Exploring Trial Endpoints in Geographic Atrophy Based on Localized Functional Changes in Microperimetry and AI-Quantified OCT Biomarkers.

PURPOSE: The introduction of novel therapeutics for geographic atrophy (GA) highlights the need to define functional correlates of high-risk optical coherence tomography (OCT) biomarkers, in particular retinal-pigment-epithelium (RPE) loss and ellipsoid zone (EZ) loss. We conducted a pointwise structure/function correlation between OCT-based markers and retinal sensitivity (RS) in microperimetry (MP) in GA.

METHODS

Patients from the phase III OAKS clinical trial (NCT03525613) examined by OCT (Heidelberg Spectralis) and Macular Integrity Assessment (MAIA, iCare, and Centervue) MP in a 68-point grid were analyzed. Deep-learning-(DL)-based algorithms quantified RPE, EZ loss, and EZ thickness from OCT. Co-registration between MP and OCT was established between each MP stimulus and OCT B-scan location. A multivariable mixed effect model was implemented to identify RS for each OCT biomarker, accounting for eccentricity.

RESULTS

Six hundred seventy-eight study and fellow eyes of 406 patients with 41,925 MP points were included. Mean RS was 17 ± 7 decibel (dB), 9 ± 7 dB, and 2 ± 6 dB in intact retina, EZ, and RPE loss, respectively. Increased EZ thickness improved RS by 0.2 dB/µm (95% confidence interval [CI] = 0.2 to 0.2, P < 0.001). In areas of EZ loss, RS was significantly reduced compared to intact retina (-8 dB, 95% CI = -9 to -8]), whereas RPE loss decreased RS by -14 dB (95% CI = -15 to -14), accounting for eccentricity (all P < 0.001).

CONCLUSIONS

A significant association between RS in MP and EZ and RPE loss on OCT was established using DL. A reliable quantitative structure/function association provides the base for developing functional endpoints in clinical care and approval of novel therapeutics for GA.