Visual Sensitivity Loss in Geographic Atrophy: Structure–Function Evaluation Using Defect-Mapping Microperimetry

Abstract

Purpose: To examine the structure–function relationship in eyes with geographic atrophy (GA) using defect-mapping microperimetry, a testing strategy optimized to quantify the spatial extent of deep visual sensitivity losses.

Methods: Fifty participants with GA underwent defect-mapping microperimetry testing of the central 8°-radius region (208 locations tested once with a 10-decibel stimuli) and fundus autofluorescence imaging in one eye. The GA extent in the corresponding central 8°-radius was derived by manual annotations and image co-registration to examine the global structure–function relationship. The distance of each test location from the GA margin was also derived, and regions defined, to examine the local structure–function relationship.

Results: GA extent in the central 8° explained a substantial proportion of variance in the percentage of locations missed (nonresponse) on microperimetry at the global level (R² = 0.90). At a local level, the probability of missing stimuli at the outer junctional zone (0–500 µm outside the GA margin) and GA margin (probability = 7% and 34%, respectively) was higher than at the outer nonlesional zone (>500 µm outside the GA margin; probability = 2%; P < 0.001 for both). The probability of missing stimuli at the inner junctional zone (0–250 µm inside the GA margin) was also lower than at the inner lesional zone (>250 µm inside the GA margin; probability = 64% and 88%; P < 0.001).

Conclusions: This study confirms the expected functional relevance of the region with GA on fundus autofluorescence imaging and underscores the potential effectiveness of defect-mapping microperimetry testing for capturing visual function changes when evaluating new GA treatments.

Geographic atrophy (GA) is a late stage of AMD characterized by the progressive loss of the RPE and accompanied by degeneration of the overlying photoreceptors.^1–5 Numerous clinical trials have been conducted or are underway that are aimed at slowing GA growth.⁶ Whilst the standard measure of visual acuity has often been used as an outcome measure in such trials,⁶ it is limited in its ability to comprehensively capture the functional impact of the GA lesions.^7–9 

Early studies using fundus-controlled perimetry (commonly termed “microperimetry”) have demonstrated that regions with GA are characterized by deep visual sensitivity losses.^10,11 Changes in visual sensitivity on microperimetry have thus been used for over a decade as an outcome measure in treatment trials aiming to slow GA progression,^12–16 and testing is generally performed using a thresholding strategy that allows the precise level of visual sensitivity (in decibels [dB]) to be determined at each location. Although threshold-based microperimetry testing can capture the decline in the level of visual sensitivity over time with GA progression, it is unfortunately limited in the number of test locations that can typically be assessed in a single test session when seeking to maintain test durations comparable with standard automated perimetry (approximately 6 minutes on average in eyes with visual field loss).¹⁷ This is because of the need to present multiple stimulus at each location when measuring visual sensitivity thresholds. 

It has been previously observed that although both the mean sensitivity on microperimetry worsened and GA area increased significantly over time, there was no significant association between these two parameters.¹⁸ Instead, a significant association was only found between the change in number of absolute scotomas on microperimetry and GA area.¹⁸ Note, however, that the GA area in this previous study was measured within a 30° × 30° region, and microperimetry testing was performed in a 20° diameter region. A recent study using data from the Chroma and Spectri phase III trials—measuring GA area and performing microperimetry in the same regions as the previous abovementioned study—also observed that there were only moderate correlations between mean sensitivity on threshold-based microperimetry testing and GA area at cross-section, but that there was a stronger correlation when evaluating the number of absolute scotomas with GA area.¹⁵ However, there were only weak correlations between the change in either of these visual function measures and GA area growth,¹⁵ and these above findings highlight how conventional threshold-based microperimetry might be limited as an outcome measure in GA treatment trials aiming to slow disease progression. This was indeed the case in the recent phase III DERBY and OAKS trials of intravitreal pegcetacoplan, where there was no significant difference in the prespecified analyses of the change in mean sensitivity from baseline at 24 months on threshold-based microperimetry between the treated and sham groups, despite a significant decrease in GA lesion growth rate with treatment.¹⁹ There was only some evidence of functional preservation in a post hoc analysis of the average sensitivity for test locations falling within the perilesional zone (a region that encompassed 250 µm each within and outside the GA margin) in one of the two treatment arms.¹⁹ 

Given that progressive GA growth is expected to result in a corresponding enlargement of regions with deep visual sensitivity losses, a microperimetry testing strategy optimized to quantify the spatial extent of such deep visual sensitivity losses could provide a more effective means for capturing progressive visual function decline over time in eyes with GA. We previously explored this strategy, which we termed a “defect-mapping” approach, in a proof-of-principle study using the optic nerve head of healthy individuals as a model for deep scotomas.²⁰ This approach involves presenting a suprathreshold stimulus only once at each location, allowing testing to be performed with a much higher spatial density than is possible with standard threshold-based microperimetry. When compared with threshold-based testing, we observed that the defect-mapping strategy exhibited significantly lower levels of test–retest variability and captured a significantly greater extent of change relative to measurement variability with an increase in the size of the deep scotoma more effectively.²⁰ This was achieved whilst maintaining a slightly (but statistically significantly) shorter test duration, with the average test duration being 5.3 minutes with the defect-mapping strategy compared with 5.8 minutes for the threshold-based strategy.²⁰ 

The current study, thus, sought to prospectively assess the usefulness and reliability of a defect-mapping microperimetry test in eyes with GA with a particular focus on the structure–function association and test–retest repeatability. 

Methods

This study included individuals enrolled in a prospective observational study of atrophic AMD conducted at the Centre for Eye Research Australia. It was conducted in accordance with the International Conference on Harmonization Guidelines for Good Clinical Practice and with the tenets of the Declaration of Helsinki. Institutional review board approval was obtained for this study, and all participants provided written informed consent. 

Participants

The prospective observational study included individuals who were 50 years or older who had at least one eye with GA that was ≥175 µm in diameter on fundus autofluorescence (FAF) imaging (visualized as a region of definite decreased autofluorescence)²¹ and a best-corrected visual acuity of 20/100 or better. Eyes with evidence of neovascular AMD or that have undergone any prior treatments for AMD were excluded from this study. In addition, individuals with any ocular, systemic, or neurological condition(s) that could affect the reliable assessment of visual function or the retina were also excluded from this study. If both eyes were eligible, the eye with the greater extent of GA present was selected to undergo testing. 

Retinal Imaging

Macular-centered OCT volume scans covering a 20° × 20° region were acquired using the Spectralis HRA+OCT device (Heidelberg Engineering GmbH, Heidelberg, Germany). Scans consisted of 97 B-scans, all with 1024 A-scans and with automatic real-time averaging of 16 frames. FAF images were acquired using the same device and covered a 30° × 30° region centered on the macula, and with 100 frames averaged, using an excitation light source of 488 nm and emitted fluorescence signals between 500 and 700 nm were detected. Nonstereoscopic, 45° macular-centered CFPs were obtained using the Canon CR6-45NM device (Canon, Saitama, Japan). 

Microperimetry Testing

Microperimetry testing was performed using the Macular Integrity Assessment device (CenterVue, Padova, Italy) after pupillary dilation and before performing any study assessments that could bleach the retina or compromise the ocular surface. Fundus tracking on this device was performed using 36.5° × 36.5° near-infrared fundus images (with a central wavelength of 850 nm) acquired using a line-scanning laser ophthalmoscope at 25 frames per second. Perimetry testing was performed using achromatic Goldmann Size III (0.43° diameter, or approximately 124 µm at the retinal surface) stimuli against an achromatic background with a luminance of 1.27 cd/m². Stimuli can be presented with a luminance of between 1.35 and 318 cd/m², thereby providing a dynamic range of 36 dB of differential contrast. 

In this study, an assessment of the spatial extent of deep visual sensitivity abnormalities—or “defect-mapping” perimetry—was performed by presenting stimuli with a luminance of 32.9 cd/m² (equivalent to a 10-dB stimulus based on the dynamic range of the device) once at each location. This stimulus luminance was chosen because it corresponds with the floor of the effective dynamic range of the device, which is defined as the stimulus luminance where 5% of the retest values include 0 dB (i.e., deemed statistically indistinguishable from the physical floor of the dynamic range of the device, which is the brightest stimuli it can present, based on measurements from a single test).²² Testing was performed using an isotropic stimulus pattern consisting of 208 locations within the central 8° radius (as shown in Fig. 1), to provide an interstimulus interval of 1° and to ensure that test duration was maintained at approximately 6 minutes. Testing was performed with a red central fixation target that was 1° in diameter, and when fixation was nonfoveal (e.g., owing to the presence of subfoveal GA), the stimulus pattern was manually centered on the location of the anatomical fovea estimated using the OCT volume scans. 

Figure 1.

 

(A) FAF image of an eye with GA, and (B) with areas of GA annotated (orange overlay). (C) Stimulus pattern used on defect-mapping microperimetry testing in this study (yellow = seen; black = missed) depicted on the near-infrared fundus image. The near-infrared fundus image captured on microperimetry was then co-registered with the FAF image, as shown in (D). (E) Illustration of the categorization of test locations based on whether they fell on the GA margin, or entirely inside or outside an area of GA, and the distance from the GA margin for these points.

All participants included in this study were required to have completed at least two reliable defect-mapping microperimetry tests within a single session, to assess the intrasession test–retest repeatability of this technique. The reliability of each test was assessed using false-positive catch trials, based on presentations of a 10-dB stimuli to the optic nerve head. Any test with a greater than 25% false-positive rate was considered unreliable, discarded, and repeated when feasible. All tests after the first were performed using the follow-up function, where the same retinal locations assessed in the first test was evaluated on subsequent tests. 

A consecutive subset of participants in this study underwent testing with the stimulus luminance of the defect-mapping microperimetry testing set at 318 cd/m² (equivalent to a 0-dB stimulus) during a different study visit to compare the results of presenting stimuli at the floor of the physical dynamic range, compared with the effective dynamic range of the device.²² Their results were only included if two reliable tests were completed during that visit. This comparison was performed to understand if the structure–function correlations and the extent of deep visual sensitivity abnormalities detected differed based on the stimulus intensity used for defect-mapping microperimetry testing. 

Image Annotations and Processing

Areas of GA that were ≥175 µm in diameter on FAF imaging were annotated using a custom software (Cross-Modality Annotation Software, Ophthalmic Neuroscience Unit, Centre for Eye Research Australia)^23,24 by a grader (E.E.G.), which were then all reviewed together with a second senior grader (L.A.B.H.). Instances of disagreement were immediately resolved through open adjudication and the annotations were revised where necessary. Note therefore that small regions of hypoautofluorescence of less than 175 µm were not annotated, in a similar manner as performed in a recent study.²⁵ The near-infrared fundus image acquired with the microperimeter with the best image quality between the two tests performed at each visit was then manually co-registered with the FAF images. These steps are illustrated in Figure 1. Note that co-registration of the near-infrared and FAF images were performed using fiducial points manually placed at retinal vessel bifurcations, and then the two images were then registered automatically, as described previously²⁶ through radial transformations of each modality to account for radial distortions and affine transformation (including rotation, scaling, and translation) 

The percentage of the corresponding 8° radius region sampled by microperimetry that was occupied by GA was derived to enable an evaluation of the global structure–function relationship based on the percentage of locations missed (PLM) (locations with a nonseeing response to the single 10-dB stimuli presented) on defect-mapping microperimetry testing. Local structure–function relationships were examined at each test location by determining the percentage of overlap with GA in the 0.43° diameter region sampled by each stimulus, and the distance from the closest edge of the stimulus to the nearest GA margin. This information was then used to categorize each point as either falling on the GA margin, or entirely outside or inside an area of GA. Points falling entirely outside or inside the GA lesion were grouped into 125-µm bins based on their distance from the GA margin (0–125 µm, 125–250 µm, etc.), up to a maximum of 1000 µm and 500 µm for points outside and inside the GA lesion, respectively. This information was also used to categorize each test location more broadly as either falling within (i) the outer nonlesional zone (stimuli >500 µm outside the GA margin), (ii) outer junctional zone (stimuli entirely within the 0-500 µm area outside the GA margin), (iii) GA margin (stimuli overlapping with the GA margin), (iv) inner junctional zone (stimuli entirely within the 0- to 250-µm area inside the GA margin), and (v) inner lesional zone (stimuli >250 µm from the margin inside the GA margin). An illustration of these five categories is also shown in Figure 1. Points falling on the GA margin were grouped into 25% bins (>0%–25%, >25%–50%, etc.), based on their degree of overlap with GA lesion. 

Statistical Analyses

All participants who underwent defect-mapping microperimetry testing with the 10-dB stimuli were included in the analyses. The global structure–function association between the PLM on defect-mapping microperimetry and the proportion of the corresponding 8° radius region with GA was evaluated using a random intercepts linear regression model to account for the correlations between the two microperimetry tests from each participant. The coefficient of determination (R²) for this relationship was derived from an ordinary least squares linear regression model, with the CIs derived using a bootstrap resampling approach (n = 1000 resamples; only 1 out of the 2 tests from each participant was randomly included per resample). 

Local structure–function associations were evaluated by assessing the probability of missing stimuli based on their location described by the categories above for points entirely inside or outside the GA lesion in 125-µm bins, or by the degree of overlap with the GA lesion for points at the GA margin in 25% bins. These probabilities were derived using a logistic regression model with random intercepts at the person and microperimetry test levels, to account for the correlations between two tests of each participant and between locations within each test. The probability of missing stimuli based on their location in the five broader categories described above was also calculated using a similar approach. 

The test–retest repeatability of the PLM on defect-mapping microperimetry was determined by deriving its coefficient of repeatability (CoR) (where 95% of the test–retest difference is expected to lie). The presence of systematic changes in the PLM between the two tests was also evaluated using a random intercepts linear regression model, and the relationship between the test–retest difference and magnitude of its measurements was examined using Kendall's tau. The test–retest repeatability of the response at each test location relative to the GA area(s) was determined by calculating its mean percentage of agreement using a random intercepts linear regression model. 

These analyses were repeated when evaluating the subset of participants in this study who underwent defect-mapping microperimetry testing with the 10-dB and 0-dB stimuli on different visits, and an additional random intercept at the visit level was included to also account for the correlations between tests performed across the two visits with the different stimulus intensities. Comparisons between the findings based on stimulus intensity was also performed by specifying an additional interaction term based on stimulus intensity where appropriate. All analyses were conducted using Stata software version 18 (StataCorp, College Station, TX, USA). 

Results

A total of 50 participants were included in this study, and they were on average 76 ± 8 years old (range, 52–93 years old) and 32 participants (64%) were female. The median total GA area in the eyes that underwent microperimetry testing was 3.02 mm² (interquartile range [IQR], 1.62–4.84 mm²), and the median percentage of the total GA area that was within the central 8° radius region tested on microperimetry was 91% (IQR, 85%–93%). A total of 25 eyes (50%) had subfoveal GA. The median duration of the defect-mapping microperimetry tests was 6.0 minutes (IQR, 5.7–6.4 minutes). 

Relationship Between Visual Sensitivity and GA

At the global level, there was a significant association between the PLM on microperimetry and the percentage of the corresponding central 8° radius region with GA seen on FAF (P < 0.001; plotted in Fig. 2). A substantial proportion of variance in the PLM was explained by the GA extent in the central 8° radius (R² = 0.90; 95% confidence interval, 0.76–0.96).  

Figure 2.

 

Plot of the percentage of locations missed (PLM) on defect-mapping microperimetry testing (out of the 208 stimuli presented per test) against the percentage of the central 8° radius region with GA determined on FAF imaging.

The mean probability of missing a stimulus on defect-mapping microperimetry testing based on its location in relation to the GA lesion(s) is shown in Figure 3. This analysis revealed that the probability of missing stimuli between >0 µm and 500 µm outside of the GA margin was significantly higher than those >1000 µm outside (all P ≤ 0.005), but not for stimuli between 500 and 1000 µm of the GA margin (all P ≥ 0.082; total of 8 comparisons). It also showed that the probability of missing stimuli at the GA margin increased significantly with an increasing percentage of overlap with the GA lesion(s) (P < 0.001). Finally, it revealed that the probability of missing stimuli between >0 and 250 µm inside the GA margin was significantly lower than those >500 µm inside (all P ≤ 0.001), but the probability of missing stimuli between 250 and 500 µm of the GA margin was not significantly different from those >500 µm inside (all P ≥ 0.016; not statistically significant based on a total of 4 comparisons after Bonferroni correction for multiple testing). 

Figure 3.

 

Plot of the probability of missing the 10-dB stimuli based on its location in relation to the GA lesion(s). *P < 0.001 compared with stimuli whose closest edges were >1000 µm outside from the GA margin (“Ref. #1”). **P < 0.001 for the association between the probability of missing a stimulus and the percentage of overlap between the stimulus and GA. ^#P ≤ 0.001 compared with stimuli whose closest edges were >500 µm inside the GA margin (“Ref. #2”). An illustration is shown on the right to provide three examples of stimuli that fell in the extralesional region, at the GA margin, and within the GA lesion.

The mean probability of missing stimuli based on location in relation to the GA lesion(s) is also summarized in the Table. This illustrates the significant increase in the probability of missing stimuli at the outer junctional zone (stimuli within 0–500 µm outside the GA margin; mean probability = 7%) and GA margin (mean probability = 34%) when compared with the outer nonlesional zone (mean probability = 2%; P < 0.001 for both). It also illustrates the significantly lower probability of missing stimuli in the inner junctional zone (stimuli within 0–250 µm inside the GA margin; mean probability = 64%) than in the inner lesional zone (stimuli falling >250 µm inside the GA margin; mean probability = 88%; P < 0.001). 

Table.

 

Probability of Missing Stimuli and the Test–Retest Agreement in the Response on Defect-Mapping Microperimetry Testing Based on Its Location in Relation to the GA Lesion(s)

 

Test–Retest Repeatability of Defect-Mapping Microperimetry

There was no significant systematic change in the PLM between the first and second test (P = 0.101). There was also no significant relationship between the test–retest difference and the magnitude of its measurements (based on its average of the two tests; P = 0.146). The intrasession CoR of the PLM was 5.6% (95% confidence interval, 3.5%–7.8%), and the CoR did not differ significantly between eyes with and without subfoveal GA (P = 0.890). 

The repeatability of the response to stimuli based on location relative to the GA lesion(s) is also presented in the Table. These findings illustrate that repeatability was the lowest at the GA margin and inner junctional zone (mean agreement = 81% and 80%, respectively), followed by the inner lesional and outer junctional zones (mean agreement = 87% and 90%, respectively), and all these regions had lower levels of repeatability compared with the outer nonlesional zone (mean agreement = 97%; all P < 0.001). 

Impact of Stimulus Intensity

In a subset of 14 participants (28%) who underwent defect-mapping microperimetry testing with the 10-dB and 0-dB stimuli on different visits, similar analyses as performed as described elsewhere in this article were repeated in this cohort and described in detail in Supplementary Materials. In brief, the analyses showed that the proportion of variance accounted for by GA area in the central 8° radius was significantly greater for PLM derived from testing with the 10-dB compared with the 0-dB stimulus (R² = 0.89 and 0.75 respectively; P = 0.044). The probability of missing the 10-dB stimuli was also significantly higher at the outer junctional zone and the GA margin compared with the 0-dB stimuli (P ≤ 0.003 for both), but not for the outer nonlesional zone, inner junctional zone, and inner lesional zone (P ≥ 0.097). The intrasession test–retest repeatability of the PLM with the 10-dB and 0-dB stimulus was comparable (P = 0.648). 

Discussion

This study showed that there was a strong global structure–function relationship between deep visual sensitivity losses quantified by defect-mapping microperimetry and the extent of GA present in the area tested on FAF imaging. This finding was further supported by the finding that the probability of deep visual sensitivity loss at a given location was associated with the distance from, or degree of overlap with, the GA margin. These findings confirm the expected functional significance of the extent of these atrophic lesions and demonstrate how defect-mapping microperimetry could be an effective approach for capturing visual function changes when evaluating new treatments for GA. 

The findings from this study are broadly consistent with those reported by Sunness et al.¹¹ nearly three decades ago, who noted that 95% of dense scotomas on suprathreshold microperimetry testing also fell within atrophic regions when evaluating a cohort of 50 eyes with GA. The strength of the correlation observed between the PLM on defect-mapping microperimetry and GA area in this study was also higher than those observed with mean sensitivity on threshold-based microperimetry and GA area in the recent phase III Chroma and Spectri trials, which was reported as having a Spearman correlation coefficient of –0.42 to –0.46¹⁵ (the equivalent Spearman correlation coefficient between the PLM and total GA area in the current study was 0.90). The markedly higher level of structure–function correlation observed with our defect-mapping approach compared with conventional threshold-based microperimetry testing is likely accounted for by its greater efficiency of characterizing the spatial extent of deep visual sensitivity losses associated with GA, by testing each location only once with a suprathreshold stimuli to achieve a higher sampling density of testing. The efficiency of conventional threshold-based microperimetry testing in eyes with deep visual sensitivity losses would be hampered by the large degree of measurement variability typically present in areas with reduced visual sensitivities.^22,27 Interestingly, a recent computer simulation study of eyes with advanced glaucoma demonstrated that a suprathreshold adaptive mapping procedure using a higher density grid on standard automated perimetry could provide higher levels of accuracy and repeatability than a conventional 4–2 staircase strategy.²⁸ This goal was achieved when the test duration of both approaches were matched, and broadly supports the notion that our defect-mapping approach may be more effective and efficient at characterizing the spatial extent of deep visual sensitivity losses in eyes with GA. 

Our observations that the probability of missing stimuli on defect-mapping microperimetry was significantly decreased at the GA margin compared with the more distal nonlesional regions is also consistent with similar findings from threshold-based microperimetry testing in previous studies by Hartmann et al.²⁹ and by Hariri et al.³⁰ However, the study by Hariri et al. reported that the visual sensitivity of stimuli falling entirely within a junctional zone that was 0 to 500 µm outside the GA margin was not significantly different from more distal nonlesional regions, suggesting that visual sensitivity decreased precipitously at the GA margin. A study by Pfau et al.³¹ instead reported that visual sensitivity was significantly reduced for test locations between approximately 125 to 250 µm outside the GA margin, when compared with those further away, between approximately 375 and 875 µm. Finally, a study by Rinella et al.³² reported that visual sensitivity showed a strong correlation with the distance from the GA margin (Pearson's r = 0.79), with progressive decreases reported for locations between 0 to >5 mm outside the GA margin in 1-mm bins, although statistical testing for the difference between each bin was not performed. Our finding that the probability of missing stimuli on defect-mapping microperimetry testing was significantly higher for locations in the outer junctional zone (0–500 µm outside of the GA margin) compared with those in the outer nonlesional zone (>1000 µm away) falls generally within the range of the observations reported in the aforementioned studies.^30–32 

These findings are also in agreement with those from histopathological studies of eyes with GA, which indicate that photoreceptor loss and progressive RPE dysmorphia extends to within the perilesional region.^1,4,33 However, note that although the difference in the probability of missing stimuli that fell between the outer junctional zone and the more distal outer nonlesional zone was statistically significant, its magnitude was relatively small (7% and 2%, respectively) especially when compared with stimuli falling within the GA lesion (64% and 88% for the inner junctional and inner lesional zone, respectively). 

The findings that the probability of missing stimuli on defect-mapping microperimetry was significantly lower at the inner junctional zone (0–250 µm inside the GA margin) compared with the inner lesional zone (>250 µm inside the GA margin) is also consistent with the findings from another study by Pfau and colleagues,³⁴ who reported that test locations that were approximately 187 µm inside the GA margin showed a higher visual sensitivity than expected for an absolute scotoma, even when accounting for false-positive response rates. Our finding that the probability of missing stimuli within the inner lesional zone was 88%—rather than closer to 100%—is likely accounted for by both false-positive responses (as expected during perimetry testing) and the possibility that there are surviving photoreceptors within the atrophic regions.^1,2,4 

The test–retest repeatability of defect-mapping microperimetry testing at the global level for the PLM was slightly worse in this study (CoR = 5.6%) compared with our previous proof-in-principle study (CoR = 3.5%).²⁰ However, our previous study included a much younger cohort of healthy individuals than those with GA in this study (average age = 35 years old and 76 years old, respectively) and used a stimulus pattern with fewer test locations (n = 169 and 208, respectively). We also observed in this study that the repeatability of the point-wise response on defect-mapping microperimetry was significantly worse at the GA margin when compared with the outer nonlesional zone (average agreement = 81% and 96%, respectively). This observation was similar in previous studies at the border of the optic nerve head of healthy individuals³⁵ and at the border of degeneration in eyes with choroideremia³⁶ with threshold-based microperimetry testing. 

There was some evidence in a subset of the participants that underwent defect-mapping microperimetry with both the 10-dB and 0-dB stimuli that the former appeared to show a slightly stronger correlation with the extent of atrophic lesion(s) present. The probability of the 10-dB stimuli being missed at the GA margin, and at the inner and outer junctional zone, was also higher when compared with the 0-dB stimuli, as expected owing to the greater difficulty to detecting stimuli with a lower luminance in areas of retinal dysfunction. Nonetheless, both stimulus intensities are likely to be useful with defect-mapping microperimetry testing, but future studies are needed to examine whether the effectiveness of capturing progressive visual function loss differs based on the stimulus intensity used. 

These findings reinforce the expected functional relevance of the measurements of GA area on FAF imaging, which may be underestimated based on microperimetry testing using a thresholding strategy.^15,18 Treatments that effectively slow the rate of GA growth on FAF imaging—which is currently accepted as a clinical trial endpoint by regulatory authorities^37,38—should thus be expected to also slow the spatial enlargement of deep visual sensitivity losses. Nonetheless, regulatory authorities such as the European Medicines Agency have expressed the desire to see the beneficial effects of treatments for slowing GA progression supported by a similar effect on preserving function.³⁸ Our findings suggest that defect-mapping microperimetry testing would be a useful approach for capturing treatments effects on preserving visual function. Accordingly, we recommend this approach in clinical trials of treatments for GA, aiming to slow growth of the atrophic lesions. 

One of the main limitations of this study is its cross-sectional design. Future studies are needed to confirm whether defect-mapping microperimetry can capture progressive enlargement of deep visual sensitivity losses over time and whether such changes are strongly associated with the structural changes. Other limitations of this study relate to the errors inherent with the fundus tracking during microperimetry testing, the manual co-registration of the FAF and microperimetry images, and the delineation of the GA margin during the manual annotations of the FAF images. These errors may partly account for the greater probability of missing stimuli on defect-mapping microperimetry testing at the outer junctional zone and lower probability of missing stimuli at the inner junctional zone compared with the inner lesional zone. Nonetheless, the strengths of this study include using two microperimetry tests per session to evaluate the structure–function relationships, as well as the size of the cohort (n = 50 participants). 

In conclusion, this study found that there was a strong structure–function relationship between the extent of deep visual sensitivity losses on defect-mapping microperimetry and GA area on FAF imaging, providing supportive evidence for the expected functional relevance of this structural measure. The findings also underscore the potential effectiveness of defect-mapping microperimetry as a tool for monitoring visual function in GA treatment trials. 

Acknowledgments

Supported by the National Health & Medical Research Council of Australia (#2008382 [Z.W.] and #1194667 [R.H.G.]), the BrightFocus Foundation (#M2019073 [ZW and RHG]) and Apellis Pharmaceuticals (investigator-initiated trial grant [ZW and RHG]). CERA receives operational infrastructure support from the Victorian Government. The funders had no role in the manuscript writing and the decision to submit the manuscript for publication. 

Disclosure: Z. Wu, None; X. Hadoux, None; M. Jannaud, None; E.K. Glover, None; E.E. Gee, None; L.A.B. Hodgson, None; P. van Wijngaarden, reports personal fees from Roche/Genentech, Bayer, Novartis, and Mylan outside the submitted work; R. H. Guymer, reports personal fees from Roche/Genentech, Bayer, Novartis and Apellis outside the submitted work