Version 2.1 (4th June 2009).
1. Study summary
Full title
Improving the value of screening for diabetic macular oedema using surrogate photographic markers.
Short title
Screening for diabetic MO using surrogate photographic markers.
Funding
£432,174 from the National Institute for Health Research (NIHR) Health Technology Assessment (HTA) programme. Grant reference 06/402/49.
Ethical approval
Main REC: North of Scotland Research Ethics Committee.
REC ref: 07/S0801/107.
Date of approval: 17/12/07.
Important dates
Start date: 1st May 2008.
50% target recruitment point: 1st May 2009.
100% target recruitment point: 31st October 2009.
End date: 30th April 2010.
Draft final report due: 14th May 2010.
Collaborators
Aberdeen University School of Medicine, NHS Grampian, NHS Tayside, NHS Lothian and Borders, Royal Liverpool and Broadgreen University Hospital NHS Trust, Queen Margaret Hospital, Dunfermline, Oxford Radcliffe Hospitals NHS Trust, Heart of England NHS Trust, Norfolk and Norwich University Hospitals NHS Trust.
Study objectives
The purpose of this study is to determine the best method for detecting sight-threatening macular oedema using photographic surrogate markers for people with diabetes in the context of national screening programmes.
Background
Macular oedema is associated with several conditions which cause irreversible vision loss, including diabetic retinopathy. It may be classified according to the fluid distribution: diffuse oedema is a general thickening of the central retina caused by either extensive capillary dilation or capillary closure, while focal oedema is centred on specific vascular abnormalities, such as microaneurysms. Accumulated fluid defocuses the image on the retina, reducing visual acuity. If oedema persists, the increased pressure may lead to irreparable photoreceptor damage or retinal detachment.
Since retinal thickening is not visible directly on the retinal photographs used by screening programmes, people are referred to ophthalmology clinics on the basis of a range of surrogate photographic markers, such as exudates within a certain distance of the foveal centre. Evidence from the Early Treatment of Diabetic Retinopathy Study suggests that exudates may be a sensitive marker of macular oedema (Bresnick 2000). However, as the ETDRS excluded patients with mild retinopathy in the absence of exudate, the results are not applicable to screening programmes in the United Kingdom, where 60% of patients have no visible signs of retinopathy and over 30% have only mild retinopathy. Evidence from the Grampian Retinal Screening Programme suggests that only 12% of patients with surrogate markers referred to an ophthalmologist have indications of macular oedema when examined by slit-lamp biomicroscopy. Similarly, a retrospective analysis from Liverpool, including 257 patients referred from the screening programme to the ophthalmology clinic between December 2001 and June 2002, found that only 14% had evidence of macular oedema (unpublished data).
Macular oedema has traditionally been assessed clinically using a combination of slit-lamp biomicroscopy, stereo photography and stereo fluorescein angiography. However, these techniques have a number of limitations. Foremost is that they are only qualitative assessments, which are relatively insensitive to thickness changes. Furthermore, slit-lamp examination does not provide a pictorial record and, together with stereo photography, is known to be biased by the presence or absence of exudates. Although the angiogram is a sensitive test for leakage, the assessment of thickening is very subjective. Best corrected visual acuity has also been used as a surrogate indication of thickening, but is neither sensitive nor specific, being affected by several factors besides macular thickness.
Study design
A total of 4000 patients with photographic signs of maculopathy (exudates within two disc diameter radius, blots or dot haemorrhages within one disc diameter radius) shall be recruited from the seven study centres. Each subject will have photography and optical coherence tomography on both eyes where possible. 10% of patients are expected to have ungradeable images and will be unsuitable for the study.
All relevant lesions visible in the colour photographs will be annotated by the research nurse in Aberdeen using computer-assisted software. The optical coherence tomography images will be analysed both quantitatively and qualitatively by the research nurse. Software will be developed to analyse the distribution of retinal lesions in order to find the significance of the photographic marker patterns on the likelihood of clinically significant macular oedema.
Proposed outcome measures
A sample of patients attending diabetic retinopathy-screening programmes will receive an optical coherence tomography examination in addition to the standard digital photograph. An expert grader will assess all digital images independently for the presence of different surrogate photographic markers. The sensitivity and specificity of standard referral criteria (based on the presence of surrogate markers) will be assessed using optical coherence tomography as the reference standard.
Although manual grading will always have a role in the assessment of retinal images, the scale of the current diabetes epidemic means that automation will have to play a key role if the Liverpool Declaration's target of offering annual systematic screening to at least 80% of the estimated 35 million people with diabetes in Europe is to be achieved. Furthermore, given the low specificity of current referral criteria, it is important to investigate whether specificity can be improved using more complex patterns of surrogate markers, using either computer-assisted annotation or fully automated computer analysis, so that scarce ophthalmology resources are used most effectively.
Derived sensitivity/specificity estimates will be used to assess the costs and consequences (i.e. the number of appropriate/inappropriate ophthalmology referrals) of using alternative surrogate markers, and patterns of surrogate markers, for the detection of macular oedema, using manual, computer-assisted annotation and fully automated detection systems.
It is important to look at the implications for health care delivery of introducing this technology into systematic screening programmes for diabetic retinopathy. We will therefore model long-term costs and outcomes (visual loss and quality adjusted life years) of the alternative screening strategies using epidemiological literature and available cost estimates.
Sample size calculation
If 33333 patients are screened across five centres, 4000 patients would be expected to have surrogate markers (12%) of whom 400 would be expected to have macular oedema. If there are 4000 patients with any markers, the power for detecting a change in the referral specificity of at least 20% will be much higher than 80%.
However, a new diagnostic test (using a more specific combination of features) which has increased specificity may also have decreased the sensitivity. Therefore, since there are fewer true cases with macula oedema than true controls, the crucial issue for the power of this study is whether or not a small change in sensitivity can be detected.
With 400 cases of macula oedema there will be 80% power to detect a difference in sensitivity of 3% (99% vs. 96%) between two diagnostic tests (any markers vs. a specific combination of markers) when the percentage of true cases where the diagnostic tests disagree is expected to be 5%. A McNemar's test of equality of paired proportions has been used with a 0.05 significance level.
Research governance
Research activities at each of the participating centres will be carried out in accordance with the Department of Health's Research Governance Framework for Health and Social Care. The project will be registered with the Institute of Applied Health Sciences (University of Aberdeen) and each appropriate NHS trust. Copies of the protocol and records of all important documents will be cross referenced and filed so that they are available at any given time. Records relating to all procedures carried out will be kept so that the research process is clearly understandable and repeatable. NHS Grampian has agreed to act as the sponsors for the study.
A Trial Steering Committee has been set up. The chair, Dr Caroline Styles, is a consultant ophthalmologist in Fife with experience of retinal screening (she was later replaced by Dr Rod Harvey). Professor Alex Elliot, a medical physicist from Glasgow, brings medical image processing experience. Mr Steve Graham is the committee patient representative. Ms Alison Farrow is a retinal photographer in Aberdeen (she was later replaced by Dr A Manivannan). Finally Dr John Olson and Professor Peter Sharp are also members.
2. Centre contact details
View in own window
| Centre | Principal investigator(s) | Other contacts |
|---|
| University of Aberdeen and NHS Grampian | Dr John Olson | Dr Keith Goatman |
| Heart of England NHS Trust | Professor Paul Dodson | Ms Jane Pitt |
| NHS Tayside | Professor Graham Leese
Dr John Ellis | |
| NHS Lothian | Dr Ken Swa | Dr Shyamanga Borooah |
| Royal Liverpool & Broadgreen University Hospitals Trust | Professor Simon Harding
Professor Deborah Broadbent | Dr Yalin Zheng |
| Oxford Radcliffe Hospitals NHS Trust | Professor Victor Chong | Ms Sue Beatty |
| Queen Margaret Hospital, Dunfermline | Dr Caroline Styles | |
Dr John Olson is the project Chief Investigator (ten.shn@noslo.nhoj). For general enquiries about the study contact Dr Keith Goatman (ku.ca.ndba@namtaog.a.k).
Dr Caroline Styles (ten.shn@selyts.enilorac) is the chair of the Trial Steering Committee (later replaced by Dr Rod Harvey).
3. Patient selection and recruitment
Each centre shall collect data from 800 patients fulfilling the inclusion criteria below.
Inclusion Criteria
Age 18 or older.
A trained grader confirms that the fundus photograph of at least one eye shows diabetic eye disease including any of:
Microaneurysms/dot haemorrhages within one disc diameter radius of the macula.
Blot haemorrhages within one disc diameter radius of the macula.
Exudates within two disc diameter radius of the macula.
Able and willing to provide signed informed consent.
Exclusion Criteria
The patient has had macular or pan-retinal laser treatment [clarification to protocol added version 2.0].
The patient has had an intraocular injection [added in protocol version 2.0].
The patient is pregnant (since oedema may be present independent of gestational diabetes) [Added in protocol version 2.0].
Intraocular surgery within one year of enrolment.
The screening fundus photograph is a technical failure (i.e. the macula vessels are not clearly visible and/or the field of view does not include a region of diameter two disc diameter radius s about the centre of the macula).
Contraindications to pupillary dilation/intolerance or hypersensitivity to mydriatics in either eye if dilation is necessitated.
How patients are recruited will depend on the local set-up at each centre. In Aberdeen patients with exudates or blots within 1 disc diameter radius (grade code M2) are routinely referred from the screening service to see an ophthalmologist. These patients are now routinely offered optical coherence tomography scans and therefore only need to be consented to allow their data to be used in the study. Patients with exudates within 1 disc diameter radius and 2 disc diameter radius (grade code M1) will need to be invited for an optical coherence tomography scan they would not normally receive, so will need to be consented both for the imaging and use of data. Finally, patients with microaneurysms/dot haemorrhages within 1 disc diameter radius do not trigger a specific grade in the Scottish grading system and will have to be manually chosen during screening grading. As for the M1 category, these patients will have to be invited for an optical coherence tomography they would not normally receive, so will need to be consented both for imaging and use of their data.
4. Retinal photography, visual acuity and Glitazone.
Visual acuity
Log-MAR (or Snellen if Log-MAR not available) pin-hole or best-corrected visual acuity measurement for both eyes. [Best-corrected added version 2.1]
Record if the patient has a long-standing, non-diabetic reason for a low visual acuity measurement, for instance due to amblyopia. This must be recorded on the web upload form. [Added version 2.1]
Photography
Single views of both eyes:
Approximately 45 degree field of view.
Approximately macula centred.
Colour digital photograph, at least 3 megapixels (ideally less than 7 megapixels and, if using JPEG compression, set for highest quality).
Adequate field of view (should show 2 disc diameter radius around centre of macula).
Adequate clarity (i.e. adequate to see macular microaneurysms, if present). If there are problems obtaining a retinal photograph it is very likely to be difficult obtaining an adequate optical coherence tomography scan.
Mydriasis only necessary if pupil size too small for imaging. Camera small pupil facility acceptable providing 2 disc diameter radius visible around macula centre.
Note that the maximum allowed time between photography and optical coherence tomography is four weeks to reduce errors due to lesion changes between photograph and optical coherence tomography.
Example images showing adequate and inadequate clarity and field of view are available from the website. (http://www.abdn.ac.uk/ismo/).
Glitazone [Added version 2.1]
The glitazone class of drugs appears to be associated with diabetic macular oedema [1]. If the patient has used any of the following drugs within the past six months this must be recorded on the web upload form:
Rosiglitazone: Avandia, Avandamet.
Pioglitazone: Actos Competact.
[1] Fong DS and Contreras R. Glitazone use associated with diabetic macular edema. American Journal of Ophthalmology 2009;147:583–586.
5. Optical Coherence Tomography
Optical coherence tomography scanner choice
When the project was first proposed every centre had use of the Zeiss Stratus optical coherence tomography scanner (often known as OCT3). Since the funding was granted a number of new instruments based on a spectral domain method have become available. Initial tests on the Zeiss Cirrus, Topcon OCT1000 and Heidelberg Spectralis show these new scanners have some advantages for this study. Hence centres will not be required to use the Stratus but instead encouraged to use the best available equipment.
In order to reduce variation due to different manufacturers, models and operators:
Only accredited operators shall acquire scans.
It is recommended that, where possible, the same scanner is used for the duration of the study. If more than one scanner is used the scanner used must be indicated. If a scanner must be replaced it may be replaced by any model.
Dates of services and faults should be recorded in the study folder.
Optical Coherence Tomography protocol
Two types of data are required from the optical coherence tomography scanner:
A thickness map divided into nine Early Treatment Diabetic Retinopathy Study regions centred on the centre of the macula and with a diameter of 6 mm.
Orthogonal B-mode 6-mm cross-sections centred on the centre of the macula.
Specific instructions for different optical coherence tomography models are given in appendix A to the protocol.
Accreditation process
As in other multicentre imaging studies, to avoid inter-centre imaging variation all optical coherence tomography operators are required to be accredited before submitting data for the study. Operators should submit a simple portfolio of images (described below) collected using the optical coherence tomography scanner intended for the study. Images will be checked for macula position, adequate image quality and differences between thicknesses in the nine Early Treatment Diabetic Retinopathy Study regions of the repeat scans. The accreditation images are as follows:
(a) Normal eyes
Repeat macula maps of the same normal eye as per scanner model protocol above.
(b) Macular oedema
Repeat macula maps of the same eye showing obvious macular oedema (i.e. central thickness at least 300 microns).
For scanners (such as the Stratus) which do not automatically generate high-resolution orthogonal cross-sections (“cross hair scans”) also perform the cross-hair scan.
Accreditation Upload procedure
Go to the study website: http://www.abdn.ac.uk/ismo/
Login using your supplied username and password. The admin functions will appear once you are logged in. Select “Accreditation” from the menu and follow the instructions for uploading images.
6. Data transfer
Photographs and optical coherence tomography scans are transferred to Aberdeen using the web-based upload service at:
http://www.abdn.ac.uk/ismo/
Log in using your supplied username and password and the admin functions will appear. Select “upload data” and follow the on-screen instructions.
Note: the server automatically generates an anonymous subject ID. If you enter the subject name and hospital ID number these will print a front sheet for your study folder, but will not be transmitted to Aberdeen.
Data protection and personal information
The personal patient information you enter is not transferred to the webserver, it is just used to produce the print out.
However, patient information may appear on the photographs or optical coherence tomography images which are transferred to Aberdeen. Although the University of Aberdeen webserver has been registered to allow the storage of personal information, personal information is not required for this study we do not wish to keep it, and have not sought ethical permission to do so. Therefore all personal information will be automatically removed from images that are uploaded to Aberdeen. It is therefore vital that if centres wish to be able to cross reference data from their own clinical records that they keep a local copy of the anonymous study ID and real patient ID.
Image grading
(a) Colour photographs
All fundus photographs will be assessed in Aberdeen by a trained reference grader. An experienced grader will train the reference grader in the identification of the features of diabetic retinopathy and in the use of the software. The trainer will quality assure the annotations of 100 images as part of the training phase. The following features will be noted:
Presence of microaneurysms/dot haemorrhages and blots within 1 disc diameter radius.
Presence of exudates within 1 disc diameter radius and 2 disc diameter radius.
All the above lesions will be annotated on images, using computer-assisted annotation software, to enable pattern analysis of the lesion distributions.
(b) Optical Coherence Tomography data
All the optical coherence tomography images will be graded quantitatively and qualitatively for area, amount and site of retinal thickening based on the generated thickness maps. Optical coherence tomography studies suggest that oedema may be reliably detected by slit-lamp biomicroscopy if it has a retinal thickness of at least 300 µm (Hee 1998). This agrees with the Early Treatment Diabetic Retinopathy Study, which used a reference thickness of 250 µm for “clinically significant” macular oedema.
In this study, macular oedema will be defined as an optical coherence tomography retinal thickness of 300 µm or greater (relative to Zeiss Stratus measurements) in any map region within one disc diameter radius of the centre of the fovea (i.e. in any of the five central Early Treatment Diabetic Retinopathy Study map regions).
