Test–retest variability of microperimetry in geographic atrophy

Abstract Purpose Microperimetry (MP) allows for measurement of retinal sensitivity at precise locations and is now commonly employed as a clinical trial endpoint. Test–retest reliability is important when evaluating treatment effects in patients with geographic atrophy (GA). This study aimed to determine the test–retest variability of MP in patients with moderate to severe GA using the MAIA MP device. Methods In this prospective study, patients with a confirmed diagnosis of foveal-involving GA were enrolled. Participants performed three MP assessments of a selected eye over two visits with the Macular Integrity Assessment (MAIA) 2 instrument (Centervue, Padova, Italy) utilizing a wide 30° grid, consisting of 93 stimuli (Goldmann III) using a 4-2 representation strategy, encompassing the entire area of GA and beyond. Mean retinal sensitivity (MS) was expressed as an average threshold value (dB) for the entire field tested. Coefficients of Repeatability at a 95% level (CoR95) were calculated for Point Wise Sensitivity (PWS). Fixation stability (FS) was assessed by evaluating the area of an elliptical representation encompassing 95% of the cloud of fixation points (CFP) dataset generated by the MAIA MP, known as the bivariate contour ellipse area (BCEA). Results A total of 8 subjects were enrolled (21 tests), with six subjects completing 3 MP assessments. BCVA in these patients ranged from 20/100 to 20/800. The mean area of GA was 18.7 ± 12.3 mm2. The average time to complete one MP assessment was 13 min 9 s and mean BCEA@95% was 38.5 ± 19.3°2. The MS was 14.3 ± 4.5 dB. No significant increase in MS was noted between testing pairs 1&2 and 2&3. The preferred retinal locus was maintained in the same quadrant on successive tests. The mean CoR95 for PWS were similar for testing pairs 1&2 (± 3.50 dB) and 2&3 (± 3.40). Conclusion Microperimetry using a wide grid can be reliably performed in a reasonable amount of time in patients with moderate and severe vision loss secondary to GA. There was no learning effect seen between sequential assessments when analyzing MS or PWS. A change of approximately 4 dB in PWS provides a threshold for considering a true change in this patient cohort.