Dr Angelica Ly and Dr Matt Trinh summarise key findings in their recent macular disease research, and offer some key take home messages for clinicians.
As clinician-researchers committed to advancing care for patients with age-related macular degeneration (AMD), we’ve had the privilege of leading or co-authoring four recent studies that explore how eyecare professionals can better predict, manage, and support patients living with AMD.1-4 These works span clinical imaging, risk modelling, artificial intelligence (AI), and patient education – each contributing to a more holistic, evidence-based approach to AMD care (Figure 1).
1. Simplifying risk without sacrificing accuracy
Early detection of progression to late AMD is key to preserving vision and reducing the burden on patients and healthcare systems. The age-related eye disease study (AREDS) simplified severity scale is recommended to estimate a patient’s risk of AMD progression based on retinal features.5 Last year, the scale was updated to include reticular pseudodrusen into risk categorisation, which improves accuracy but also adds complexity.6 The updated model requires observation of up to nine features across both eyes.
In our recent publication, we proposed an abridged version of the updated simplified AREDS risk model.1 By focusing on just four key features we arrive at similar risk estimates. This streamlined approach essentially halves the number of observations and ~doubles efficiency, focusing on one eye of interest (Figure 2).
Key take home clinical message: To perform ‘simpler’ AMD risk prediction, firstly, assign one score per feature – large drusen, pigmentary abnormalities, or reticular pseudodrusen in the eye of interest, and late AMD in the fellow eye. The approximate risk of developing late AMD within three years increases based on the total score: Score 0 = 4%; Score 1 = 8%; Score 2 = 16%; Score 3 = 32%;
Score 4 = 64%.
2. High risk ≠ high accuracy: The OCT biomarker paradox
The identification of novel biomarkers has become central to the burgeoning field of precision medicine, offering eyecare professionals powerful tools to predict disease progression and tailor management. However, with more than 100 AMD biomarkers described to date, this can be overwhelming and impractical to apply in everyday practice.7 In this study, we evaluated the real-world performance of 10 key optical coherence tomography (OCT) biomarkers for predicting late AMD, including quantification of their prevalence, ease-of-grading, time-to-conversion, and overall accuracy.2 While some recently popular biomarkers, like nascent geographic atrophy and shallow irregular retinal pigment epithelium elevations, showed very high risk associations, they occurred infrequently or were difficult to grade reliably. Interestingly, pigmentary abnormalities detected via colour fundus photography was the single most useful biomarker for predicting late AMD, even more so than individual newer OCT biomarkers, likely owing to its relatively high prevalence and ease-of-grading. At least three additional OCT signs were required to significantly lift predictive accuracy, emphasising the need for automated, multimodal grading tools in future.
Key take home clinical message: Pigmentary abnormalities may be the strongest individual predictor of AMD progression, even amidst newer OCT biomarkers. But prepare for automated, multimodal tools in the future.
3. AI in AMD: Promise, caution, and practicality
The need for structured clinical risk scales discussed in the prior sections has arisen because AMD is often misdiagnosed, diagnosed or prognosticated too late, delaying treatment.8,9 AI-based tools using retinal imaging (like fundus photos or OCT) show near-specialist accuracy in detecting and prognosticating AMD, but are not yet widely used in practice.10 Barriers like cost, trust, and data privacy concerns are slowing their adoption, and the specific challenges for implementing these tools in real-world eyecare remain unclear.11
We recently conducted a qualitative study exploring multi-stakeholder perspectives on AI for AMD diagnosis in Australia.3 Clinicians, patients, developers, and healthcare leaders collectively supported AI as a clinical decision support tool – particularly for remote detection and monitoring of neovascular AMD. However, they also emphasised the need for:
• Human oversight and shared decision-making
• Equity across socioeconomic and cultural groups
• Transparency in algorithm performance and data use
• Integration into existing workflows and patient management systems
Stakeholders envisioned a fee-for-service model for AI-supported diagnosis, with potential to improve access in rural and underserved communities. Yet, concerns about trust, false positives, and the risk of over-reliance on AI highlight the importance of cautious, evidence-based implementation.
Key take home clinical message: AI tools using retinal imaging show near-specialist accuracy in AMD diagnosis and prognostication, but their safe and effective adoption in clinical practice depends on addressing key barriers like trust, equity, and integration into existing care workflows.
4. Education matters – especially early on
Finally, anti-VEGF injections, smoking cessation and nutritional supplements are key strategies to slow AMD progression; however, many patients don’t follow recommended care. Over half skip supplements, and more than a third miss injection appointments.12,13 Poor communication and recall of care advice contribute to low adherence, highlighting the need for clearer patient education and support.14
In our recently-conducted randomised controlled trial of 125 participants with AMD, we compared enhanced education, using take-home pamphlets, posters, and tailored messaging, to standard care.4 The enhanced education focused on simple calls to action e.g., the main message used on the poster about nutritional supplements was: ‘Save your vision. If you have intermediate AMD, take your eye supplements every day’.
While overall confidence in AMD-related care did not differ significantly between groups, patients diagnosed within the past five years showed a meaningful improvement in confidence when exposed to targeted educational materials. This suggests that education, within five years of diagnosis, can positively shape patient attitudes and potentially influence care-seeking behaviour.
Key take home clinical message: Timing matters in eye health education. While this form of enhanced education didn’t boost confidence across all patients – for patients diagnosed with AMD in the last five years, it made a clinically meaningful difference. This highlights the importance of tailoring education to the stage of diagnosis using simple calls to action.
Looking ahead
Together, these studies highlight the importance of integrating clinical imaging, simplified risk stratification, AI, and patient education into AMD care. Pigmentary abnormalities remain a cornerstone of AMD prognostication, but the future lies in automated, multimodal clinical decision support tools that enhance without diminishing clinician judgment. Meanwhile, empowering patients with timely, relevant information is essential for fostering engagement and adherence.
As AMD care evolves, so must our models and methods. The goal is not just to predict progression accurately, but to support patients holistically. With further validation and thoughtful implementation, these findings could help shape a future where AMD care is not only more accurate, but more accessible and compassionate.
References
1. Trinh M, Duong A, Cheung R et al. Proposal of a Simpler Eye-Level Risk Model Incorporating Reticular Pseudodrusen for the Clinical Prediction of Late Age-Related Macular Degeneration.n/a(n/a).
2. Trinh M, Cheung R, Nam J et al. High risk does not guarantee high accuracy—Evaluating the prognostic accuracy of OCT biomarkers for predicting late AMD.n/a(n/a).
3. Ly A, Herse S, Williams M-A et al. Artificial intelligence for age-related macular degeneration diagnosis in Australia: A Novel Qualitative Interview Study.n/a(n/a).
4. Wang E, Doig GS, Ly A. An enhanced educational intervention for improving confidence in the eye health benefits of appropriate care for age-related macular degeneration: a randomized controlled trial. Health Education Research. 2025;40(4).
5. Ferris FL, Davis MD, Clemons TE et al. A simplified severity scale for age-related macular degeneration: AREDS Report No. 18. Arch Ophthalmol. 2005;123(11):1570-1574.
6. Agrón E, Domalpally A, Chen Q et al. An Updated Simplified Severity Scale for Age-Related Macular Degeneration Incorporating Reticular Pseudodrusen: Age-Related Eye Disease Study Report Number 42. Ophthalmology. 2024;131(10):1164-1174.
7. Trinh M, Cheung R, Duong A et al. OCT Prognostic Biomarkers for Progression to Late Age-related Macular Degeneration: A Systematic Review and Meta-analysis. Ophthalmol Retina. 2024;8(6):553-565.
8. Neely DC, Bray KJ, Huisingh CE et al. Prevalence of Undiagnosed Age-Related Macular Degeneration in Primary Eye Care. JAMA Ophthalmol. 2017;135(6):570-575.
9. Ly A, Nivison-Smith L, Zangerl B et al. Advanced imaging for the diagnosis of age-related macular degeneration: a case vignettes study. Clin Exp Optom. 2018;101(2):243-254.
10. Dong L, He W, Zhang R et al. Artificial Intelligence for Screening of Multiple Retinal and Optic Nerve Diseases. JAMA Netw Open. 2022;5(5):e229960.
11. Tseng RMWW, Gunasekeran DV, Tan SSH et al. Considerations for Artificial Intelligence Real-World Implementation in Ophthalmology: Providers’ and Patients’ Perspectives. 2021;10(3):299-306.
12. Ng WT, Goggin M. Awareness of and compliance with recommended dietary supplement among age-related macular degeneration patients. Clin Exp Ophthalmol. 2006;34(1):9-14.
13. Kusenda P, Caprnda M, Gabrielova Z et al. Understanding Loss to Follow-Up in AMD Patients Receiving VEGF Inhibitor Therapy: Associated Factors and Underlying Reasons. Diagnostics (Basel). 2024;14(4).
14. Bott D, Huntjens B, Binns A. Nutritional and smoking advice recalled by patients attending a UK age-related macular degeneration clinic. J Public Health (Oxf). 2018;40(3):614-622.
