Can deep phenotyping using retinal images predict response to intravitreal aflibercept therapy in patients with neovascular age-related macular degeneration?
- Conditions
- Eye DiseasesWet age-related macular degeneration
- Registration Number
- ISRCTN28276860
- Lead Sponsor
- Moorfields Eye Hospital
- Brief Summary
2023 Results article in https://pubmed.ncbi.nlm.nih.gov/37109349/ (added 17/07/2023)
- Detailed Description
Not available
Recruitment & Eligibility
- Status
- Completed
- Sex
- All
- Target Recruitment
- 3000
Inclusion criteria for both retrospective and prospective parts:
1. Adults who are = 50 years and = 100 years
2. Treatment naïve neovascular AMD at baseline
3. Media clarity, pupillary dilation and patient cooperation for adequate imaging
4. Ability to give informed consent
Inclusion criteria for retrospective part only in addition to the above:
1. Have received 3 loading injections of intravitreal aflibercept therapy at monthly intervals as per standard care
2. Review up to 10 weeks after the 3rd loading dose with or without injection at this visit
3. Had Heidelberg OCT at least at baseline and after the loading phase but ideally 4 Heidelberg OCTs for the 4 visits
4. Heidelberg OCTA images if available for baseline and any visit thereafter (2nd, 3rd or 4th visit) provided there is a baseline OCTA (optional criteria)
1. Co-existent ocular disease: any other ocular condition that, in the opinion of the investigator, might affect or alter visual acuity during the course of the study
2. Any patient who has opted out of their information being used for research nationally or locally at any site
Study & Design
- Study Type
- Observational
- Study Design
- Not specified
- Primary Outcome Measures
Name Time Method Diagnostic accuracy of artificial intelligence over human graders in assessing the response of loading phase of intravitreal aflibercept injections for wet age-related macular degeneration
- Secondary Outcome Measures
Name Time Method 1. The analyses will be repeated excluding patients who appeared in the training set and the primary validation set2. Performance of the AI will be evaluated using higher-quality images with no media opacity (eg, cataracts) as noted by professional graders 3. AUC subgroups will be computed stratified by age and sex, smoking or medical history4. The analysis will be repeated by calculating the AUC, sensitivity, and specificity of the AI and the proportion of concordant and discordant eyes on the external validation datasets, compared with the reference standards