Retinal Sensitivity in Macular Subfields and Their Association with Contrast Sensitivity in Early and Intermediate Age-Related Macular Degeneration

Best corrected visual acuity (BCVA) is the main outcome measure in many eye diseases to date. However, BCVA performs poorly at detecting functional loss in the early stages of AMD because it remains unaffected until more advanced stages of AMD are reached [1‒4].

Retinal sensitivity and contrast sensitivity are important to consider in the early stages of AMD. Even patients with relatively good visual acuity in early and intermediate stages of AMD complain of subjective visual impairment in dim light or contrast sensitivity [5‒7]. Consequently, there is a high interest for the identification of improved clinical markers that would enable earlier diagnosis and quantification of disease severity, progression of functional loss, and response to treatment in AMD.

In recent years, there has been a growing interest in the use of microperimetry for examination of retinal sensitivity. Studies using microperimetry in AMD were able to differentiate between healthy, early age-related macular degeneration (eAMD), and intermediate age-related macular degeneration (iAMD) based on group mean sensitivities (MSs) [8]. However, the evidence for individual categorization is questionable due to the lack of defined criteria for optimal differentiation between disease stages using microperimetry [3]. The natural history of AMD degeneration involves the central area. It was reported that lesion start begins with rods degenerating more than cones. This degeneration of rods appears to be most severely affected parafoveally between 3.5 and 10° [9]. It is reasonable to hypothesize that specific intraretinal degeneration patterns may influence retinal sensitivity changes observed in AMD patients. Only few studies have investigated retinal sensitivity using microperimetry in the form of comparing intraretinal subfields in AMD [10, 11].

Visual field model analysis (VFMA) is a volumetric presentation of microperimetric results, which may provide a more accurate assessment combining both extent and degree of sensitivity in retinal pathologies. This representation is expressed as total volume (VTOT) and may be valuable for patients in advanced stages of disease where the MS is near the lower end of the test capabilities and VFMA shows localized residual retinal function. So far, there are only few studies employing this new metric [12, 13].

Contrast sensitivity describes the minimum contrast required to separate objects from one another [14]. Visual contrast plays an important role in daily life, and studies have reported that contrast sensitivity perception may decline while visual acuity remains intact in the beginning of AMD [15, 16].

Known “traditional” methods measuring contrast sensitivity such as Pelli-Robson chart or conventional contrast sensitivity tests have limitations. As opposed to time-consuming conventional contrast sensitivity tests [17] or Pelli-Robson chart, which is limited by measurement at one spatial frequency and low test-retest reliability [18], the quantitative contrast sensitivity function (qCSF) overcomes these disadvantages by being performed in a reduced time of two to 5 min per eye with high sensitivity and consistent test-retest reliability [19, 20]. qCSF, introduced by Lesmes et al. [21] in 2010, describes a novel method to assess contrast sensitivity in patients. The test is based on a Bayesian active learning algorithm and estimates contrast sensitivity across multiple spatial frequencies and contrasts. This procedure has been computerized and commercialized in a clinical device, the Manifold Contrast Vision Meter (Adaptive Sensory Technology (AST), San Diego, CA, USA) [20]. As a novel metric for measuring contrast sensitivity function (CSF), only a few studies have used the qCSF method in the context of AMD so far [22‒25].

Therefore, microperimetry and qCSF may both be potential functional clinical markers in the evaluation of AMD. However, comparative studies between these two methods are currently lacking in the literature. It is therefore plausible to speculate that both parameters may decrease with AMD progression, suggesting a potential association between them at identical stages of the disease. The goal of this study arises from two research questions that we would like to address in relation to the use of microperimetry and qCSF as functional markers in AMD.

First, differences in retinal sensitivities are investigated in subfields. Second, we report the associations of the subfield retinal sensitivity with qCSF. Thereby, we hope to contribute new insights on retinal function in healthy controls and AMD patients.

This prospective cross-sectional observational study was conducted at the University Hospital Basel and the Institute of Molecular and Clinical Ophthalmology Basel (IOB) in collaboration with the Ecole Polytechnique Fédérale de Lausanne (EPFL). This study analyses baseline data of the longitudinal Multimodal Functional and Structural Visual System Characterization (MuMoVi) study.

The study protocol was approved by the Institutional Review Board “Ethikkommission Nordwest- und Zentralschweiz EKNZ” (Swiss Ethics Project-ID: 2021-00029) and adhered to the tenets of the Declaration of Helsinki, the principles of Good Clinical Practice (GCP), the Human Research Act (HRA), and the Human Research Ordinance (HRO).

Eligible patients with early to intermediate age-related macular degeneration and healthy subjects were recruited consecutively between June 2021 and June 2022 by the study team at the Department of Ophthalmology, University Hospital Basel, Switzerland. Inclusion criteria for participants in the AMD group were as follows: clinical diagnosis of early (medium drusen of 63 μm until 125 μm without pigmentary abnormalities) or iAMD (large drusen >125 µm and/or pigmentary abnormalities according to the classification system published by Ferris et al. [26]). For controls, the “non-dominant” eye was chosen. For the AMD groups, if both eyes were eligible, “dominant eye” was defined study eye. The rationale behind this methodology is to compare the “best” eye in AMD patients with the “worst” eye in controls (the non-dominant one). Controls were healthy eyes with no additional or coexisting ocular disease other than refractive error or previous cataract surgery. Other requirements included age ≥50 years and ability to perform ophthalmic and psychophysical examinations. Exclusion criteria for both AMD and control subjects were coexisting ocular disease affecting visual function or ocular morphology, inadequate pupil dilation for imaging and functional testing, and the presence of cataract sufficient to interfere with retinal imaging. Written informed consent was obtained from all patients in the study.

Study visits included recording of medical history, subjective refraction for BCVA using Early Treatment Diabetic Retinopathy Study (ETDRS) charts, ophthalmic slit-lamp examination, microperimetry testing (MAIA, CenterVue, Padova, Italy), and qCSF (Adaptive Sensory Technology (AST), San Diego, USA). The participants performed examinations, except for microperimetry, with the best refraction. Visual function testing was conducted in the dark.

Before the full test, a trial run was performed to demonstrate the principles of the MAIA exam. Application and settings of MAIA have been described elsewhere [27, 28]. The mesopic testing was performed without prior dark adaption. A customized stimulus grid, the PINNACLE standard grid, consisting of 24 points, was used. These points were located at the center of the fovea and covered a circle of 10° [13]. In a Cartesian coordinate system, the test points are situated at positions of −5°, −3°, −1°, 1°, 3°, and 5° along both the x-axis and the y-axis. MSs in dB and volumetric representations, as “VTOT” for total volume, for all 24 points and different subfields (see Fig. 1) were calculated [12, 29]. The result obtained from VFMA through an integration process is referred to as the volume beneath the surface in dB-sr.

qCSF was measured in darkness on a large LED screen 1,920 × 1,080 pixels (46″) from Manifold Contrast Vision Meter (Adaptive Sensory Technology (AST), San Diego, CA, USA). Subjects were seated 3 m in front of the device with the BCVA. Study eyes were tested monocularly. A series of 25 examinations was performed. The stimuli presented were filtered Sloan letters within 19 possible spatial frequencies (1.19–30.95 cycles per degree [CPD]) and 128 possible contrasts at a luminance of 95.4 cd/m² [30].

The following variables were exported from the qCSF system: the area underneath the logarithmic contrast sensitivity function in logCS; the contrast acuity in logCPD, defined as the intersection point on the x-axis with lowest contrast sensitivity and highest resolvable spatial frequency of the CSF curve; and the contrast sensitivity values of the spatial frequencies at 1CPD, 1.5CPD, 3CPD, 6CPD, 12CPD, 18CPD.

Statistical analysis for this study was conducted with Microsoft Excel (Version 16.80), GraphPad Prism (Version 9.5.1), and the R software environment (Version 2022.12.0+353) with additional packages lme4 and sjPlot. The data were examined for normal distribution using the D’Agostino and Pearson test, and either parametric or non-parametric statistics were applied as appropriate. Accordingly, cohort characteristics were compared using ANOVA or Friedman test and post hoc multiple comparison tests.

In this study, multiple multivariable linear regression models were constructed for analysis. These models consistently used various functional metrics as dependent variables. Variables of interest were introduced as fixed effects. Specifically, age was found to have a significant association with all the tested functional metrics in this dataset. As a result, age was included in the linear regression models to account for its impact on the dependent variables. It was considered that the effect of age on the dependent variables in this case could be generalized to other cohorts; hence, age was introduced as a fixed effect. Hypothesis tests were carried out with a significance level of 5% (0.05) (ns, p > 0.05; *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001). Mean and standard deviation were presented for data that followed a normal distribution, while median and interquartile range were reported for non-normally distributed data.

This study featured baseline data from the MuMoVi study, for which 99 subjects with one study eye each were recruited. For this study, 7 participants were excluded due to incomplete data or advanced AMD stage, resulting in a total of 92 study eyes included in the final analysis. The control group consisted of 48 eyes, while the 44 eyes with AMD were further subdivided into 13 eAMD and 31 iAMD eyes (Table 1). A significant age difference (p = 0.004) was found between control and AMD, while eAMD and iAMD showed no significant difference (p = 0.6). Due to the significant correlation between all functional measures tested and age (p < 0.001), age was defined as a fixed effect for the following regression analyses. The control group comprised 62.5% females, whereas the AMD group had an even sex distribution. The majority of participants identified as white (98.9%) except for one participant. Most study eyes were phakic in control (89.6%), eAMD (92.3%), and iAMD (61.3%). Univariate linear regression analysis revealed no significant correlation between lens status and any of the functional measures tested.

As shown in Figure 1a, only the control group showed a significant difference (p < 0.0001) in retinal sensitivity of the superior and inferior meridian. The superior hemifield was observed to have higher retinal sensitivity than the inferior hemifield. For eAMD and iAMD, there was no significant difference between the superior and inferior halves, although there was a nonsignificant trend toward lower retinal sensitivities in the superior half for iAMD (p = 0.237).

When comparing retinal sensitivities between the nasal and temporal halves, no significant differences were identified for control, eAMD, and iAMD (Fig. 1b). However, there was a trend toward higher retinal sensitivities in the nasal half compared to the temporal half for both controls (p = 0.08) and iAMD (p = 0.22).

Retinal sensitivities were significantly higher in the central subfield compared to the peripheral subfield in controls (p < 0.0001) and eAMD (p < 0.001) groups (Fig. 1c). In iAMD, however, this was not significantly measurable.

There was a significant difference between all four quadrants (p < 0.0001) for the control group. Further analysis revealed the significance of the superior nasal quadrant compared to the inferior nasal quadrant (p < 0.001) and inferior temporal quadrant (p < 0.001). A significant difference was found between superior temporal and inferior temporal quadrants (p < 0.05). In the control group, the highest retinal sensitivity was observed in the superior nasal quadrant. In eAMD, no significance was observed between the four quadrants. In the iAMD group, a significant difference over all four quadrants (p < 0.05) was found, and the highest sensitivity was observed in the superior nasal quadrant.

As shown in Figure 1e, the analysis revealed overall strong significant differences in the control group for all five subfields (p < 0.0001). Retinal sensitivity was highest in the central square with the significant differences to all but the nasal square. Overall, the differences between the eAMD squares were statistically significant (p < 0.05). The central square showed the highest retinal sensitivity, which was significantly higher than retinal sensitivity in the inferior square. For iAMD, a significant difference was observed overall (p < 0.05). However, the nasal cube showed the highest retinal sensitivity. No significant differences were found in post hoc multiple comparisons.

The association of qCSF metrics with retinal sensitivity was determined by multivariate linear regression analysis for control, eAMD, and iAMD. Age was set as fixed effect. In general, there was a trend toward stronger associations of retinal sensitivity with contrast sensitivity further along the disease trajectory (Fig. 2). Overall associations were weakest in the control group, stronger in the eAMD group, and strongest in the iAMD group as can be seen color-coded by effect size in Figure 2. Area underneath the logarithmic contrast sensitivity function and CSF acuity showed strong associations with retinal sensitivity in all groups. CS at 3 CPD and CS at 6 CPD showed strong associations with retinal sensitivity in the control and eAMD groups, whereas in the iAMD group, CS at 12 CPD displayed a strong association with retinal sensitivity. In all groups, CS at 1 CPD and CS at 1.5 CPD showed weak associations with retinal sensitivity.

In the control group, strong associations for MS were observed for nasal microperimetry subfields with qCSF metrics (Table 2). MS in temporal subfields showed weaker associations with qCSF metrics (Table 2). In the volumetric representation of retinal sensitivity, no significant association of retinal sensitivity with qCSF metrics was found in the control group.

The combined AMD group (eAMD and iAMD) revealed significant associations of nasal and temporal MS and volumetric representation with qCSF metrics (Table 2). No significant association was found between fixation stability and qCSF metrics in any of the group.

Although the potential of microperimetry to quantify retinal sensitivity in eAMD and iAMD has been demonstrated [2, 8, 31‒35], only few publications comparing retinal sensitivity between subfields in controls and AMD exist. This may be due to the lack of a standardized test array or defined test procedures for microperimetry in clinical practice and research [36].

For healthy controls, our results showed a significantly higher sensitivity of the superior hemifield compared to the inferior hemifield and of the central subfield compared to the periphery. These conditions are in line with other studies. Pfau et al. [10] reported significantly higher retinal sensitivity within the central 1° and higher retinal sensitivity in the superior hemifield compared to the inferior hemifield.

With respect to retinal sensitivity in the AMD disease trajectory, we showed that the nasal macula showed higher sensitivity in iAMD than other macular subfields. This trend was only evident in iAMD, while controls and eAMD probands showed highest retinal sensitivity in the central subfield. Curcio et al. [37, 38] have described a nasal cone streak, which features a greater cone density in the nasal compared to the temporal macula. Also, Curcio et al. [37, 38] reported that cones are initially less vulnerable in the AMD disease trajectory. Possibly the resilience of cones in the nasal cone streak contributes to our finding of relatively higher retinal sensitivity in the nasal macula compared to early disease stages or healthy controls. A study by Trinh and colleagues [11] reported that in iAMD functional impairment in retinal sensitivity was mostly biased toward central macula and superior quadrant compared to normal controls. Although the different grating complicates comparison, a trend of relative decrease in central retinal sensitivity from control to iAMD has been identified. Our data further provide overall pronounced differences of retinal sensitivities between subfields for the control group. In general, the eAMD and iAMD exhibit less pronounced differences in multiple comparison. Smaller sample size might contribute to this finding; however, we also hypothesize that progressing AMD levels out physiologic differences in the macular retinal sensitivity.

Some studies have investigated the CSF as a surrogate endpoint for AMD [39‒45]. In a previous study, we showed that contrast sensitivity at specific spatial frequencies as determined by the qCSF technology can detect functional differences between eAMD and iAMD, where VA metrics cannot [23].

In this study, we calculate the associations of retinal sensitivity with contrast sensitivity in order to gain a better understanding of the interaction between the two metrics of retinal sensitivity and contrast sensitivity as measured by the qCSF method. To the best of our knowledge, this is the first study looking into association of retinal sensitivity with contrast sensitivity metrics acquired by the novel qCSF method.

Overall, associations are weaker in controls and increase in eAMD and iAMD. This could be due to greater spread of values in eAMD and iAMD. Therefore, in these groups, a specific parameter can explain more variability of other values.

In specific, in the present study, the patterns of association seem to change between subgroups. In the control group, there is a strong association of CS at 3 CPD with retinal sensitivity. In the eAMD group, there is a strong association of CS at 6 CPD with retinal sensitivity. And in the iAMD group, there is a strong association of CS at 12 CPD with retinal sensitivity. In summary, there is a shift of association of retinal sensitivity with smaller optotypes with progressive AMD-related retinal changes. This finding suggests that variability in retinal sensitivity in healthy controls is best explained by CS for medium-sized optotypes, and for eAMD and iAMD, it is best explained by CS for small-sized optotypes. To our knowledge, such shift in associations has not been described so far. A possible explanation for this could be that location of structural damage in the AMD disease trajectory has an effect on the ability to recognized contrast grated optotypes at certain image sizes on the retina. According to this, the image size for optotypes presented at 12 CPD (6 CPD) might best explain variability in retinal sensitivity in iAMD (eAMD). These findings might correspond to the observations of Curcio and colleagues [37], who found that age-related rod loss starts in the parafovea and that also functional decline begins in the parafovea before advancing to the visual center. In another study, Ooto et al. [46] described the association between reduced parafoveal retinal sensitivity and reduced contrast sensitivity. Early incapacities in recognition of medium-sized optotypes with progression to smaller optotypes might reflect this disease trajectory.

Moreover, in healthy controls, our data provide significant associations of nasal retinal sensitivity with CS metrics, whereas temporal retinal sensitivity is predominantly not significantly associated with CS metrics. As referenced above, the nasal macula comprises a higher density of cones than the temporal macula. A strong contribution of these nasal cones to contrast sensitivity might explain its distinct association to retinal sensitivity in this region. Himmelberg et al. explored the evidence of regional differences, specifically highlighting asymmetries in contrast sensitivity perception in visual field. Asymmetries in perception following the phenomenon of horizontal vertical anisotropy were confirmed [47]. However, regional nasal and temporal differences were not addressed.

We also found that the regional differences in associations are more pronounced in the MS representation of retinal sensitivity, whereas the volumetric representation does not show significant associations of nasal retinal sensitivity to qCSF metrics. As described before, the volumetric might be advantageous in the representation of inhomogeneous visual field results as in end-stage disease [13]. However, the volumetric representation mitigates the effect of particularly heterogeneous perimetric values, which would serve to explain variability in associations.

There are several limitations to this study. There is a significant difference in age between the healthy control and AMD group. As contrast sensitivity showed a significant association with age, age was included as a fixed effect in our mixed model analysis. Additionally, AMD patients exhibit a broader spectrum of visual function measures than healthy controls, limiting the possibility for correlation in the control group. Second, given the descriptive nature of this cross-sectional study, it lacks the predictive power of longitudinal studies. Further downsides are the limited number of subjects in our eAMD group and the fact that our cohort consisted almost exclusively of white participants, making generalization to other ethnicities difficult.

In summary, our study contributes to the understanding of retinal functional capacities, which is essential in evaluating disease and in planning targeted therapies. For example, our data provide localized differences in retinal sensitivities as evaluated in subfields. It is important to recognize that retinal sensitivity both in healthy controls and also in AMD is not evenly distributed throughout the retina. This cross-sectional study also suggests a different trajectory of loss of retinal sensitivity throughout the retina, which would have implications to the application of localized treatments. It will be important to investigate these localized differences longitudinally as well. Finally, we report insights on the association of retinal sensitivity with contrast sensitivity as measured with the novel qCSF method.

The study protocol was approved by the Institutional Review Board “Ethikkommission Nordwest- und Zentralschweiz EKNZ” (Swiss Ethics Project-ID: 2021-00029) and adhered to the tenets of the Declaration of Helsinki, the principles of Good Clinical Practice (GCP), the Human Research Act (HRA), and the Human Research Ordinance (HRO). Written informed consent was obtained from all patients in the study.

Philipp Anders is supported by Bayer (F). Maximilian Pfau is a consultant for Apellis Pharmaceuticals (C), Johnson & Johnson (C), and is supported by Apellis Pharmaceuticals (F) and Novartis (R). Kristina Pfau is a consultant for Daiichi Sankyo (C) and is supported by Heidelberg Engineering (R) and Bayer (R). Hanna Camenzind Zuche is a consultant for Apellis Pharmaceuticals (C) and is supported by Bayer (R). Hendrik P.N. Scholl is a consultant for Astellas Pharma Global Development, Inc./Astellas Institute for Regenerative Medicine (C), Belite Bio (C), Boehringer Ingelheim Pharma GmbH & Co (C), Gerson Lehrman Group (C), Guidepoint Global, LLC (C), Janssen Research & Development, LLC (Johnson & Johnson) (C), Novo Nordisk (C), ReNeuron Group Plc/Ora Inc. (C), ReVision Therapeutics, Inc. (C), Stargazer Pharmaceuticals, Inc. (C), Tenpoint Therapeutics Ltd. (C), Third Rock Ventures, LLC (C), and is supported by Kinarus AG (F), Okuvision GmbH (F), and Novartis Pharma AG (F). All other authors have no conflicts of interest to declare.

H.S. is supported by the Swiss National Science Foundation (Project funding: “Developing novel outcomes for clinical trials in Stargardt disease using structure/function relationship and deep learning” #310030_201165, and National Center of Competence in Research Molecular Systems Engineering: “NCCR MSE: Molecular Systems Engineering (phase II)” #51NF40-182895), the Wellcome Trust (PINNACLE study), and the Foundation Fighting Blindness Clinical Research Institute (ProgStar study). P.A. is supported by the foundation “Freiwillige Akademische Gesellschaft Basel” and the foundation “OPOS zugunsten von Wahrnehmungsbehinderten” (Project funding: “Validation of novel microperimetry metrics as functional outcome parameters for clinical trials targeting macular diseases”). The funding organizations had no role in the design, data collection, data analysis, and reporting of this study.

Eric J. Chan and Philipp Anders: conceptualization, data curation, formal analysis, investigation, methodology, project administration, validation, writing – original draft, and writing – review and editing; Simona A. Garobbio: formal analysis, methodology, and writing – review and editing; Ursula Hall: data curation, project administration, and writing – review and editing; Chrysoula Gabrani: data curation, investigation, and writing – review and editing; Kristina Pfau, Hanna Camenzind Zuche, Stefan Futterknecht, and Maximilian Pfau: data curation, formal analysis, and writing – review and editing; Michael Herzog: conceptualization, validation, and writing – review and editing; Ghislaine L. Traber and Hendrik P.N. Scholl: conceptualization, validation, supervision, and writing – review and editing.

Eric J. Chan and Philipp Anders contributed equally to this work.Ghislaine L. Traber and Hendrik P.N. Scholl share last authorship.

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.