Test–Retest Repeatability of a Novel Photopic Microperimetry Approach in Patients with Geographic Atrophy Authors: Neha R Tandon, Eric Vasquez, Karl G Csaky Published: 2026 ABSTRACT AbstractPurposeTo evaluate test–retest repeatability of novel focal photopic microperimetry approach targeting central 1 mm of the fovea and explore its relationships with geographic atrophy (GA) lesion area and other visual function metrics.DesignA prospective, single arm, observational study.SubjectsParticipants with nonsubfoveal GA as determined by spectral-domain OCT and fundus autofluorescence.MethodsTwelve participants (17 eyes) performed focal photopic tests using a 45-loci focal grid (0.5° spacing; Goldmann II stimuli) on Nidek MP-3, along with other visual function assessments. To assess test–retest repeatability of MP-3, 2 tests up to 4 weeks apart were performed. The interval was short relative to expected GA progression, and thus measurements are interpreted as reflecting test–retest repeatability.Main Outcome MeasuresCoefficient of repeatability (CoR), intraclass correlation coefficient (ICC), and Bland–Altman were analyzed for mean sensitivity (MS) and point-wise sensitivity (PWS). Correlations of adjusted MS (excluding floor effect) with GA lesion size, best-corrected visual acuity (BCVA), low-luminance visual acuity (LLVA), and contrast sensitivity were also analyzed.ResultsFocal photopic MP-3 showed good test–retest repeatability with CoR of MS at 2.42 dB (intraclass correlation coefficient [ICC] = 0.98) and CoR of PWS at 8.2 dB (ICC = 0.89). Excluding the floor effect improved the CoR of MS and PWS to 2.03 dB and 7.06 dB, respectively. Mean test duration was 4.33 minutes per eye with fixation stability >89%. Structure–function analysis showed weak correlation between GA lesion size and adjusted MS (P = 0.05; R2 = 0.44). Adjusted MS showed a significant correlation with LLVA (P = 0.001) and low-luminance quick contrast sensitivity function (qCSF) (P = 0.01), a weak association with qCSF (P = 0.05), but no correlation with BCVA.ConclusionsFocal photopic microperimetry using Goldmann II stimuli provides a highly repeatable measure of central cone-mediated function in patients with nonsubfoveal GA. Mean sensitivity demonstrated high repeatability and supports its use as a robust, clinically relevant endpoint for longitudinal and interventional studies, while PWS may provide spatial information for deeper structure–function analyses. Weak associations with GA lesion size and BCVA suggest that conventional measures may underestimate visual impairment, highlighting the value of focal photopic microperimetry within central 1 mm region as a sensitive, patient-centered outcome in age-related macular degeneration.Financial DisclosuresProprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.Keywords: AMD, GA, Nonsubfoveal GA, Microperimetry, Cone function Although individuals with geographic atrophy (GA) associated with age-related macular degeneration (AMD) often retain good visual acuity (VA), they experience progressive loss of visual function. These functional changes are subtle and typically go undetected by standard clinical tests, underscoring the need for reliable, cone-specific measures that reflect meaningful central vision loss.1 For patients, enlarging GA lesions have direct impact on their daily lives, translating into slower reading, unstable fixation, and reduced contrast sensitivity, functional losses that significantly affect common visual tasks despite unchanged acuity.2 Quantifying subtle sensitivity losses that precede direct foveal loss may provide clinicians an objective tool to monitor disease progression and to tailor follow-up and intervention timing. For patients, this means functional decline in the fovea can be recognized earlier, before central vision is lost, allowing for timely discussions about emerging treatment options and adaptive strategies. This need for earlier and more sensitive functional assessment is amplified by the high prevalence of AMD, one of the leading cause of vision loss in the aging population, affecting nearly 200 million people, including 1.5 million Americans.3,4 Geographic atrophy, the non-neovascular advanced form of AMD, is defined by progressive degeneration of the retinal pigment epithelium (RPE) and outer retinal layers.5,6 Advanced retinal imaging technologies such as spectral-domain OCT and fundus autofluorescence (FAF) have proven to be very useful in identifying the structural changes associated with AMD progression such as appearance and volume of drusen, hyper-reflective foci, loss of retinal RPE and photoreceptor integrity, and GA area.5, 6, 7, 8 While GA lesion growth is an accepted structural endpoint in clinical trials, it has limited correlation with functional outcomes, such as VA.9 Lesions often expand centrifugally, sparing the foveal center for extended periods. This spatial dissociation between lesion enlargement and central visual function may partially explain why US Food and Drug Administration-approved therapies like pegcetacoplan (Syfovre) and avacincaptad pegol (Izervay) have reduced the rate of GA progression but shown no statistically significant benefit for visual function outcomes.10, 11, 12 As new treatments emerge, the need for repeatable visual function endpoints becomes increasingly critical to detect subtle changes in cone function and to capture preservation of central vision.13,14 Cone photoreceptors are concentrated in the central 1 mm of the fovea, enabling sharp color vision under photopic (well-lit) conditions. In contrast, rods are more abundant in the parafoveal and perifoveal regions and mediate scotopic and mesopic (low light) vision. Histopathologic and clinical evidence indicate that photoreceptor degeneration in AMD follows a rod-before-cone sequence, suggesting distinct biochemical vulnerabilities between these cell types.15,16 Curcio et al demonstrated that rod loss exceeded cone degeneration in the parafoveal region, whereas cones in the foveal center remained relatively preserved, even in advanced AMD. A growing body of evidence implicates early mitochondrial dysfunction and oxidative stress within the RPE as initiating events that disrupt photoreceptor homeostasis and activate chronic inflammation through the complement pathway.17,18 Rod photoreceptors, which have high metabolic demand and depend heavily on RPE support, appear to degenerate first under these conditions. As rods degenerate, they trigger a cascade of metabolic and structural changes that ultimately impair cone function and central vision.19 As such, traditional high-contrast best-corrected visual acuity (BCVA) often underestimates functional deficits in AMD, particularly in cases with foveal-sparing GA.20 Functional measures like low-luminance best-corrected visual acuity (LLVA), contrast sensitivity, and microperimetry techniques currently used in clinical trials capture aspects of AMD progression that traditional VA cannot, but their variability and inconsistent structure–function agreement limit broader application.13,21, 22, 23, 24, 25 Microperimetry, a psychophysical visual field test, allows us to spatially map retinal sensitivity by presenting light stimuli to the retina and recording the patient’s response, enabling topographic correlation between functional loss with structural changes across the macula.25 Microperimetry can be performed under different light adaptation conditions—photopic, mesopic, and scotopic—each targeting different photoreceptor functions.26 Mesopic and scotopic microperimetry can detect early rod-based dysfunction in AMD even before cones are affected. Several studies have shown a decline in mesopic retinal sensitivity with progression of macular degeneration.27,28 However, variability in test–retest repeatability, especially near scotoma borders,29 and differences across devices (such as CenterVue Macular Integrity Assessment [MAIA] and Nidek MP-1) have resulted in difficulties in utilizing mesopic microperimetry, in its present format, as a clinical trial endpoint. Photopic microperimetry, conducted under luminance conditions that suppress rod activity, primarily assesses cone-mediated sensitivity and is functionally relevant for patients with foveal sparing GA. The Nidek MP-3 (MP-3) device offers the ability, under illuminated background conditions, to selectively measure photopic sensitivity in the central cone-dense fovea and uniquely to utilize smaller Goldmann II stimuli (not available at present on the MAIA). In addition, the MP-3 features improved eye tracking, wide dynamic range (0–34 dB) and easily customizable grids. This test–retest repeatability study was designed to evaluate novel focal photopic microperimetry using the MP-3 to measure cone-mediated retinal sensitivity within the central 1 mm region in patients with nonsubfoveal GA. By employing a smaller 63 μm spot size and a dense central grid, this approach was designed to capture subtle functional loss at the fovea with greater precision and repeatability than conventional protocols. Establishing robust test–retest repeatability is a critical step toward validating focal microperimetry as a sensitive, trial-ready endpoint for monitoring GA progression and therapeutic response. This study also explored the relationship between the central cone-mediated sensitivities with GA lesion size and other visual function metrics such as BCVA, LLVA, quick contrast sensitivity function [qCSF], and low-luminance quick contrast sensitivity function (LL-qCSF). Methods This study was conducted with the approval of the University of Texas Southwestern and WCG Institutional Review Board. All study protocols were adherent to the tenets set forth in the Declaration of Helsinki and the Health Insurance Portability and Accountability Act. All study participants provided written informed consent. Participants Patients with nonsubfoveal GA (determined by a retina specialist [K.G.C.]) were included in this study. Subjects were screened using assessments of BCVA, slit-lamp examination, spectral-domain OCT volume scans (Spectralis), and FAF images (Spectralis). Foveal involvement status was determined using multimodal imaging. Fundus autofluorescence images were used to delineate GA boundaries; however, given the reduced autofluorescence signal at the fovea, foveal sparing was confirmed using corresponding spectral-domain OCT B-scans. The foveal center was identified on OCT, and preservation of outer retinal layers was required to classify eyes as having nonsubfoveal (foveal sparing) GA. Subjects with BCVA scores better than or equal to 20/80 were included, as eyes with poorer VA had difficulty completing focal microperimetry testing, with increased fatigue, reduced fixation stability, and inconsistent responses. Patients with choroidal neovascularization, any corneal pathology that could compromise vision, cataract, significant glaucoma, and physical or mental impairment that would prevent them performing the microperimetry test were not included in the study. Visual Function Tests Best-corrected visual acuity and LLVA were measured using the trial frame refraction and ETDRS charts with standard light box method. Best-corrected visual acuity was assessed under room luminance of 130 cd/m2 and LLVA was assessed by placing a 2.0-log unit neutral density filter in front of the trail lenses. Best-corrected visual acuity and LLVA results were recorded in letter scores. Contrast sensitivity was measured as qCSF and LL-qCSF using the Manifold Contrast Vision Meter (Adaptive Sensory Technology), as described previously.21 Testing was performed with the patient’s spectacle correction, and a 2.0-log unit neutral density filter was used for LL-qCSF. The results were recorded as area under line contrast sensitivity function for our analysis. Imaging Spectral-domain OCT and FAF was performed using the Heidelberg Spectralis HRA + OCT device (SN: Spec-Cam-11496-S3610; Heidelberg Engineering). OCT volume scans were obtained with dimensions of 10° × 15° and 20° × 20°, centered on the fovea and consisting of 97 sections in high resolution mode. Fundus autofluorescence images were obtained in high resolution mode centered on the fovea. Total GA area was measured as the total hypo-autofluorescence area using the built-in RegionFinder tool available on the Heidelberg Eye Explorer (HEYEX) software (version: 2.6.7). For patients with multifocal GA lesions, the total GA area was reported as the sum of each of the measured hypo-autofluorescence areas. Microperimetry Testing To assess test–retest repeatability, the subjects performed 2 photopic microperimetry tests on the MP-3 (Nidek Technologies). MP-3 photopic testing uses a background of illumination constant at 10 cd/m2 with maximum stimulation luminance at 3193 cd/m2 (0 dB) and minimum luminance at 11 cd/m2 (34 dB). MP-3 captures a 45° field of view infrared (IR) image for motion capturing with a tracking frame rate of 30Hz. The study eye was dilated using 1% Tropicamide and 2.5% Phenylephrine, and the test was conducted, without dark adaptation, and with the fellow eye patched. Testing room lights were turned off right before the test, and room luminance was then maintained at <0.1 lux during the test. The testing stimuli pattern consisted of 45 points in a 4° by 2° grid (approximately 1.2 mm × 0.6 mm) centered on the fovea, with spacing of 0.5° between adjacent locations (Fig 1). The fixation target was set as a white single cross with 1° width. The stimulus size was set at Goldmann II (0.21° or 63 um diameter) with color and duration set as white and 200 ms, respectively. The testing was performed using a 4-2 fast staircase strategy with the starting threshold set at 17 dB. (A) Schematic of Nidek MP-3 focal microperimetry grid pattern. (B) Representative image showing MP-3 fundus image output with grid and resulting sensitivities overlay; zoomed-in image of grid with sensitivity values to the right, and (C) FAF image with grid overlay manually generated using Photoshop (Adobe); zoomed-in image of grid with sensitivity values to the right. FAF = fundus autofluorescence. On test day 1, a trained staff provided detailed instructions to the subject before starting the test and guided the subject during the test to maintain fixation stability. The retest was performed within 2 to 4 weeks of test 1, using the built-in follow-up feature of MP-3. The MP-3 utilizes the IR image from test 1 (baseline) as the reference image to align the IR image of the follow-up examination to position the stimuli pattern (the grid) on the same location. The device generates an overlay image to confirm the alignment of baseline and follow-up IR images. If the automated alignment was not accurate, the examiner manually adjusted the alignment before starting the follow-up test. Statistical Analysis Microperimetry outcome measures used for statistical analysis were extracted from the PDF report generated by Navis-EX software. Raw data and fundus image with grid overlay images were also exported from the system. Mean sensitivity, fixation stability, and bivariate contour ellipse area (BCEA) scores were extracted from the PDF report. The data for point-wise sensitivity (PWS) at specific coordinates were obtained from the raw data files. Image files were used to create OCT and FAF enface images with grid overlay using Photoshop (Adobe) to confirm grid position between the 2 tests. Statistical analysis was performed using GraphPad Prism 11 and SAS software. Descriptive statistics and Bland–Altman analysis were used to assess agreement between test sessions. Bland–Altman plots were interpreted descriptively, and not for formal inference, given the clustered structure of data. Intraclass correlation coefficient and coefficient of repeatability (CoR) with 95% confidence intervals (CIs) were calculated to assess intersession test–retest reproducibility and repeatability. Intraclass correlation coefficient (ICC) is an index of reliability and represents the ratio of variance between subjects and total variance. For mean sensitivity (MS), variance measurements for ICC and CoR were estimated using linear mixed-effects model to account for clustering of eyes within subjects, with subject included as a random effect. For PWS, the hierarchical structure of the data (loci nested within eyes, nested within subjects) was accounted for using the linear mixed-effects model. Random intercepts were included to model clustering at the subject level, with loci and eye-level measurements incorporated within the model structure. Coefficient of repeatability was calculated as recommended by Bland–Altman,30,31 to represent the difference between 2 measurements on the same subject to fall within ± CoR absolute value on 95% of occasions. Intraclass correlation coefficient, CoR, and 95% CI of CoR were calculated as follows: Where σP2 is between subject variance, σE2 or σws2 variance due to random error or within subject variance. Correlation analysis was performed to assess correlations between GA lesion size, adjusted MS, and other visual function metrics—BCVA, LLVA, qCSF, and LL-qCSF. These analyses were also conducted using the linear mixed-effects models to account for intereye correlation. Correlation coefficient (r), 2-tailed P value, and 95% CIs were calculated. To control for multiple comparisons across the 5 correlations, P values were adjusted using the Bonferroni–Dunn correction, with the significance threshold set at an adjusted α = 0.05. The relationship between GA lesion size and adjusted MS was further quantified within the linear mixed-effects modeling framework, and the coefficient of determination (R2) was derived to estimate the proportion of variance in MS explained by GA lesion size. Results A total of 17 eyes from 12 subjects (6 men and 6 women, age ≥65 years) with nonsubfoveal GA were included in the study. When both eyes met eligibility criteria, both were included. Five participants contributed both eyes, while 7 contributed a single eye. Quick contrast sensitivity function and LL-qCSF data were unavailable for a subset of eyes (n = 15) due to technical issues during testing and data acquisition, resulting in incomplete or unusable measurements. The median BCVA score for the cohort was 20/32, and the median GA lesion size was 3.76 mm2. The median number of weeks between test 1 and test 2 was 2.6 weeks. The average (± standard deviation [SD]) MS for test 1 was 14.15 dB (±5.5) and test 2 was 13.61 dB (±5.7). The difference between average MS from test 1 to test 2 was –0.53 dB (P = 0.072). The test–retest repeatability for mean sensitivities was high with CoR (95% CI) of 2.42 dB (1.26–3.54). The ICC score between the 2 tests was 0.98, demonstrating strong agreement between the 2 tests. The average (± SD) PWS for 765 loci (45 points in the grid ∗ 17 eyes) for test 1 and test 2 was 14.14 dB (±8.7) and 13.61 dB (±9.2), respectively. The CoR (95% CI) and ICC score for PWS between test 1 and test 2 was 8.20 dB (7.62–8.8) and 0.89, respectively (Table 1). Table 1Assessment of Test–Retest Repeatability for MS and PWS with and without Exclusion of Floor Effect (≤3 dB)nTest 1Mean (±SD)(dB)Test 2Mean (±SD)(dB)PCoR (dB)95% CI of CoR (dB)ICCMS1714.15 (5.5)13.61 (5.7)0.0722.421.26 – 3.540.98Adjusted MS (excluding ≤3 dB)1717.82 (3.2)17.58 (3.4)0.352.031.06 – 3.00.95Average PWS76514.14 (8.7)13.61 (9.2)<0.0018.27.62 – 8.80.89Average adjusted PWS (excluding ≤3 dB)56118.43 (5.2)18.25 (5.6)0.277.066.5 – 7.640.79ICC: >0.90 = excellent, 0.75-0.90 = good, 0.50-0.75 = moderate, and <0.50 = poor reliability.CI = confidence interval; CoR = coefficient of repeatability; dB = decibels; ICC = intraclass correlation coefficient; MS = mean sensitivity; P = P value; PWS = point-wise sensitivity; SD = standard deviation. Assessment of Test–Retest Repeatability for MS and PWS with and without Exclusion of Floor Effect (≤3 dB) ICC: >0.90 = excellent, 0.75-0.90 = good, 0.50-0.75 = moderate, and <0.50 = poor reliability. CI = confidence interval; CoR = coefficient of repeatability; dB = decibels; ICC = intraclass correlation coefficient; MS = mean sensitivity; P = P value; PWS = point-wise sensitivity; SD = standard deviation. Bland–Altman analysis also showed good agreement between test 1 and test 2, with limits of agreement (LoA) between –1.71 and +2.78 dB for mean sensitivities (Fig 2A) and between –7.6 and +8.68 dB for PWSs (Fig 2B). Bland–Altman analysis of PWS demonstrated significant floor effect shown by the reduced variability and clustering of sensitivities <3 dB (Fig 2D). Repeat analysis of adjusted MS after exclusion of floor effect (loci ≤3 dB) improved the CoR (95% CI) to 2.03 dB (1.06 – 3.0) with no significant change to ICC score. After excluding the floor effect, 561 loci were used to analyze the CoR, ICC, and LoA for PWS. The CoR (95% CI) for adjusted PWS after exclusion improved to 7.06 dB (6.5–7.64); however, the ICC decreased from 0.89 to 0.79 (Table 1). The Bland–Altman analysis after exclusion did not change LoA for adjusted MS (Fig 2C) but showed slight reduction in LoA for adjusted PWS (Fig 2D). Bland–Altman analysis for MS and PWS with and without exclusion of floor effect: (A) Bland–Altman plot for MS for all patients, (B) Bland–Altman plot for adjusted MS for all patients after excluding loci with sensitivities ≤3 dB, “Adjusted,” (C) Bland–Altman plot for average of PWS for all 765 stimuli from 17 patients, and (D) Bland–Altman plot for average of PWS after excluding loci with stimuli threshold ≤3 dB, 561 stimuli from 17 patients, “Adjusted.” Horizontal line represents mean of the difference between test 1 and test 2 (i.e., the bias from Bland–Altman plots) and the horizontal dotted line represents 95% LoA. dB = decibels; LoA = limits of agreement; MS = mean sensitivity; PWS = point-wise sensitivity. The mean elapsed time taken for testing each eye across both tests was 9.56 (±1.9) minutes. The elapsed time is total duration of the session and includes time taken for retina alignment. The average tracked time of the test was 4.33 (±0.76) mins. The tracked time is the actual total duration of the test, when the stimuli is presented and patient is actively responding. The difference between the average tracked time from test 1 to test 2 was only –0.62 min but was significant with P = 0.015. The average fixation stability within 2° diameter across both tests was 89.32% (±12.56) and the average BCEA at 95.4% SD across both tests was 3.81 (±2.86) degree square. There was no significant difference in average fixation stability and BCEA scores between test 1 and test 2. Eye tracked time, fixation stability, and BCEA at 95.4% SD showed high CoR and poor ICC. The false positive cutoff was set at 33.33% (Table 2). Table 2Assessment of Test–Retest Subject PerformancenTest 1Mean (±SD)Test 2Mean (±SD)PCoR95% CI of CoRICCEye tracked time (mins)174.64 (0.8)4.02 (0.6)0.0011.760.92 – 2.60.34Average fixation within 2° Ø (%)1790.29 (8.04)88.35 (16.09)0.6129.715.6 – 43.810.36BCEA at 95.4% SD (degree sq.)173.8 (2.35)3.82 (3.37)0.976.673.5 – 9.840.37Ø = diameter; BCEA = bivariate contour ellipse area; CI = confidence interval; CoR = coefficient of repeatability; ICC = intraclass correlation coefficient; P = P value; SD = standard deviation; sq. = square. Assessment of Test–Retest Subject Performance Ø = diameter; BCEA = bivariate contour ellipse area; CI = confidence interval; CoR = coefficient of repeatability; ICC = intraclass correlation coefficient; P = P value; SD = standard deviation; sq. = square. Frequency distribution of the difference between PWS from test 1 to test 2 showed that 80% of the stimuli had a difference of ±4 dB (Table 3). Further analysis of the test locations within GA lesion showed that 92.5% of test locations that were recorded at 0 dB during test 1, returned a value of 0 dB or ≤3 dB during test 2. Table 3Frequency of Difference in PWSDifference (t2 - t1) (dB)Number of StimuliPercentage of Stimuli (%)Cumulative Percentage (%)024031.431.4–1 or +1121.632.9–2 or +223330.563.4–4 to –3 or 3 to 412716.680.0–6 to –5 or 5 to 67710.190.1–8 to –7 or 7 to 8364.794.8–10 to 9 or 9 to 10172.297.0Other233.0100.0Total stimuli765dB = decibels; PWS = point-wise sensitivity; t1 = test 1; t2 = test 2. Frequency of Difference in PWS dB = decibels; PWS = point-wise sensitivity; t1 = test 1; t2 = test 2. Exploratory structure–function correlation analysis using the mixed effects model showed moderate negative association between GA lesion size and adjusted MS (r = –0.66, P = 0.01, R2 = 0.44) (Fig 3). After Bonferroni–Dunn correction for multiple comparisons (significance threshold P ≤ 0.01), the association between GA lesion size and adjusted MS was not statistically significant (corrected P = 0.05). When examining the relationship between adjusted MS and other visual function measures, adjusted MS showed significant correlations with LLVA (r = 0.78, P = 0.0003), qCSF (r = 0.66, P = 0.01), and LL-qCSF (r = 0.76, P = 0.002), but no significant association with BCVA. After Bonferroni–Dunn correction, the associations between adjusted MS and LLVA (corrected P = 0.001), and MS and LL-qCSF (corrected P = 0.01) remained significant (Table 4). Post hoc analysis indicated that the study had limited statistical power, with estimated power of 0.56 for full dataset (n = 17) and 0.50 for the qCSF and LL-qCSF data subset (n = 15). Regression analysis of GA lesion area and adjusted MS: ordinary least square regression model showing linear trend (solid line) and 95% confidence bands (dotted line) between GA size and adjusted MS. dB = decibels; GA = geographic atrophy; MS = mean sensitivity. Table 4Correlations of Adjusted MS with GA Lesion Size and Other Visual Function MetricsnCorrelation Coefficient (r)95% CIP (raw)P (corrected)GA lesion area Adjusted MS (excluding ≤3 dB)17–0.66–0.88 to –0.210.010.05Adjusted MS (excluding ≤3 dB) BCVA170.34–0.19 to 0.710.201.0 LLVA170.780.48 – 0.920.00030.001∗ qCSF150.660.19 – 0.880.010.05 LL-qCSF150.760.38 – 0.920.0020.01∗ρ = Spearman correlation coefficient; CI = confidence interval; BCVA = best-corrected visual acuity; dB = decibels; GA = geographic atrophy; LL-qCSF = low-luminance quick contrast sensitivity function; LLVA = low-luminance visual acuity; MS = mean sensitivity; P (corrected) = Bonferroni-Dunn corrected; P (raw) = raw P value; qCSF = quick contrast sensitivity function.∗P (corrected) ≤ 0.01 is significant. Correlations of Adjusted MS with GA Lesion Size and Other Visual Function Metrics ρ = Spearman correlation coefficient; CI = confidence interval; BCVA = best-corrected visual acuity; dB = decibels; GA = geographic atrophy; LL-qCSF = low-luminance quick contrast sensitivity function; LLVA = low-luminance visual acuity; MS = mean sensitivity; P (corrected) = Bonferroni-Dunn corrected; P (raw) = raw P value; qCSF = quick contrast sensitivity function. P (corrected) ≤ 0.01 is significant. Discussion We evaluated the test–retest repeatability of a novel focal photopic microperimetry approach using the MP-3 to measure retinal sensitivity within the central, cone-dense 1 mm zone in patients with nonsubfoveal GA. Our custom grid included 45 stimulus points spaced 0.5° apart in a 1.2 mm × 0.6 mm pattern and Goldmann II stimulus size. We achieved efficient testing with a mean testing time of ∼4.3 min per eye along with high fixation stability (>89% of points within 2°) across sessions, indicating repeatable performance in eyes with nonsubfoveal GA. We analyzed 17 eyes from 12 subjects and found an average MS of 13.9 dB, across 2 visits, with a negligible difference between visits (–0.53 dB). Average MS showed excellent repeatability (CoR = 2.42 dB; ICC = 0.98), indicating that changes exceeding the threshold of 2.42 dB are likely to represent true functional change rather than test variability. Point-wise sensitivity across 765 loci demonstrated moderate repeatability (CoR = 8.20 dB; ICC = 0.89). Repeatability improved after excluding floor effect, that is, locations with sensitivity threshold <3 dB (adjusted MS CoR = 2.03 dB and adjusted PWS CoR = 7.06 dB). Bland-Altman analysis confirmed these findings. The structure-function analysis revealed a weak linear relationship between GA lesion size and adjusted MS (r = –0.66, P = 0.01; R2 = 0.44), which did not remain significant after Bonferroni-Dunn correction (corrected P = 0.05). This finding suggests that total lesion area spanning para- and perifoveal regions may not reliably reflect central cone-mediated functional loss in patients with foveal-sparing GA. Notably, global GA lesion area may underestimate structure–function relationships within the foveal region, where preservation of cone photoreceptors can persist despite substantial extrafoveal atrophy. Central retinal sensitivity is more likely to reflect local outer retinal layer integrity such as the ellipsoid zone and retinal pigmented epithelium (RPE). However, the relatively small sample size may have limited the ability to detect statistically significant associations, particularly for structure–function analyses. Post hoc power analysis suggests that the study was underpowered to detect small to moderate effect sizes. Therefore, the lack of significant association between GA lesion size and adjusted MS should be interpreted cautiously and may reflect limited statistical power rather than a true absence of relationship. Adjusted MS showed strong correlations with low-luminance and contrast-based visual measures; LLVA (r = 0.78), qCSF (r = 0.66), and LL-qCSF (r = 0.76). Both low luminance measures—LLVA and LL-qCSF—remained significant after multiple-comparison adjustment. These findings indicate that photopic MP-3 sensitivity better reflects visual performance under low-luminance or reduced-contrast conditions than under high-luminance high-contrast conditions of standard BCVA. The strongest visual function relationships were between MS and low-luminance metrics, consistent with the concept that parafoveal cone dysfunction affects visual performance before measurable BCVA loss.2 It should be noted that qCSF and LL-qCSF data were missing in a small subset of eyes due to technical issues during data acquisition. Although the missingness was technical in nature and not attributable to clinical factors, the reduced sample size for these measures may have limited statistical power and influenced the strength of observed correlations. Wu et al demonstrated that a focal mesopic testing protocol, using a high density 1.75° grid with 0.5° interstimulus spacing and Goldman III stimuli on the MAIA platform, achieved excellent repeatability and shorter testing times than a standard 6° radial grid.32 This method was applied only to small areas overlying nascent or established GA. While valuable for tracking localized progression of retinal sensitivity loss, it does not provide information about central cone-mediated function, which is critical for patient-reported visual outcomes and clinical relevance in AMD. Our study extends this focal approach into the cone-mediated photopic domain using the MP-3 with a small, highly dense grid. The lower average MS (13.9 dB) in our study reflects the inclusion of patients with nonsubfoveal GA, in which complete or partial foveal sparing atrophy reduces central cone-mediated function. Mean sensitivity strongly influences repeatability metrics: eyes with lower baseline sensitivity generally show greater test–retest variability because responses approach the psychophysical detection limit. Despite this inherent challenge, the MP-3 achieved a CoR of 2.42 dB and ICC of 0.98 for MS, demonstrating excellent repeatability even in a cohort with reduced retinal function. The minimal difference (–0.53 dB) observed between the average MS of 2 test sessions suggests that photopic MP-3 does not exhibit a significant learning effect, thereby supporting its reliability for repeated functional assessment without the need for discarding the first test.29,33 In contrast, mesopic microperimetry using the CenterVue MAIA or Nidek MP-1 platforms often demonstrate a significant learning effect and underestimation of sensitivity during the initial test, and necessitate a separate training session or exclusion of the first test for improved accuracy.34, 35, 36 By operating under light-adapted, cone-mediated conditions, the photopic protocol ensures consistent cone-mediated responses and improves reliability for longitudinal monitoring. No direct head-to-head comparison with MAIA or MP-1 was performed in this study, as these platforms are limited to the use of a Goldmann III stimulus size and mesopic testing conditions. Accordingly, any comparisons should be interpreted as indirect and intended to provide methodological context rather than definitive cross-platform evaluation. The MP-3 introduces several performance advantages that enhance both sensitivity measurement and overall testing precision. Its expanded dynamic range (0–34 dB) prevents the ceiling effects reported with the MP-1 (0–20 dB).37 As a result, our study showed no ceiling effect in PWS or MS measurements. The MP-3’s high-speed fundus tracking (30 Hz) and automated follow-up alignment precisely reregister test loci between sessions, reducing spatial variability that can alter results. These technical refinements, combined with the availability of Goldmann II stimuli, allow for finer assessment of residual cone-mediated sensitivity in the fovea. Together with a short average test duration (∼4.3 min per eye) and high fixation stability (>89% within 2°), these features demonstrate that patients with nonsubfoveal GA can perform photopic testing efficiently and reproducibly. Patients with GA reliably detected Goldmann size II stimuli, indicating that this stimulus size is appropriate for assessing residual retinal sensitivity in this population. The use of Goldmann size II stimulus, with its smaller stimulus size, enhances spatial resolution and reduces spatial summation and light spillover from adjacent retinal areas, improving the specificity of functional measurements within a localized region. Frequency distribution analysis showed that 80% of test loci differed by only ±4 dB between sessions, confirming that the smaller stimulus preserved precision without increasing variability. This high level of agreement indicates that Goldmann II stimuli preserve test-retest reliability and can enhance measurement specificity. Test locations registering 0 dB within the GA lesion indicate a complete loss of cone-mediated sensitivity, suggesting advanced photoreceptor degeneration in these regions. Notably, 92.5% of 765 loci during test 1 remained at 0 dB or ≤3 dB on retesting, indicating that these locations are functionally inactive or exhibit minimal residual sensitivity. This result mirrors the <3 dB functional floor defined by Wu et al using mesopic microperimetry on the MAIA platform, where responses below this threshold no longer correlate with structural integrity on OCT.33 The concordance between studies supports interpreting loci ≤3 dB as representing true loss of a measurable function rather than test variability. However, their proximity to the lower end of the dynamic range (0–34 dB) reduces their utility for detecting change or for structure–function analyses. This photopic protocol provides the first repeatability dataset focused specifically on cone-mediated sensitivity in central 1 mm of the fovea, where residual function remains measurable and clinically meaningful. The repeatability of MS in this study (CoR 2.42 dB), closely matches prior MP-3 reports in GA and other macular diseases that used wider grids with Goldmann III stimulus size.38 Point-wise sensitivity showed greater variability (CoR 8.2 dB) than reported by Palkovits et al (∼5 dB). This difference likely reflects the use of the 4-2-1 staircase stimulus presentation by Palkovits et al and the floor effect observed in our data. Loci with sensitivity ≤3 dB comprised approximately 20% of all the point-wise measurements and were clustered near the lower bound of the MP-3 dynamic range (0 dB), where further functional decline cannot be reliably quantified. Empirically, Bland–Altman analysis of MS demonstrated LoA of approximately ±2 dB, indicating that variability in this range approaches the measurement noise of the device. A threshold of ≤3 dB was selected as a conservative cutoff to minimize the influence of the lower-bound censoring and repeatability error near the measurement floor. In addition, floor effects in the MP-3 device have been previously reported, and a similar cutoff of ≤3 dB has been previously applied by Palkovits et al. Excluding the floor effect markedly improved repeatability, confirming that very low-sensitivity points inflate variability. While exclusion of loci near the measurement floor reduces variability and improves agreement metrics such as CoR, it simultaneously reduces between-subject variability and compresses the dynamic range, resulting in lower ICC (adjusted MS ICC = 0.95 and adjusted PWS ICC = 0.79). This underscores the importance of interpreting repeatability metrics in the context of both absolute and relative reliability, particularly in datasets affected by floor effects. Results were consistent when alternative thresholds for floor effects (e.g., 0 dB, ≤2 dB, and ≤5 dB) were explored (Supplemental Table 1, available at www.ophthalmologyscience.org). Anatomical variability of the GA lesions in our cohort might also be a contributing factor to the higher PWS CoR in our study. Wu et al showed lower test–retest repeatability near the edges of scotomas, where sensitivity transitions steeply.29 Consistent with this, a recent study comparing MAIA and MP-3 across GA zones found PWS repeatability to be similar (∼±6.6 dB) with slightly higher variability in regions of photoreceptor or RPE loss.39 Together, these results demonstrate that variability in our data largely reflects biological and anatomical factors rather than patient or instrument-related noise. The dense focal grid centered on the fovea may have contributed to fixation stability across the cohort. Of the 17 eyes tested, 13 demonstrated stable fixation (75% of fixation points within a 2° circle), and only 4 exhibited relatively unstable fixation (>75% of the fixation points within a 4° circle and <75% fall within a 2° circle). The grid design likely facilitated more consistent central fixation, especially in patients with preserved foveal function, thereby enhancing the reliability of central sensitivity measurements. Moreover, the inclusion of eyes with BCVA ≥20/80 may have also contributed to the high fixation stability observed in this cohort. However, the subject performance metrics—eye-tracked time, fixation stability, and BCEA—demonstrated lower repeatability, as reflected by the lower ICC values. This likely represents the intersubject variability in fixation behavior, which is well recognized in GA, where patients may adopt different preferred retinal locus and exhibit variable fixation patterns across sessions. In this study, fixation metrics were not used as primary outcomes but were assessed to ensure adequate test performance. Future studies will further explore the role of preferred retinal locus location, fixation stability metrics, and its association with visual function and structure changes in patients with GA. This focal approach achieved better MS repeatability between test sessions than several previous studies using MAIA and MP-1 devices with wider grids spanning 10° to 20° of macula and employing Goldmann III stimulus size.37 Wider grids comprising >45 test loci increased test time and variability, whereas the compact grid with photopic testing on the MP-3 increased reliability and reduced test time to ∼4.3 minutes per eye. These attributes make the method practical for clinical settings and scalable for interventional trials.33 Our findings validate focal photopic microperimetry with the MP-3 as a highly repeatable and efficient functional measure for central visual function in patients with nonsubfoveal GA. This approach may be applied to examine localized losses in retinal sensitivity and how it relates to photoreceptor disruption. Establishing these relationships may clarify how central cone function declines as atrophy encroaches toward the fovea and may help identify potential predictors of preserved central vision. Structure–function assessments reinforce the clinical relevance of cone-mediated photopic testing for detecting functional impairment that precedes VA loss. The modest relationship between GA lesion size and focal sensitivity highlights that structural enlargement alone may underestimate disease burden. Functional endpoints such as focal photopic microperimetry provide complementary, patient-centered insight into early cone dysfunction and may serve as more sensitive indicators of therapeutic effect in interventional trials. While mesopic microperimetry and rod-based testing are important for detecting early disease, they are less effective for capturing functional changes in foveal-sparing GA. Shen et al40 showed that in foveal-sparing GA, atrophic damage typically affects the 0.5–1.5 mm parafoveal ring surrounding the fovea, a region where cone vulnerability begins to manifest. Similarly, Pfau et al7 demonstrated that photoreceptor degeneration precedes RPE loss and occurs in areas outside the atrophic lesion, indicating structural compromise before lesion expansion. Together, these findings demonstrate that even eyes with preserved VA may already experience subtle cone-mediated functional decline. Measuring sensitivity in this region could serve as a sensitive biomarker of early functional loss. By targeting this region with a photopic, cone-mediated conditions, focal MP-3 testing captures early foveal dysfunction that directly impacts patient-relevant tasks such as reading and facial recognition. As regression analyses integrate focal sensitivity with OCT and FAF features, they may reveal how GA progression toward fovea may forecast compromise of the structural and functional integrity of photoreceptors in the fovea, clarifying disease trajectory and identifying candidates most likely to benefit from interventions aimed at preserving central vision. Lee et al41 showed that photoreceptor loss often precedes visible outer retinal atrophy on OCT and that residual cone function can persist despite ellipsoid zone disruption or outer retinal tubulation. Erb et al42 further demonstrated that macular tissue integrity and ellipsoid zone preservation within the central 1 mm of the fovea exhibited a stronger correlation with BCVA than similar structural metrics obtained from the broader 3 mm region. Together, these findings support the value of high-resolution functional testing in regions that appear functionally compromised. Photopic microperimetry can identify “functionally viable” photoreceptors within these zones, providing a sensitive means of assessing the efficacy of treatments aimed at preserving foveal function rather than merely slowing peripheral lesion expansion. By focusing on the central 1 mm, focal photopic MP-3 testing directly targets the cone-rich region responsible for high-resolution vision, aligns with patient-reported functional concerns such as reading and facial recognition, and offers a repeatable, trial-ready endpoint for monitoring meaningful visual change. Defect-mapping microperimetry enables dense sampling of retinal sensitivity at lesion borders and offers unique value in research settings by employing customized, high-density testing patterns for detailed structure–function correlation.33,43 However, it is more time-intensive and difficult to standardize, making it less practical for large-scale or longitudinal trials. It remains better suited to derive mechanistic insights into disease progression. In contrast, by using the Nidek MP-3, our approach leverages its high repeatability, robust fixation tracking, and accurate follow-up alignment to detect early functional changes that precede observable structural damage, while retaining the flexibility to apply customized grids when needed. Focal photopic microperimetry with Goldmann II stimulus size using MP-3 demonstrated excellent test–retest repeatability for within the central cone-rich foveal region in GA. The high repeatability and short testing time observed here supports focal photopic microperimetry as a reliable, patient-centered functional endpoint for assessing central visual function in GA. While PWS demonstrated substantially greater variability, even after exclusion of floor effect, MS across the focal grid showed excellent repeatability, supporting its use a robust endpoint for longitudinal and interventional studies. The higher variability of PWS limits their utility as primary outcomes, although it may be better suited for exploratory or spatially resolved analysis of localized structure–function relationships. Collectively, these findings provide a strong foundation for future structure–function analyses that align microperimetry results with OCT and FAF-derived markers of retinal integrity and disease progression. Despite these strengths, several limitations warrant discussion. The sample size (17 eyes), although reasonable to establish test–retest repeatability, limits extrapolation to all GA phenotypes. Larger, multicenter studies will be needed to confirm these findings and to determine whether repeatability holds across varying disease stages. Longitudinal follow-up will also be essential to quantify how photopic sensitivity changes with as GA progression. Moreover, inclusion of eyes with BCVA ≥20/80 may limit generalizability to patients with more advanced central vision loss and may have contributed to high repeatability and fixation stability observed in this cohort. However, this is consistent with the intended application of this approach in clinical trials targeting early central visual function loss, where sensitive and repeatable functional endpoints are critical for detecting and preserving remaining vision. Given that the microperimetry grid was limited to central foveal region, we plan to further investigate how localized sensitivity loss corresponds to patterns of photoreceptor and RPE integrity, which may provide more sensitive and clinically meaningful correlations. This structure–function mapping will localize and quantify functional loss in relation to underlying retinal morphology, potentially identifying early markers of photoreceptor compromise before atrophy develops. Declaration of Generative AI and AI-Assisted Technologies in the Writing Process During the preparation of this work the author(s) used ChatGPT in order to assist with grammar, spelling, and sentence structure. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication. Acknowledgments The authors gratefully acknowledge Laura Erker, PhD and Sanchareeka Dey, PhD for their professional assistance with scientific writing support, including critical review, editorial support, proofreading, and grammatical refinement of the manuscript. Manuscript no. XOPS-D-25-01030. Footnotes line: Presented at the Association for Research in Vision and Ophthalmology (ARVO) Annual Meeting, 2024, May 5-9, Seattle, Washington. Supplemental material available atwww.ophthalmologyscience.org. Disclosure(s): All authors have completed and submitted the ICMJE disclosures form. The authors made the following disclosures: N.R.T.: Consultant — Aviceda Therapeutics, Boehringer Ingelheim, Novartis, Vivid Vision. K.G.C.: Financial support — AbbVie, Annexon Biosciences, Boehringer Ingelheim, EyeBio, Genentech/Roche, Astellas, Heidelberg Engineering, Johnson & Johnson, Merck & Co, NGM Biopharmaceuticals, Novartis Pharma AG, Regeneron, Aavantgarde; Grants — Genentech/Roche, Astellas, Annexon Biosciences, Novartis Pharma AG, Regeneron, AbbVie, Johnsons & Johnson, Aavantgarde, Ionis, Boehringer Ingelheim, Alexion; Honoraria — Genentech/Roche; Regeneron, Payment for expert testimony — Genentech/Roche; Patents — Gaming - AI, Ocular Drug Delivery Device; Participation on a Data Safety Monitoring Board — PYC Therapeutics, Ray Therapeutics, Adverum Biotechnologies, Aavantgarde Bio; Stock — Vinci Pharmaceuticals, Inc. Supported by Still Water Foundation. The funding organization had no role in the design or conduct of this research. Support for Open Access publication was provided by Retina Foundation of the Southwest. HUMAN SUBJECTS: Human subjects were included in this study. This study was conducted with the approval of the University of Texas Southwestern (Dallas, Tx) and WCG Institutional Review Board (Cary, NC). All research adhered to the tenets of the Declaration of Helsinki. All participants provided informed consent. No animal subjects were used in this study. Author Contributions: Conception and design: Csaky, Tandon Analysis and interpretation: Csaky, Tandon, Vasquez Data collection: Csaky, Tandon Obtained funding: Csaky Overall responsibility: Csaky, Tandon Supplementary Data Supplemental Table 1 References Associated Data Supplementary Materials Supplemental Table 1