A study of the progression of sensitivity loss in GA over time using inferred data

(Image Credit: AdobeStock)

The recent OMEGA study¹ found that “inferred sensitivity” mapping may be a potential functional surrogate end point for geographic atrophy (GA) based on its ability to predict retinal function with high spatial resolution and without extensive psychophysical testing, according to Georg Ansari, MD, first author of the report published in Investigative Ophthalmology and Visual Science. He is from the Institute of Molecular and Clinical Ophthalmology Basel, and the Department of Ophthalmology, University Hospital Basel, both in Basel, Switzerland.

Currently, the treatments available for GA, ie, pegcetacoplan (Syfovre, Apellis Pharmaceuticals) and avacincaptad pegol (Izervay, Astellas Pharmaceuticals) in the US, slow the disease progression. However, no treatment with a prospectively established functional treatment benefit is currently available.^2-4

“The challenge of selecting appropriate functional outcome measures for GA in treatment trials remains significant,” the researchers said.

While the best-corrected visual acuity (BCVA), low-luminance VA, and dark adaptometry have been evaluated as functional metrics, they have failed “to effectively capture the progression of GA, primarily because GA may form in the parafovea, leading to foveal sparing. This phenomenon results in patients maintaining high VA despite having substantial parafoveal scotomas. Consequently, traditional measures such as the BCVA are insufficient for accurately assessing the functional impact of GA,^5,6” Dr. Ansari and colleagues commented.

Microperimetry, which addresses these limitations by evaluating retinal sensitivity across a broader area, including regions affected by foveal sparing,^7,8 is limitedbecause of the sizable retest variability of microperimetry,^9,10 and the burden associated with testing.

Machine-learning and deep-learning algorithms have potential to infer retinal function from domain spectral-domain optical coherence tomography (SD-OCT) images in a fully automated manner in non-neovascular and neovascular AMD,^8,11 macular telangiectasia type 2 (MacTel),¹² inherited-retinal degeneration,^13,14 and toxic retinopathies.¹⁵ However, using these methods in clinical trials has been problematic because of the current lack of consensus. They theorized that incorporating patient-specific baseline data, such as microperimetry, could enhance the accuracy of inferred functional assessments in subsequent visits.

In the study under discussion, the investigators evaluated the effectiveness of different machine-learning models for predicting retinal sensitivity in GA secondary to age-related macular degeneration (AMD) and compared the progression of sensitivity loss using observed versus inferred data over time, they explained.

A total of 30 (37 eyes) were included in the study, all of whom underwent microperimetry and SD-OCT imaging at baseline and weeks 12, 24, and 48. The retinal layers were segmented using a custom-written deep-learning algorithm.

They used random forest, LASSO (Least Absolute Shrinkage and Selection Operator) regression, and multivariate adaptive regression splines, all machine-learning models, to predict the retinal sensitivity in three scenarios: unknown patients, known patients at later visits, and interpolation within visits. The study goals were the predictive accuracy and the models’ ability to reduce test variability over time, the investigators explained.

What were the results in machine-learning models?

Ansari and colleagues reported, “The random forest model demonstrated the highest accuracy across all scenarios, with mean absolute errors of 3.67 decibels (dB) for unknown patients, 2.96 dB for known patients at follow-up, and 3.10 dB for within-visit interpolation. The inferred sensitivity data significantly reduced the variability compared to the observed data in longitudinal mixed-model analysis, with a residual variance of 2.72 dB² versus 8.67 dB², respectively.”

“This inferred sensitivity mapping approach provides a spatially resolved, automated, and reproducible alternative to traditional psychophysical assessments, which are often limited by variability and practical constraints. Although prior studies have explored cross-sectional structure-function correlations in AMD, our analysis represents a systematic comparison in a longitudinal setting, demonstrating their capacity to track functional progression over time. The findings indicate that inferred sensitivity can serve as a robust functional surrogate endpoint, particularly for natural history studies where psychophysical testing may be impractical,” the authors commented.

Ansari and colleagues summarized, “Our study made it possible to use inferred sensitivity mapping as a potential functional surrogate end point for GA, demonstrating its capacity to predict retinal function with high spatial resolution and without the need for extensive psychophysical testing. In addition, our study demonstrates the superiority of SD-OCT informed ‘interpolation’ over mere hill-of-vision interpolation. Future research should focus on validating this approach in larger, more diverse cohorts and exploring its integration into the design of clinical trials for retinal diseases.”

References

1. Ansari G, Schärer N, Pfau K, et al. Evaluating the progression of retinal sensitivity loss in geographic atrophy using machine-learning-based structure-function correlation (OMEGA 2). Invest Ophthalmol Vis Sci. 2025;66:34. https://doi.org/10.1167/iovs.66.11.34

2. Heier JS, Lad EM, Holz FG, et al. Pegcetacoplan for the treatment of geographic atrophy secondary to age-related macular degeneration (OAKS and DERBY): two multicentre, randomised, double-masked, sham-controlled, phase 3 trials. Lancet. 2023;402:1434–1448.

3. Patel SS, Lally DR, Hsu J, et al. Avacincaptad pegol for geographic atrophy secondary to age-related macular degeneration: 18-month findings from the GATHER1 trial. Eye (Lond). 2023;37:3551–3557.

4. Liao DS, Grossi FV, El Mehdi D, et al. Complement C3 inhibitor pegcetacoplan for geographic atrophy secondary to age-related macular degeneration: a randomized phase 2 trial. Ophthalmology. 2020;127:186–195.

5. Sadda SR, Chakravarthy U, Birch DG, Staurenghi G, Henry EC, Brittain C. Clinical endpoints for the study of geographic atrophy secondary to age-related macular degeneration. Retina. 2016;36:1806–1822.

6. Sunness JS, Rubin GS, Broman A, Applegate CA, Bressler NM, Hawkins BS. Low luminance visual dysfunction as a predictor of subsequent visual acuity loss from geographic atrophy in age-related macular degeneration. Ophthalmology. 2008;115:1480–1488.

7. Pfau M, Lindner M, Steinberg JS, et al. Visual field indices and patterns of visual field deficits in mesopic and dark-adapted two-colour fundus-controlled perimetry in macular diseases. Br J Ophthalmol. 2018;102:1054–1059.

8. Pfau M, von der Emde L, Dysli C, et al. Determinants of cone and rod functions in geographic atrophy: AI-based structure-function correlation. Am J Ophthalmol. 2020;217:162–173.

9. Pfau M, Lindner M, Fleckenstein M, et al. Test-retest reliability of scotopic and mesopic fundus-controlled perimetry using a modified MAIA (macular integrity assessment) in normal eyes. Ophthalmologica. 2017;237:42–54.

10. Alibhai AY, Mehta N, Hickson-Curran S, et al. Test–retest variability of microperimetry in geographic atrophy. Int J Retina Vitreous. 2020;6:16.

11. von der Emde L, Pfau M, Dysli C, et al. Artificial intelligence for morphology-based function prediction in neovascular age-related macular degeneration. Sci Rep. 2019;9:11132.

12. Kihara Y, Heeren TFC, Lee CS, et al. Estimating retinal sensitivity using optical coherence tomography with deep-learning algorithms in macular telangiectasia type 2. JAMA Netw Open. 2019;2: e188029.

13. Sumaroka A, Garafalo AV, Semenov EP, et al. Treatment potential for macular cone vision in Leber congenital amaurosis due to CEP290 or NPHP5 mutations: predictions from artificial intelligence. Invest Ophthalmol Vis Sci. 2019;60:2551–2562.

14. Müller PL, Odainic A, Treis T, et al. Inferred retinal sensitivity in recessive Stargardt disease using machine learning. Sci Rep. 2021;11:1466.

15. Jayakar G, De Silva T, Cukras CA. Visual field sensitivity prediction using optical coherence tomography analysis in hydroxychloroquine toxicity. Invest Ophthalmol Vis Sci. 2022;63:15.