Assessing the Variability and Repeatability of OCT-A in Uveitis Patients

Abstract

Purpose : Optical Coherence Tomogaphy-Angiography(OCT-A) is a recent imaging modality that is increasingly utilized for assessment of ocular disease. However, its variability and repeatability have been poorly characterized. We performed a retrospective, observational study to characterize both intra-visit and inter-operator variability of OCT-A in uveitis patients.

Methods : 69 Uveitis patients were evaluated with 2 OCT-A scans per eye, per visit in a standardized fashion as part of a clinical research protocol. Consecutive OCT-A scans were recorded by the same technician per visit, separated temporally by several minutes. 2 separate operators retrospectively analyzed images using an algorithm developed by investigators to generate: total vessel area(TVA) and the foveal avascular zone(FAZ) area of both the superficial capillary plexus(SCP) and the deep capillary plexus(DCP), and the total flow-void area of the choriocapillaris(CC). The right eye of each patient was chosen from their earliest available OCT-A scans for variability analyses. Images with low signal intensity or motion artefacts were excluded. Bland-Altman modeling was used to generate a coefficient of repeatability.

Results : Intra-visit variability of total flow-void areas in patients with CC pathology was 0.243 ± .05 mm² with an average flow void area of .585 mm² across all tested eyes(n=18).Inter-operator variability in these patients were .141 ± .05mm² with an average flow void area of .468mm² across all tested eyes(n=24). Among panuveitis patients, intra-visit variability was .286 ±.050 mm²(SCP FAZ),.439 ± .059mm²(DCP FAZ), .553 ± .106 mm²(SCP TVA) and .755 ± .171 mm²(DCP TVA).Inter-operator variability was.349±.056 (SCP FAZ),.586± .0946 mm²(DCP FAZ), 5.8x10^-5± 1x10^-5(SCP TVA), and 2.47x10^-4± 4x10^-5(DCP TVA).Average values for measured eyes were .609(SCP FAZ), .936(DCP FAZ), 2.83(SCP TVA) and 3.55 (DCP TVA) mm² (n=30).Bland-Altman models of data showed no bias and nearly all data points falling within the limits of agreement.

Conclusions : Preliminary data indicates that while OCT-A reasonably fulfills Bland-Altman models of repeatability, with minimal exclusion criteria, it may not be sensitive enough to detect subtle changes within disease as shown by the high coefficients of repeatability relative to average measurements, a finding attributed to the presence of outliers. Further analysis is underway to determine specific criteria for high reproducibility.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.