Association of Macular Thickness With Age and Age-Related Macular Degeneration in the Carotenoids in Age-Related Eye Disease Study 2 (CAREDS2), An Ancillary Study of the Women's Health Initiative

Purpose To evaluate the relationship of retinal layer thickness with age and age-related macular degeneration (AMD) in the Carotenoids in Age-Related Eye Disease Study 2. Methods Total retinal thickness within the macular area, and individual layer thickness was determined for CAREDS2 participants (n = 906 eyes, 473 women) from the Women's Health Initiative using Heidelberg optical coherence tomography (OCT). Mean measurements within the OCT grid were compared across age tertiles (69–78, 78–83, and 83–101 years) and AMD outcomes. Results Mean retinal thickness in the central circle, inner ring, and outer ring were 277 ± 34 µm, 326 ± 20 µm, and 282 ± 15 µm, respectively. Thickness did not vary by age in the central circle, but decreased with age in the inner and outer circles (P ≤ 0.004). Specifically, ganglion cell (GCL), inner plexiform, and outer nuclear (ONL) layer thickness decreased with age (P ≤ 0.003). Age-adjusted retinal thickness in all three circles did not vary by AMD outcomes (486 without AMD and 413 with AMD). However, individual layers showed changes with GCL and photoreceptor thinning and retinal pigment epithelial thicknening in eyes with late AMD. After controlling for age and AMD, higher ONL thickness was associated with better visual acuity. Conclusions In this cohort of older women, a decrease in perifoveal thickness was associated with increasing age, particularly in the inner retinal layers. Variabilty in thickness in AMD eyes was primarily due to outer retinal layers. Among all retinal layers, the ONL plays an important role in preserving visual acuity. Translational Relevance The study provides a deeper understanding of age related changes to the retinal layers and their effect on visual loss.


Introduction
Optical coherence tomography (OCT) provides noninvasive in vivo cross-sectional imaging of the retina. Spectral domain OCT (SD-OCT) devices permit the demarcation of individual retinal layers at higher acquisition speeds and image resolution from previous devices. 1 In addition, SD-OCT enables automated segmentation and quantification of individual retinal layers through the application of various software and segmentation algorithms. Previous studies have demonstrated that SD-OCT segmentation software produces reproducible thickness measurements. 2 SD-OCT is used extensively in clinical practice and clinical trials to measure the thickness of the retina and identify anatomic changes that occur with diseases, including those related to aging. For example, in age-related macular degeneration (AMD), SD-OCT segmentation has been used to monitor drusen, choroidal neovascularization, and geographic atrophy. 3 SD-OCT segmentation is also used to identify thinning of the nerve fiber layer in neurodegenerative diseases, such as Alzheimer's disease. 4 The description of SD-OCT changes in diseases of aging have been constrained by small sample sizes of populations younger than 75 years and an overrepresentation of severe disease. In addition, normative thickness data from proprietary SD-OCT manufacturers are not publicly available.
Previous studies that have used SD-OCT segmentation to evaluate the distribution of retinal layer thickness measurements across population subgroups were vital to understanding the epidemiological associations and pathophysiological mechanisms of disease, as well as to define normal ranges. [5][6][7][8][9][10][11][12] However, these studies have been limited to population studies without retinal diseases. We used data from the Carotenoids in Age-Related Eye Disease Study 2 (CAREDS2), a unique, and well-characterized cohort of women aged 69 to 101 years at the time of their participation in 2016 to 2019, to provide the macular thickness measurements and to study the relationship with age and AMD. We also explored relationships between the thickness of individual retinal layers and best-corrected visual acuity. CAREDS2 is an ancillary study of the Women's Health Initiative Observation Study (WHI-OS).

Participants
The details of the CAREDS study population has been previously published. 13 In summary, the WHI-OS, a prospective cohort study at 40 sites throughout the United States investigating the most common causes of morbidity and mortality among postmenopausal women aged 50 to 79 years (n = 3143). 14 Eligibility criteria included having high or low dietary lutein plus zeaxanthin intake levels (>78th or <28th percentiles) at WHI enrollment (1994)(1995)(1996)(1997)(1998) , of which 487 participated in study visits inperson. The analysis dataset for this project includes 473 women who completed SD-OCT scans. The age of the participants at the time of OCT assessment in CAREDS2 ranged from 69 to 101 years. The study was approved by the institutional review boards associated with each center. All participants provided written informed consent, and the study adhered to the tenets of the Declaration of Helsinki.

SD-OCT Imaging
SD-OCT scans were acquired by certified photographers with a Heidelberg Spectralis (Heidelberg Engineering, Heidelberg, Germany) OCT machine following the CAREDS2 reading center (Fundus Photograph Reading Center, University of Wisconsin-Madison) approved protocol. Volume scans centered on the macula were obtained from both eyes for all study participants. The macular scan protocol includes a 20°× 20°volume scan at high speed with 97 B scans, interscan distance of 60 μm and an image averaging of five frames.

Retinal Layer Segmentation
Retinal layers were automatically segmented using Heidelberg Spectralis software (version 1.9.13.0). Layers were defined using the boundaries generated by the software. The layers, including internal limiting membrane (ILM), retinal nerve fiber layer, ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer, outer plexiform layer, outer nuclear layer (ONL), photoreceptor layer (PR), and retinal pigment epithelium (RPE), are shown in Figure 1. The boundaries of the total retina were defined as ILM to BM. The total retina was divided into the inner and outer retinal layers by the software. The inner retinal layers were defined as ILM to external limiting membrane and the outer retinal layers as external limiting membrane to BM, respectively. The segmentation was reviewed throughout the entire Early Treatment in Diabetic Retinopathy Study (ETDRS) grid and segmentation errors were manually edited within the Heidelberg Spectralis software in all scans as needed. Segmentation review and editing was performed by masked, study certified graders at the CAREDS2 reading center. A total of 946 SD-OCT images from 473 CAREDS2 study participants were available for segmentation. ETDRS grid centration was assessed and scans with additional artifacts, such as z-offset (n = 3) and B-scan inversion (n = 2) were excluded. Additional SD-OCT scans were excluded because of grossly inaccurate segmentation throughout all retinal layers beyond that correctable by manual editing (n = 2). The remainder of excluded SD-OCT scans were due to confounding pathology, including macular hole (n = 8), significant intraretinal or subretinal fluid (n = 4), and severe macular degeneration (n = 21) prohibiting the identification of individual retinal layers.
Thickness maps were generated within Heidelberg Spectralis software, providing mean macular thickness (μm) and volume (mm 3 ) measurements for the nine sectors of the ETDRS grid, which were combined to form three concentric circles, including the central circle (1 mm diameter), inner ring (3 mm diameter), and outer ring (6 mm diameter) as displayed in Figure 1.

AMD Assessment
AMD presence and severity at CAREDS baseline (2001)(2002)(2003)(2004) was evaluated using 30°stereoscopic fundus photographs and at CAREDS2 follow-up (2016-2019) using 30°stereoscopic digital images. Of the 487 CAREDS2 participants, AMD status was determined from fundus photographs in 470 (96.5%) participants. AMD status of an additional five participants was determined from provider reports or medical records. Three field stereoscopic color photographs were evaluated for presence and severity of AMD using the Age-related Eye Disease Study (AREDS) severity scale. 15 Based on CAREDS2 protocol defined outcomes, we grouped the original AREDS 12-step AMD scale into early AMD (AREDS levels 1-5), intermediate AMD (AREDS levels 6-8) and late AMD (AREDS levels 9-11b). Late AMD was further classified as geographic atrophy and neovascular AMD.

Best Corrected Visual Acuity
Best corrected visual acuity (BCVA) was measured using the standardized ETDRS protocol modified for the AREDS trials. 16

Statistical Analysis
Macular thickness and volume measurements of the central circle (1 mm diameter), inner ring (3 mm diameter), and outer ring (6 mm diameter) were expressed as means ± standard deviation (SD) or least squares mean ± standard error (SE). Age was categorized into tertiles, and a P value for continuous trend over years was computed. AMD outcomes were expressed as a categorical variable, including no AMD, early AMD, intermediate AMD, and late AMD. We investigated the association between mean macular thickness and age tertile, as well as mean macular thickness and AMD outcomes, adjusted for age. Generalized estimating equations were used, which enabled the use of scans from both study eyes of each participant. Empirical standard error estimates were used for generalized estimating equations parameter estimation considering within person correlation. A two-tailed P value < 0.05 was considered significant. We also examined the ageadjusted associations between BCVA and individual layer thickness in the central circle, including sensitivity analyses, excluding participants with cataract and late AMD. Statistical analysis was performed using SAS software, version 9.4 (SAS Institute Inc., Cary, NC, USA).

Macular Thickness by Age
Overall, 946 SD-OCT scans of both eyes of 473 CAREDS2 participants were available for the current analysis. After excluding 40 scans, 906 scans were evaluated as presented in Table 1. Mean (SD) macular thickness overall, and in individual retinal layers are described by age tertile, within each of the nine ETDRS subfields ( Fig. 2A and Table 2). In the central circle, we did not observe significant variations in the total retinal thickness by age. However, in the analysis of separate layers, the IPL, ONL, and PR layers were significantly thinner with age. In the more peripheral inner and outer rings, the overall thickness of the retina and some specific layers (GCL, IPL, and ONL) also significantly decreased with age.

Macular Thickness by AMD Presence and Severity
The thickness of the total retina was similar in eyes with and without AMD in the central circle, as well as inner and outer rings ( Fig. 2B and Table 3). Eyes with AMD were further classified into early AMD (n = 205), intermediate AMD (n = 185), and late AMD (n = 23). Of the 23 eyes with late AMD, five (22%) had geographic atrophy (both central and noncentral) and 18 (78%) had neovascular AMD. The thickness of the total retina did not vary by AMD severity in the central circle, as well as inner and outer rings (P = 0.35, P = 0.26, and P = 0.51). The thickness of the inner retinal layers was also similar across AMD outcomes in the central circle and outer ring (P = 0.11 and P = 0.25) but decreased significantly with increasing AMD severity in the inner ring (P = 0.003). The thickness of the outer retinal layers decreased significantly with increasing AMD severity in the central circle, as well as inner and outer ring (P = 0.009, P = 0.002, and P = 0.009). In the analysis of the individual retinal layers, the thickness of the RPE increased with AMD severity within the central circle, as well as inner and outer rings (P = 0.001, P = 0.0005, and P = 0.002). These results were likely due to 78% of late AMD eyes being neovascular compared to 22% having geographic atrophy. The PR layer in the central circle, but not the inner and outer rings, was thinner in eyes with AMD compared to those without AMD (P = 0.0008). The thickness of the GCL decreased by AMD severity in the inner ring (P = 0.05).

Discussion
The histologic correlates of OCT images have been studied extensively, [17][18][19] and results of histological [20][21][22] and OCT studies [23][24][25] indicate that many regions of the macula thin with age and neurodegenerative diseases. In the current analysis, we provide normative data on the macular thickness of women age 69 to 101 years and describe the variations in macular thickness across age and AMD outcomes. Results of this report corroborate other population-based studies demonstrating reduced retinal thickness with age, particularly outside the central circle. [5][6][7][8]26 The difference in outer retinal thickness between AMD and non-AMD eyes was expected considering AMD is a disease primarily affecting the RPE. However, the difference in thickness between no AMD and early AMD was not discernible, which makes this measurement inadequate for identifying the transition. Outer retinal quantification in advanced AMD is an important biomarker for disease monitoring. 27,28 Our data showed increasing outer retinal thickness in all three circles from early to intermediate AMD indicating the role of quantitative monitoring of disease severity.
The positive association between the thickness of the ONL and BCVA, adjusted for age, is noteworthy. Previously, individual retinal layers and impact on visual acuity has mostly been evaluation in relation to epiretinal membrane surgery or other retinal diseases such diabetes and glaucoma. [29][30][31] Histopathologic studies have shown degeneration of ONL with loss of nuclei due to aging. 32 The current study shows a thinning of ONL with age and that the age-adjusted association between ONL and BCVA persists after excluding participants with cataract, indicating that vision is preserved in those with delayed aging effects  on retinal layers. This is explained, only in part, by the presence late AMD in some women. Vital considerations when evaluating retinal thickness include the accuracy of the SD-OCT segmentation algorithms applied and the boundaries of the individual retinal layers used. 33 Our study used the Heidelberg Spectralis SD-OCT device and its proprietary segmentation software. We noted errors in the automated segmentation that required manual correction throughout most scans (>80%). Errors in boundary lines, particularly ILM can occur because of commonly seen age-related pathologies, such as epiretinal membranes and thickened vitreous. The graders ensured that the line followed the ILM and made corrections as needed. The boundaries for Heidelberg's segmentation software differ from other devices and algorithms, such as those performed by Cirrus (Carl Zeiss Meditex, Dublin, CA, USA). 34 These discrepancies likely contribute to the variation in measurements across population-based studies, even among similar population subgroups. For example, despite a similar cohort of elderly Caucasian individuals, our results differed slightly from those obtained by the Beaver Dam Eye Study, which used Topcon 3D-OCT software. 7 Other factors that likely contribute to differences in retinal thickness measurements across populationbased studies include the differences in thickness by gender [5][6][7]35 and race. 36,37 For example, the Beaver Dam Eye Study population (also predominantly Caucasian) was not gender specific, 7 whereas the CAREDS2 population consisted of only women. We observed thinning of the total retina with increasing age, mainly because of reduced thickness of the inner retinal layers, particularly the GCL, IPL, and ONL. Thinning of these layers was only observed in the inner (3 mm diameter) and outer (6 mm diameter) rings and not in the central circle (1 mm diameter). Our results corroborate other population-based studies demonstrating reduced retinal thickness with age, particularly outside the central circle. [5][6][7]35 Our study has some limitations. The CAREDS cohort includes only women, of primarily (97%) white race, and higher levels of socioeconomic status than similarly aged women in the American population. For this reason the results may lack generalizability to men,  and to women of other races and socioeconomic status (Mares et al., unpublished data). Another limitation is the cross-sectional study design, which limits inferences about causal relationships between the observed associations. In addition, we did not control for axial length and refractive error, whereas previous studies reported that both axial length and refractive errors can influence the macular thickness measurements, particularly in eyes with high axial lengths. [5][6][7]35,38,39 The mean axial length in the population was 23.6 (SD 1.1) and percentage of women with axial length >26 mm was small at 3% and not expected to affect study data. In addition, unlike other SD-OCT devices, the Heidelberg software applies model eye parameters to correct for ocular magnification. 40 Last, we did not control for other diseases that may affect thickness measurements, such as neurodegenerative diseases. 4 Despite these limitations, our study has several strengths. CAREDS2 involved a unique and wellcharacterized cohort of older women, for the study of age-related changes in the eye. An important strength is the age range of CAREDS2 study participants, with few studies containing participants with a mean age of 83 years and age range up to 101 years. We did manual editing of all segmented layers throughout the entire ETDRS grid, which provides the most accurate retinal thickness measurements. In addition, expert graders

Conclusions
We provide normative macular thickness data for a unique and well-characterized cohort of women, mostly white, aged 69 to 101 years in the CAREDS2 study, an ancillary study of the WHI-OS. The peripheral macula thinned with increasing age, particularly in the GCL, IPL, and ONL of the inner retina. Retinal thickness did not differ between eyes with AMD compared to without AMD in the neural retina (retina without the RPE layer). Greater thickness in the ONL may confer advantages for visual acuity, independent of age and AMD. Additional studies of retinal thickness in relation to vision function could enhance our understanding of the value in retinal thickness measurements in clinical practice.
With the population aging in the United States and a likely increase in the prevalence of age-related diseases, including AMD, our study will help to better understand AMD among older women. In addition, the results of the current analysis can be used in future epidemiological and clinical studies evaluating the associations between the macular thickness and age-related diseases.