevent the progression of diabetic retinopathy [7,9]. Appropriately timed laser photocoagulation therapy and, to a certain extent, Correspondence to: Justin B. Echouffo-Tcheugui. E-mail: [email protected]1272 ª 2013 The Authors. Diabetic Medicine ª 2013 Diabetes UK DIABETICMedicine DOI: 10.1111/dme.12274
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Systematic review or Meta-analysis
Screening intervals for diabetic retinopathy and
incidence of visual loss: a systematic review
J. B. Echouffo–Tcheugui1, M. K. Ali1, G. Roglic2, R. A. Hayward3,4,5 and K. M. Narayan1
1Hubert Department of Global Health, Rollins School of Public Health, Emory University, Atlanta, GA, USA, 2Department of Chronic Diseases and Health
Promotion, World Health Organization, Geneva, Switzerland, 3Department of Internal Medicine, 4Department of Health Management and Policy, University of
Michigan, and 5VA Center for Practice Management and Outcomes Research at the University of Michigan, Ann Arbor, MI, USA
Accepted 18 June 2013
Abstract
Screening for diabetic retinopathy can help to prevent this complication, but evidence regarding frequency of screening is
uncertain. This paper systematically reviews the published literature on the relationship between screening intervals for
diabetic retinopathy and the incidence of visual loss. The PubMed and EMBASE databases were searched until
December 2012. Twenty five studies fulfilled the inclusion criteria, as these assessed the incidence/prevalence of
sight-threatening diabetic retinopathy in relation to screening frequency. The included studies comprised 15 evaluations
of real-world screening programmes, three studies modelling the natural history of diabetic retinopathy and seven
cost-effectiveness studies. In evaluations of diabetic retinopathy screening programmes, the appropriate screening
interval ranged from one to four years, in people with no retinopathy at baseline. Despite study heterogeneity, the overall
tendency observed in these programmes was that 2-year screening intervals among people with no diabetic retinopathy
at diagnosis were not associated with high incidence of sight-threatening diabetic retinopathy. The modelling studies
(non-economic and economic) assessed a range of screening intervals (1–5 years). The aggregated evidence from both the
natural history and cost-effectiveness models favors a screening interval >1 year, but ≤2 years. Such an interval would be
appropriate, safe and cost-effective for people with no diabetic retinopathy at diagnosis, while screening intervals
≤1 year would be preferable for people with pre-existing diabetic retinopathy. A 2-year screening interval for people
with no sight threatening diabetic retinopathy at diagnosis may be safely adopted. For patients with pre-existing diabetic
retinopathy, a shorter interval ≤1 year is warranted.
of sight-threatening diabetic retinopathy or blindness, and
measures of efficiency of the screening interval. Disagree-
ments were resolved by a third reviewer (KMN). To our
knowledge, there is no commonly agreed-upon unifying
framework to evaluate screening programmes and/or studies
of the natural history of diabetic retinopathy; we therefore
focus on the individual characteristics of each study, giving
more credit to well-designed, large, prospective studies with
appropriate measures of outcomes. Although not originally
designed for use in review articles, the Drummond and
Jefferson evaluation scheme [21] for evaluating the quality
of economic studies appeared to be a consensual tool
that has been previously used [22]; we therefore used it for
economic studies (see also Supporting Information,
Appendix S2).
Results
Of the 25 studies included in this review (Fig. 1), 15 could be
characterized as evaluations of actual screening programmes
[17,19,23–34], three as studies modelling the natural history
of the disease [35–37], seven were economic modelling
studies that explored screening interval (five cost-effective-
ness studies [38–42], one a cost-utility study [18] and one
combined cost-effectiveness and cost-utility analyses [43]).
Screening studies
Screening programme evaluations examined the relationship
between sight-threatening diabetic retinopathy occurrence
and the frequency of screening (Table 1), either as a primary
[17,27–30,32] or a secondary objective [19,23–26,31,33,
34,44]. None of these studies were conducted in regions
other than the USA, Europe and Australia. Four studies were
hospital-based [23,24,28,30] and the remainder were popu-
lation-based. Their sample size varied from 185 to 57 199.
Six of these studies exclusively recruited Caucasians [19, 23–
26, 32] and seven included non-white participants [17,27,
29,31,33,34,44], but had a majority of Caucasians. In two
studies, the ethnicity of participants was not clearly reported,
but given the setting of these studies it was logical to infer
that the vast majority of their participants were Caucasians
[28,30]. When clearly reported, the age of participants
ranged from 15 to 99 years. One study focused on children
and adolescents exclusively [28]. The average duration of
ª 2013 The Authors.Diabetic Medicine ª 2013 Diabetes UK 1273
Systematic review or Meta-analysis DIABETICMedicine
diabetes at first screening varied between 0 and 15 years.
Thirteen of the 15 studies were retrospective cohorts and two
were prospective cohorts. One study reported using ophthal-
moscopy alone to ascertain diabetic retinopathy [23], fundal
photography alone was used in nine studies [25,27–30,32–
34,44], and a combination of ophthalmoscopy and fundal
photography was used in four studies [19,24,26,31].
The vast majority of the 15 screening studies addressed
screening for diabetic retinopathy in Type 1 and Type 2
diabetes together [19,23–26,29–31,44]. The screening stud-
ies provide unique information based on actual risk among
screened individuals, the majority of whom were not
receiving ophthalmic care. The appropriate screening inter-
val was variable, ranging from 1 year [44] to 4 years [30] in
people with no diabetic retinopathy at baseline. Despite the
between-studies variation, the overall tendency observed was
that a screening interval > 1 year would be appropriate and
safe in people with no diabetic retinopathy at diagnosis,
based on the extremely low rate of patients advancing from
no diabetic retinopathy to sight-threatening diabetic retinop-
athy in less than 2–3 years. Twelve of the 15 studies
supported a diabetic retinopathy screening interval > 1 year
[17,19,24,26–29,31–34,44]. The reported screening compli-
ance rate varied from 21% [27] to 28% [26].
A single study assessed the appropriate surveillance
intervals for those with diabetic retinopathy at diagnosis,
showing that a 1-year screening interval in the case of
background retinopathy and 0.3 of a year for mild prolif-
erative diabetic retinopathy, respectively, would be associ-
ated with a 95% probability of remaining free of
sight-threatening diabetic retinopathy for patients with
Type 2 diabetes [17]. Corresponding figures for Type 1
diabetes were 1.3 of a year for background and 0.4 of a year
for pre-proliferative diabetic retinopathy [27]. These were
the only studies that assessed the appropriate surveillance
intervals for those with diabetic retinopathy at diagnosis
[17,27]. Their findings are consistent with current consensus
in the medical community that yearly or more frequent
screening for people with any sign of diabetic retinopathy
should be the norm [12].
Some of the screening studies supporting an interval
> 1 year (e.g. a 2-year interval) included large-sample-size,
population-based cohorts with extended follow-up and/or
were specifically designed to assess the relationship between
screening interval and incidence of sight-threatening diabetic
retinopathy or blindness [17,27,29,33,34], thus offering
more robust evidence on the frequency of screening for
diabetic retinopathy. However, other studies putting forward
4886 citations identified essentially through PubMed (n = 1905) and EMBASE (n = 2981) and screened for identification of potentially relevant studies
4819 articles excluded—960 duplicates and 3859 articles on the basis of the titles and/or abstracts not fulfilling the inclusion criteria
67 studies retrieved for a more detailed evaluation
25 studies included in the review
15 screening studies
3 studied modelling natural history
7 economic modelling studies
42 studies excluded as, on close reading of the full text, these did not provide clear information on the impact of screening intervals on health outcomes (incidence/prevalence of retinopathy) or costs
FIGURE 1 Flow of selection of studies for inclusion.
1274ª 2013 The Authors.
Diabetic Medicine ª 2013 Diabetes UK
DIABETICMedicine Screening interval for diabetic retinopathy � J. B. Echouffo–Tcheugui et al.
Table
1Screeningstudiesassessingtheappropriate
intervalofscreeningfordiabetic
retinopathy
Author
andyear
publication
Studydesign/
period(length
offollow-up)
Sample
size/
country
(ethnicity)
Ageat
diagnosis
orentryin
thescreening
programme
Diabetes
type
Setting
Screening
modality/
coverage
Screening
frequency
Averagedisease
durationatthe
firstscreening
round
Incidence/
prevalence
ofsight-
threatening
diabetic
retinopathy
Assessm
ent
ofscreening
interval
Authors’
conclusion
aboutscreening
interval
Agardh
etal.,
1993[23]
Retrospective
cohort/
5years
858/Sweden
(Caucasian)
Meanage34.9
years
forpeople
with
Type1diabetes/
53.8
years
for
people
withType2
diabetes
Type1
diabetes
(n=431)
and
Type2
diabetes
(n=367)
Hospital-
based
Biomicroscopic
indirect
ophthalm
oscopy
withmagnifying
lensandaslit
lampthrough
dilatedpupils/by
ophthalm
ologist
1–2
years
ifno
diabetic
retinopathyor
minim
al
diabetic
retinopathy
19.8
years
for
people
with
Type1diabetes/
9.0
years
for
people
with
Type2diabetes
5-yearincidence
ofblindnessin
Type1
diabetes:0.5%
and0.6%
in
Type2
diabetes
5-yearincidence
ofmoderate
visual
impairment
(macular
oedem
aor
proliferative
diabetic
retinopathy):
1.2%
inType1
diabetes
and
1.7%
inType2
diabetes
Noform
al
testing/
implicit
Screening
interval
1to
2years
is
appropriate
Kristinsson
etal.,
1995[24]
Retrospective
cohort/
2years
185/Iceland
(Caucasian)
≥15years
Type1
diabetes
(n=87)
and
Type2
diabetes
(n=119)
Hospital-
based
Dilated
biomicroscopic
fundoscopyand
fundal
photography/by
ophthalm
ologist
Yearly
Notreported
2-yearincidence
ofsight-
threatening
diabetic
retinopathy
from
no
retinopathy
was0%
2-yearincidence
ofany
retinopathyfrom
noretinopathy
was23%
in
patients
with
Type1diabetes
and16%
in
those
with
Type2diabetes
Noform
al
testing/
implicit
2-yearlyscreening
forthose
with
Type1diabetes
andType2
diabetes
without
retinopathy
atdiagnosis
issafe.
Henricsson
etal.,
1996[25]
Retrospective
cohort/
2.9
years
1769/Sweden
(Caucasian)
30–6
0years
Type1
diabetes
(n=370)
and
Population-
based
Colourfundal
photography
(covered
fields
1–3
ofthe7
Yearlyforpeople
diagnosed
between20and
30years
ofage
18.7
years
for
those
aged
<30years
and
8.3
years
for
Incidence
of
blindness:
1.0
per
1000
person-years
Noform
al
assessm
ent
ofthe
relationship
Suggestionthata
1-yearscreening
intervalis
effective
ª 2013 The Authors.Diabetic Medicine ª 2013 Diabetes UK 1275
Systematic review or Meta-analysis DIABETICMedicine
Table
1(C
ontinued
)
Author
andyear
publication
Studydesign/
period(length
offollow-up)
Sample
size/
country
(ethnicity)
Ageat
diagnosis
orentryin
thescreening
programme
Diabetes
type
Setting
Screening
modality/
coverage
Screening
frequency
Averagedisease
durationatthe
firstscreening
round
Incidence/
prevalence
ofsight-
threatening
diabetic
retinopathy
Assessm
ent
ofscreening
interval
Authors’
conclusion
aboutscreening
interval
Type2
diabetes
(n=1399)
standard
fields,
withstereopairs
ofthemacula
(field2)in
eyes
withoutdiabetic
retinopathy;if
retinopathy,at
least
two
photographswere
added
of
standard
fields
4–7
)
andafter
5years
of
diabetes
duration
2-yearlyif
diabetes
diagn
osedat
≥30years
of
ageanduntil
approxim
ately
75years
ofage
those
aged
≥30years
(95%
CI0.4–2
.1),
incidence
of
visual
impairmentof
4.6
per
1000
person-years
(95%
CI3.0–6
.6)
with
screening
interval/
implicit
Linget
al.,
2002[31]
Retrospective
cohort/
6years
775/England
(mostly
Caucasian)
15–9
9years
Type1
diabetes
(n=104)
and
Type2
diabetes
(n=671)
Population-
based
(primary
care)
Non-m
ydriatic
fundussingle-field
photographyand
indirect
ophthalm
oscopy
2-yearly
13
2-yearincidence
ofnon-
proliferative
diabetic
retinopathy
2.20%
atround
2ofscreening
and2.25at
round3
2-yearincidence
ofclinically
significant
maculopathy
4.79%
atround
2ofscreening
and5.18%
at
round3
Implicit
assessm
ent
screening
interval:
incidence
ofsight-
threatening
diabetic
retinopathy
appears
to
bestable
over
rounds
ofscreening
ina2-yearly
strategy
Noreasonto
believethata
2-yearscreening
intervalwould
bedetrimental
Hansson-
Lundblad
etal.,
2002[26]
Retrospective
cohort/
8years
264/Sweden
(Caucasian)
≥30years
Type1
diabetes
(n=39)
and
Type2
diabetes
(n=225)
Population-
based
(primary
care)
Mydriaric3-field
fundal
photographyor
biomicroscopy
1–2
years
ifno
ormilddiabetic
retinopathy
Notreported
8-yearrate
of
blindnessin
Type1diabetes
was0%
and
2%
forType2
diabetes
Noform
al
testing/
implicit
A2yearscreening
regim
enmay
beappropriate
Younis
etal.,
2003[27]
Retrospective
cohort/
6years
of
follow-up
501/England
(96.2%
Caucasian)
<30years
Type1
diabetes
Population-
based
(primary
care)
Non-stereoscopic
3-fieldmydriatic
photography(and
modified
Wisconsin
grading)
Yearlyfor
patients
withnon-sight-
threatening
diabetic
retinopathy(no
retinopathyor
3.0
Cumulative
incidence
of
sight-threatening
diabetic
retinopathyin
patients
without
Fora95%
likelihoodof
remaining
free
ofsight-
threatening
diabetic
Screeningat2-to
3-yearintervals,
rather
than
annually,
for
patients
without
1276ª 2013 The Authors.
Diabetic Medicine ª 2013 Diabetes UK
DIABETICMedicine Screening interval for diabetic retinopathy � J. B. Echouffo–Tcheugui et al.
Table
1(C
ontinued
)
Author
andyear
publication
Studydesign/
period(length
offollow-up)
Sample
size/
country
(ethnicity)
Ageat
diagnosis
orentryin
thescreening
programme
Diabetes
type
Setting
Screening
modality/
coverage
Screening
frequency
Averagedisease
durationatthe
firstscreening
round
Incidence/
prevalence
ofsight-
threatening
diabetic
retinopathy
Assessm
ent
ofscreening
interval
Authors’
conclusion
aboutscreening
interval
background
retinopathy)
baseline
retinopathywas
0.3%
(95%
CI0.0–0
.9)at1
yearand3.9%
(1.4–5
.4)at
5years
Ratesof
progressionto
sight-threatening
diabetic
retinopathyin
patients
with
backgroundand
mildpre-
proliferative
diabetic
retinopathyat
1yearwere
3.6%
(0.5–6
.6)
and13.5%
(4.2–2
2.7),
respectively
retinopathy,
mean
screening
intervalsby
baseline
statuswere:
no
retinopathy
5.7
(95%
CI3.5–7
.6)
years;
background
retinopathy
1.3
(95%
CI0.4–2
.0)
years;and
mildpre-
proliferative
diabetic
retinopathy
0.4
(95%
CI0–0
.8)
years
retinopathyat
diagnosis
Younis
etal.,
2003[17]
Retrospective
cohort/
3.5
years
of
follow-up
4770patients/
England
(vast
majority
Caucasian)
>30years
Type2
diabetes
Population-
based
(primary
care)
Non-stereoscopic
mydriatic3-field
photography/
graded
bytrained
graders
Yearlyforpatients
withnon-sight-
threatening
diabetic
retinopathy(no
diabetic
retinopathyor
background
retinopathy)
7.8
Yearlyincidence
sight-threatening
diabetic
retinopathyin
patients
without
retinopathyat
baselinewas
0.3%
(95%
CI0.1–0
.5)in
thefirstyear,
risingto
1.8%
(95%
CI1.2–2
.5)in
thefifthyear;
cumulative
5-yearincidence
was3.9%
(95%
CI2.8–5
.0)
Fora95%
probability
ofremaining
free
ofsight-
threatening
diabetic
retinopathy,
mean
screening
intervalsby
baselinestatus
were:
no
retinopathy
5.4
years
(95%
CI4.7–6
.3),
background
1.0
year(95%
CI0.7–1
.3)
A3-yearscreening
intervalcould
be
safely
adopted
forpatients
withno
retinopathy
ª 2013 The Authors.Diabetic Medicine ª 2013 Diabetes UK 1277
Systematic review or Meta-analysis DIABETICMedicine
Table
1(C
ontinued
)
Author
andyear
publication
Studydesign/
period(length
offollow-up)
Sample
size/
country
(ethnicity)
Ageat
diagnosis
orentryin
thescreening
programme
Diabetes
type
Setting
Screening
modality/
coverage
Screening
frequency
Averagedisease
durationatthe
firstscreening
round
Incidence/
prevalence
ofsight-
threatening
diabetic
retinopathy
Assessm
ent
ofscreening
interval
Authors’
conclusion
aboutscreening
interval
andmildpre-
proliferative
diabetic
retinopathy
0.3
years
(95%
CI0.2–0
.5)
Maguire
etal.,
2005[28]
Retrospective
cohort/
12years
668/Australia
(notreported)
Childrenand
adolescents
Type1
diabetes
Hospital-
based
Seven-field
stereoscopic
mydriaticfundal
photography/
graded
byan
ophthalm
ologist
Yearly
Notreported
Notreported
Significant
increase
in
retinopathy
after
2years
from
thefirst
eye
examination
(P=0.03)in
theagegroup
>11years,
butnotuntil
6years
(P=0.01)
inthe
agegroup
<11years
Norisk
ofmissing
clinically
significant,vision-
threateningor
treatm
ent-
requiring
retinopathyby
extendingthe
screeninginterval
to2years
Olafsdottir
etal.,
2007[19]
Retrospective
cohort/
10years
296/Caucasian
16–9
0years
Type1
diabetes
(n=97),
Type2
diabetes
(n=199)
Population-
based
Dilatedslit-lamp
ophthalm
oscopy
andcolour
fundusphoto
graphy/byan
ophthalm
ologist
using.worsteyeto
defineretinopathy
level
2-yearlyifno
retinopathyand
yearlyonce
developmentof
retinopathy
18
Nopatientwent
from
no
retinopathyto
sight-threatening
retinopathyin
less
than2years
Noform
al
testing/implicit
Every2years
Misra
etal.,
2009[29]
Retrospective
cohort/
17years
20788/UK
(mainly
Caucasian)
Meanage:
68.8
years
Type1
diabetes
(n=205)
and
Type2
diabetes
(n=20583)Population-
based
Mydriatic2-field
fundusphoto
graphy/graded
bya
diabetologistandan
opththalm
ologist
17–1
9months
6.8
91%
decrease
in
theprevalence
ofsight-
threatening
diabetic
retinopathyfrom
thefirstroundof
screening(1.7%)
tothelast
round
Comparedwith
intervalsof
12–1
8months,
intervalsof
19–2
4months
werenot
associatedwith
ahigher
risk
of
either
referable
orsight-
Screening
intervals
of18months–
2years
canbe
safe
forpatients
atlow
risk
1278ª 2013 The Authors.
Diabetic Medicine ª 2013 Diabetes UK
DIABETICMedicine Screening interval for diabetic retinopathy � J. B. Echouffo–Tcheugui et al.
Table
1(C
ontinued
)
Author
andyear
publication
Studydesign/
period(length
offollow-up)
Sample
size/
country
(ethnicity)
Ageat
diagnosis
orentryin
thescreening
programme
Diabetes
type
Setting
Screening
modality/
coverage
Screening
frequency
Averagedisease
durationatthe
firstscreening
round
Incidence/
prevalence
ofsight-
threatening
diabetic
retinopathy
Assessm
ent
ofscreening
interval
Authors’
conclusion
aboutscreening
interval
17years
later
(0.16%)
threatening
diabetic
retinopathy
(odds
ratio0.93,
95%
CI0.82–1
.05),
butintervalof
more
than
24months
associatedwith
anincrease
risk
ofdiabetic
retinopathy
(oddsratio
1.56,95%
CI1.14–1
.75)
Arunet
al.,
2009[44]
Retrospective
cohort
study/
5years
6430/England
(mainly
Caucasian)
16–6
4years
Type1
diabetes
(n=5)
and
Type2
diabetes
(n=15)
Population-
based/
relied
ona
blindness
register
Mydriaticfundus
photography/
bytrained
screeners
Assumed
tobe
1year
Notreported
Annualincidence
ofblindness:
0.22
per
1000,andof
partialsightedness
0.43per
1000
Noform
al
testingof
relationto
screening
interval/im
plicit
Soto-Pedre
etal.,
2009[30]
Retrospective
cohort/
6years
430/Spain
(mainly
Caucasian)
Meanage
52.7
years
Type1
diabetes
(n=320)
and
Type2
diabetes
(n=110)
Hospital-
based
(tertiary
care)
45ºnon-m
ydriatic
single-fieldfundus
photography
(centred
onthe
macula
andone
photo
foreach
eye)
Notreported
10.1
years
for
those
withno
diabetic
retinopathyat
firstscreen
and
14.9
years
for
those
with
diabetic
retinopathyat
firstscreening
Ifnodiabetic
retinopathyat
baseline,
the
probabilityof
remainingfree
of
sight-threatening
diabetic
retinopathywas
97%
(95%
CI94–9
9%)at
theendofthe
fourthyear.Ifmild
non-proliferative
diabetic
retinopathy
atbaseline,
the
probabilityof
remainingfree
of
Intervalforat
least
a95%
estimated
probabilityof
remainingfree
ofsight-
threatening
diabetic
retinopathy
fornobaseline
diabetic
retinopathywas
4years
for
those
whohad
Type1diabetes
and3years
for
those
whohad
Screeningata
3–4
yearinterval
forpatients
with
diabetes
patients
ofdiabetic
retinopathyissafe
because
oftheir
low
risk
of
developingsight-
threateningdiabetic
retinopathy.
Patients
withmild
non-proliferative
diabetic
retinopathy
requirescreening
ata1-yearinterval,
orata2-year
ª 2013 The Authors.Diabetic Medicine ª 2013 Diabetes UK 1279
Systematic review or Meta-analysis DIABETICMedicine
Table
1(C
ontinued
)
Author
andyear
publication
Studydesign/
period(length
offollow-up)
Sample
size/
country
(ethnicity)
Ageat
diagnosis
orentryin
thescreening
programme
Diabetes
type
Setting
Screening
modality/
coverage
Screening
frequency
Averagedisease
durationatthe
firstscreening
round
Incidence/
prevalence
ofsight-
threatening
diabetic
retinopathy
Assessm
ent
ofscreening
interval
Authors’
conclusion
aboutscreening
interval
sight-threatening
diabetic
retinopathy
was99%
(95%
CI95–1
00%)atthe
endofthefirstyear
offollow-upand
94%
(95%
CI88–9
7%)atthe
endofthesecond
year
Type2diabe-
tes. Forpatients
withmildnon-
proliferative
diabetic
retinopathyat
baseline,
the
intervalforat
least
a95%
probabilityof
sight-
threatening
diabetic
retinopathy-
free
survival
was1year,or
inthose
witha
level
ofglycated
haem
oglobin
≤7.5%.
intervalwithgood
metaboliccontrol
Agardh
and
Tababat-
Khani,
2011[32]
Prospective
cohort
1322/Sweden
(Caucasian)
Meanage
55years
(SD12)
Type2
diabetes
Population-
based
Fundal
photography
(images—1central
and1nasalfields)
3-yearscreening
intervalifno
diabetic
retinopathyat
baseline
6years
After
3years,73%
hadnodiabetic
retinopathy,29%
hadmilddiabetic
retinopathy,0%
hadseverenon-
proliferative
diabetic
retinopathy
73%
ofpatients
without
retinopathyand
28%
withmild
ormoderate
diabetic
retinopathy
after
3years
3-yearscreening
intervalsissafe
insubjectswith
mildType2
diabetes
andno
diabetic
retinopathy
Jones
etal.,
2012[33]
Prospective
cohort
20686/
England
(mainly
Caucasian)
Median66.7
(interquartile
range
58.0–7
4.5)forthose
withoutretinopathy;
68.0
(interquartile
range58.5–7
5.7)for
those
withnon-
proliferativediabetic
retinopathy,66.3
(interquartile
range
55.7–6
6.3)forthose
withproliferative
diabetic
retinopathy
Type2
diabetes
Population-
based
Fundal
photography
(twoim
ages
of
each
eye)
Assumed
tobe
1-yearinterval
(butvariable
interval,upto
24monthsafter
baseline
examination
Notreported
(range
0–1
0years)
5-yearincidence
amongpatients
withoutdiabetic
retinopathyat
baselinewas4.0%
forpre-proliferative
diabetic
retinopathy,
0.59%
forsight-
threatening
maculopathy,0.68%
forproliferative
diabetic
retinopathy;
therespective
Noform
al
testingofthe
relationship
between
screening
intervaland
incidence
of
variousstages
ofdiabetic
retinopathy/
implicit
Few
patients
withoutdiabetic
retinopathyatfirst
screen
developed
pre-proliferative
diabetic
retinopathy,
proliferative
diabetic
retinopathyor
sight-threatening
maculopathy
after
5–1
0years
1280ª 2013 The Authors.
Diabetic Medicine ª 2013 Diabetes UK
DIABETICMedicine Screening interval for diabetic retinopathy � J. B. Echouffo–Tcheugui et al.
Table
1(C
ontinued
)
Author
andyear
publication
Studydesign/
period(length
offollow-up)
Sample
size/
country
(ethnicity)
Ageat
diagnosis
orentryin
thescreening
programme
Diabetes
type
Setting
Screening
modality/
coverage
Screening
frequency
Averagedisease
durationatthe
firstscreening
round
Incidence/
prevalence
ofsight-
threatening
diabetic
retinopathy
Assessm
ent
ofscreening
interval
Authors’
conclusion
aboutscreening
interval
10-yearincidences
were16.4,1.2
and
1.5%,respectively.
Amongthose
with
non-proliferative
diabetic
retinopathy
atbaseline,
after
1year23%
developed
pre-
proliferativediabetic
retinopathy,5.2%
developed
maculopathyand
6.1%
developed
proliferativediabetic
retinopathy;the
respective
10-year
incidenceswere53,
9.6
and11%,
respectively
offollow-up.
Screeningintervals
longer
than1year
maybeappropriate
forpeople
without
diabetic
retinopathy
atdiagn
osis
Thomas
etal.,
2012[34]
Retrospective
cohort
57199(w
ith
noevidence
ofdiabetic
retinopathy
atbaseline)/
Wales
(mainly
Caucasian
)
≥30years
Type2
diabetes
Population-
based
Fundal
photography
≥1year
3.9
years
for
those
withno
retinopathy
and5.1
years
amongthose
withdiabetic
retinopathy
4-yearincidence
of
anyandreferable
diabetic
retinopathy:
360.27and11.64
per
1000,respectively.
Forthose
oninsulin
treatm
entandwith
diabetes
duration
≥10years,incidence
ofdiabetic
retinopathyat1and
4years
was9.61
and30.99per
1000,
respectively
Noform
al
testingofthe
relationship
between
screening
intervaland
incidence
of
variousstages
ofdiabetic
retinopathy/
implicit
Extensionofthe
screeninginterval
forpeople
with
Type2diabetes
withoutdiabetic
retinopathy
beyond12
months,
withthe
possible
exception
ofthose
with
diabetes
duration
≥10years
and
oninsulin
ª 2013 The Authors.Diabetic Medicine ª 2013 Diabetes UK 1281
Systematic review or Meta-analysis DIABETICMedicine
similar recommendations were relatively small in size,
[19,23,24,26,28,30,31] or were hospital based [28,30].
Studies supporting annual screening were not always specif-
ically designed to examine the relationship between less
frequent screening intervals and incidence of sight-
threatening diabetic retinopathy or blindness [25,44]. Fur-
thermore, the largest of these studies used the proportion of
blindness attributable to diabetic retinopathy as the main
outcome and information about blindness was obtained from
a registry [44].
In brief, the vast majority of evaluations of real-world
diabetic retinopathy screening programmes supported a
screening interval > 1 year.
Modelling studies
Natural history models
Modelling studies of the natural history are shown in
Table 2. A brief summary of the key specificities of each of
these studies are presented below.
Using a hypothetical population and a range of sensitivities
and specificities, to compare annual or biennial screening
until background diabetic retinopathy develops and then
examination 6 monthly or more frequently, Davies et al.
found that biennial screening is a safe and efficient strategy,
provided that patients’ compliance and screening sensitivities
are both high [35]. The net benefit of reducing the screening
interval for those with no diabetic retinopathy from 2 years
to 1 year would range from 0.25–0.42 years of sight saved
per person, depending on screening methods used or the
screener (ophthalmologist, general practitioner or optome-
trist).
Two Taiwan-based studies used data from real-world
screening programmes to derive the appropriate screening
interval for diabetic retinopathy [36,37]. None of these
models included the pathway to blindness through macul-
opathy.
Tung et al. advocated annual screening on the basis of the
incidence of blindness reduction for various screening regi-