Abstract
Purpose :
This retrospective study assesses the photoreceptor loss/retinal pigment epithelium loss ratio (PR/RPEloss) as a biomarker for classifying progression speeds in geographic atrophy (GA). Extracted from OCT scans with automatic segmentation, this biomarker aims to enhance clinical trial recruitment by identifying active and non-active GA lesions in real-world data on a population level.
Methods :
Patients with GA were identified from a substantial retrospective real-world cohort from the out-patient service at the Vienna General Hospital from 2007 to 2018. This selection was made by sifting through electronic health records (EHR) for indications of age-related macular degeneration (AMD) associated atrophic changes. The final cohort, 54 patients (82 eyes, 386 scans) with 6 months visit intervals, represents an unselected real-world clinical patient cohort with GA. The MDR-approved GA-Monitor (RetInSight, Austria) was used for segmentation and tracking of RPE/PR loss over 30 months post-initial presentation. Patients were divided based on baseline by PR/RPEloss ratio into first quartile (Q1, 21 eyes) and remaining quartiles (Q2-Q4, 61 eyes). Lesion growth was investigated as square root difference in RPEloss area to baseline using a linear mixed model with nested random factor patient/eye, correlating growth rates with baseline ratio groups, controlling for baseline RPEloss size.
Results :
The cohort had an average RPEloss lesion size of 5.62 ± 7 mm2 (median: 1.73 mm2), and PRloss averaged 10 ± 11 mm2 (5.12 mm2). 15 eyes showed no atrophy in the central millimeter, and 9 lacked photoreceptor degeneration. At month 30, the average GA lesion size had reached 9.03 ± 9.58 mm2 (5.65 mm2), and PRLoss averaged 13.4 ± 13.68 mm2 (9.58 mm2), while 9 eyes showed still no central atrophy, and 6 no sign of PR loss. The linear mixed model revealed significant growth speed differences at 24 months (p<0.02) and 30 months (p<0.002) (Figure 1), with a trend toward slower progression in the lower PR/RPE ratio group (Q1) versus higher ratios (Q2-Q4) from baseline on (Figure 2).
Conclusions :
Our study demonstrates in a real-world data set that a lower PR/RPEloss correlates with reduced progression speed in GA on a population level. This highlights the potential of AI-based segmentation in routine OCT scans for patient stratification and optimizing clinical trial design.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.