Improving the Value of Screening For Macular
Oedema usingSurrogate Photographic
Markers
Dr John Olson
NHS Grampian
Improving The Economic Value Of Photographic Screening For Optical Coherence Tomography
Detectable Macular Oedema – A Prospective Multicentre, United Kingdom Study
• Olson J, Sharp P , Goatman K, Prescott G, Scotland G, Fleming A, Philip S, Santiago C, Borooah S, Broadbent D, Chong V, Dodson P, Harding S, Leese G, Styles C, Swa K, Wharton H
• Health Technol Assess, Vol 17,, May- June 2013, In Press
A Success Story?
• Systematic screening programme for diabetic retinopathy
Missing the target?
• The health-economic case is based on the detection of people with, or at risk of– proliferative diabetic
retinopathy– before they develop
complications• Vitreous haemorrhage• Traction retinal detachment
• But 90% of referrals are for ? diabetic macular oedema
Why?
• Retinal photographs are not discriminatory for proliferative retinopathy or its precursors
• Other things may be present– e.g. diabetic “maculopathy”– We have to manage these findings
How did we get there?
• Retinopathy grades based on ETDRS
• Maculopathy grades basedon
…(GOBSAT)
Different management
New Vessels Oedema
Definitive treatment Indefinite treatment
Management independent of visual acuity
Management depended on visual acuity
2 D red structure 3 D transparent elevation
Few false +ves Many false +ves
What did ISMO question?
• Can we do it better?
• What will it cost?
• What will it mean?
The Answers In Short- Can Grading Schemes do it better ?
• Computer says nah
The Answers In Short- Can OCT do it better ?
• Yes• Increases the
specificity of referrals• With no loss of
sensitivity
The Answers-What will it cost?
• Less• If you use OCT • Whatever grading
strategy you use• Saves you money
The Answers- What will it mean?
Study Highlights
© 2008 Google-Imagery © 2008 TerraMetrics
Aberdeen
Dundee
Edinburgh
Liverpool
Birmingham
Oxford
Study centres
Aberdeen
Birmingham
Dundee
Edinburgh
Glasgow
Liverpool
Oxford
Glasgow
Every day practice
Aberdeen
Dundee
Edinburgh
Liverpool
Birmingham
Oxford
Glasgow
3450 Subjects
• Photographic signs of diabetic retinopathy– exudates ≤ 2DDr– blot haemorrhages ≤ 1DDr– dot haemorrhages/microaneurysms ≤ 1DDr
• Each subject had photography and optical coherence tomography on both eyes, where possible.
Patient Characteristics
• Median age 60
• 60.7% male
• 85.4% Caucasian
• 77.4% type 2 diabetes
370 Excluded (10.5%)
• 6 years older
• Female
• Asian/ Black
• Zeiss Stratus
• Topcon OCT 1000
Lesion Distribution
Expected % Recruited %
Ma/dot only 69.8 40.3
Blot no exudate 8.6 8.4
Exudate 21.6 20.4
No Ma/dot/blot/exudate ≤ 1DDr
28.1
Definition of Macular Oedema
• Central ETDRS region thickness > 250µm
• OR any of 5 inner regions > 300µm
• AND visible intraretinal cyst/ area of subretinal fluid
Prevalence of oedema
• 7.7% of study population
• Prevalence differed greatly by centre– 3.7% to 12.2%
• Prevalence differed greatly by scanner– 4.5% to 11.8%
Relationship to Centre
• Aberdeen 12.0%• Birmingham 3.7%• Dundee 12.2%• Edinburgh 6.4%• Liverpool 2.9%• Dunfermline 4.4%• Oxford 7.7%
All scanners are equal, but some scanners are more equal than others
• Zeiss Stratus– 4.5%
• Topcon OCT1000– 6.5%
• Heidelberg Spectralis– 8.7%
• Zeiss Cirrus– 11.8%
Relationship to patient features
• Older age– 68yrs cf 60
• Caucasian– 8.4% cf 3-4%
• Type 2 diabetes– 8.7% cf 3.9%
• Poorer vision– 5x more likely– If VA ≤ 6/9
• BUT NOT– Sex, glitazone, amblyopia
Relationship to Lesions
R Eye % L Eye %
No lesions
0.8 0.6
Ma/dot only
2.2 2.3
Blot no Exudate
10.2 11.2
Exudate 12.5 11.2
Other 1.1 1.1
Can we do any better?
• Three Grading Strategies Examined– Manual grading
• Presence/ absence of features• SDRGS 2007
– Computer-assisted manual annotation• All individual lesions ≤ 2DDr
– Fully automated annotation grading• Three versions
– Automated image analysis– +VA
– +VA + Age+ Type DM + Sex
Manual Grading (features)
Manual Grading (features)
• Scotland– 59.5% sensitivity– 79.0% specificity
• England– 72.6% sensitivity– 66.8% specificity
• England plus– 73.3% sensitivity– 70.9% specificity
Computer Assisted, Manual Annotation, Grading• Best for sensitivity &
specificity• Time-consuming
procedure • Unlikely to be
considered for routine screening practice
In Years To Come
Marvin the Manically Depressed Autograder
""I think you ought to I think you ought to know…. I'm feeling know…. I'm feeling very very depressed ......noboddepressed ......nobody likes mey likes me""
DRS in Scotland 2012
What will it cost?
• Cost per screen £33.13• Cost per OCT screen £31.96• Total cost for ?oedema £65.09
• Cost of attending ophthalmology £90.00
• (Cost of Slit lamp within DRS £27.29)
TABLE 30 Screening and referral cost per true case of macular oedema detected for 3,170 patients; Adjusted for expected frequency of different patient categories and based on Scottish screening and referral costs
* Reference strategy; a figures in table based on assumption that fully automated grading can be implemented at zero net increase in grading costs;++ Represents a cost saving per case missed relative to the reference strategy; d strategy more costly and less effective than an alternative strategy (dominated)
What does it mean?
• At present we spend £13,750,000 a year – 250,000 people @ £55– Screening + 1st visit to
ophthalmology– £2,337,500 on ? M2
• If we do nothing, other than introduce OCT into the screening pathway
• we save money
Should we grade differently?
Current Scottish Criteria + OCT is the most cost effective of all strategies
What if we do nothing? 20 year “M2” Markov Model• Only 5.6% of M2 at risk of visual loss• Repetitive nature of screening
– 12% of non-referred MO modelled to progress at 12 months cf 5% of referred (laser Rx)
• More sensitive strategies– More OCTs, more referrals
• Bilateral incidence 12%– QALY determined by VA in better seeing eye
• Additional cost per QALY going to strategy 16– £882,307 at 5 years– £353,927 at 20 years– (£20-30,00 UK threshold for “cost-effectiveness)
What should we do?
Cost-effectiveness acceptability curves for the alternative strategies based on a 20 year time horizon and using quality adjusted life years as the measure of effect
How should we manage M2s?Is this the answer?
• Photos graded as M2
• Check VA
• Do an OCT if VA 6/12 or worse?
• Otherwise rescreen in 6 months?
Thank You
Modelled visual acuity changes for “CSMO”