Eye Adult Changes in Thought (Eye ACT) Study: Design and Report on the Inaugural Cohort

INTRODUCTION

The concept of using the eye, often referred as the “window to the brain,” for understanding brain health has been explored for several decades. Remarkable recent advances in non-invasive retinal imaging and the marriage of Big Data and artificial intelligence (AI) have led to increasing enthusiasm for discovering eye-related biomarkers of Alzheimer’s disease and related dementias (AD/ADRD), either as early diagnostics, prognostics, or modifiable risk factors.

Since the 1986 optic nerve atrophy in AD report by Hinton et al. [1], additional associations between retinal imaging features and impaired cognition and/or AD have been reported, including structural changes such as thinning of the retinal nerve fiber layer (RNFL) [2 –8], macular volume [2, 9], ganglion cell-inner plexiform layer (GC-IPL) [2 , 11], and choroidal volume [12 –14]. These associations also include functional changes such as decreased blood flow [15, 16] and attenuated vasculature [17 –23].

Studies to date have conflicting findings on retinal features and AD/ADRD risk [18 –20] likely related to cross-sectional designs, small convenience samples, and the use of clinical rather than research-based AD/ADRD diagnoses [24 –29]. Longitudinal studies have limited sample sizes (n = 20) [30], and many imaging studies are not comparable due to differences in protocols and retinal imaging devices. Some promising studies rely on AD biomarker-positive cognitively intact younger adults at baseline, which may not be applicable to older adults [31 –34]. Furthermore, most studies have excluded people with eye diseases, common in older people, and therefore preclude investigation of associations between eye diseases and cognitive functioning [35 –37].

The Eye Adult Changes in Thought study (“Eye ACT”) was designed to address these limitations by following a well-characterized community-based cohort and obtaining uniquely detailed eye data via repeated prospective eye imaging (performed in clinic or at home) accompanied by enriched eye clinical information abstracted from health records. Participants in Eye ACT are recruited from the parent ACT study [38], an ongoing prospective study with a large sample and high retention rate (>90%). The ACT study provides extensive resources for Eye ACT to leverage: routine cognitive testing at the participant’s home or research clinic, research protocols to identify incident AD/ADRD [39], and state-of-the art neuropathology evaluation among deceased participants who volunteered for brain autopsy. The overall hypothesis is that visual function alterations and changes in the retinal structure and vasculature may serve as biomarkers of early AD/ADRD. In this paper, we provide the detailed Eye ACT research protocol and descriptive statistics from the first 330 Eye ACT participants who have been participating in prospective eye data collection since 2021.

METHODS

The Prospective Eye ACT study cohort

The Eye ACT study recruits and enrolls participants from the parent ACT study. Eye ACT is unique in that there are no exclusion criteria specific to the study. Instead, Eye ACT reaches out to the entire ACT cohort for potential recruitment. At the initiation of the study, only established ACT participants undergoing biennial visits at the research clinic were recruited for Eye ACT. More recently, ACT participants seen at home for biennial visits who agree to Eye ACT home study visits have been included (Fig. 1). Leveraging the parent ACT’s currently ongoing ambitious aims to increase overall cohort size with emphasis on diversity, Eye ACT also plans to increase cohort size to over 2,000 active participants and begin recruiting new ACT enrollees.

OPEN IN VIEWER

Fig. 1

Flow chart of participants. Relationship between the parent ACT study and Eye ACT study cohorts (retrospective and prospective).

RESULTS

Total enrollment of the parent ACT study was 5,763 participants as of 2020. Of these, 2,198 ACT participants were alive and at risk of developing AD/ADRD as of 12/2021. A total of 330 of these have been recruited to the Eye ACT study as of 03/2023, the most recent ACT data freeze date, and included in the analytic set, with 300 seen in the research clinic and 30 seen at home. A total of 109 individuals declined participation in Eye ACT (4% of research clinic and 75% of home participants).

DISCUSSION

Eye ACT is a prospective, longitudinal, community-based study of older adults that provides an unparalleled opportunity to investigate the relationship between the eye and brain, with unique home visit capacity and unprecedented past eye health data from both paper chart and EHR data. In this initial report, we compare demographic and clinical characteristics of Eye ACT study participants to the parent ACT study to better understand potential sources of selection bias in future analyses of associations between ophthalmic biomarkers and AD/ADRD risk. Compared to the rest of the parent ACT participants, current Eye ACT study participants were younger at ACT enrollment, more recently enrolled in the ACT study, and overall healthier. However, these differences were small. Among the Eye ACT cohort, participants seen at home had significantly worse visual acuity and contrast sensitivity compared to participants seen in the research clinic, while cognitive scores and the majority of the optical coherence tomography measurements were similar.

The Eye ACT cohort should provide opportunities to address several conflicting findings in the existing literature on retinal imaging biomarkers of AD/ADRD. In many studies, retinal thinning of the RNFL and GC-IPL, and retinal vascular changes including decreased foveal avascular zone, increased vessel tortuosity, and lower vessel density of deep capillary plexus were reported to be associated with AD/ADRD [3 , 31]. However, many other studies found no significant associations [18–20 , 27]. Conflicting results may be due to cross-sectional designs [24 –27], small sample sizes, lack of longitudinal studies [50], disparate protocols for vision measures and retinal imaging, and variability of dementia diagnosis characterization [24 –29]. The Eye ACT study can address some of these gaps with the incorporation of prospective data collection on retinal biomarkers in a large, cognitively well-characterized cohort and by including vision function data, retinal imaging, and clinical ophthalmic data. Eye ACT is also unique in incorporating eye disease history and visual function in parallel with eye imaging data collection. Previous studies have found associations between eye diseases and decreased cognition [51, 52], vision impairment and dementia [52, 53], and correlations between cataract surgery and lower risk of AD/ADRD [54, 55].

When designing a prospective, observational study with voluntary participation, it is important to understand potential selection biases that would undermine accuracy of future association studies [46 , 56–58]. We found several factors that were considerably different between Eye ACT participants and other active ACT study participants. Eye ACT participants were overall healthier, displaying decreased frequency of cardiovascular disease, cerebrovascular disease, hypertension, visual impairment, and AMD. All of these comorbidities have been shown to be associated with risks of AD/ADRD [51 , 59–61]. Eye ACT participants were more likely to be male and had slightly higher education levels compared to the parent ACT cohort. Although Eye ACT participants were generally healthier, a slightly higher percentage of them had cataracts compared to ACT participants. Even though this difference was nonsignificant, ascertainment bias is also a possibility since Eye ACT participants were more likely to seek eye care based on the higher number of eye encounters and therefore may be diagnosed with cataracts earlier. As Eye ACT recruitment from the parent ACT study expands to include more participants seen at home, these differences will likely decrease, and we will account for any remaining sources of selection bias in future studies. We routinely perform inverse probability weighting (IPW) to account for selection bias among participants with brain autopsy consent [62]. Survival bias may also be an issue, which we will address by weighting longitudinal models by the probability of death or dropout [54] or by treating death and dropout as competing risks.

Prior work by Crane et al. involving the parent ACT study revealed the critical importance of enabling study visits at home given the added difficulty in traveling to clinic settings for those who are frail, physically ill, or cognitively impaired [46]. Patients who cannot travel to research clinics are currently underrepresented in dementia research and may have differing dementia risk profiles. In that prior work, the authors found a 23%, 39%, and 75% reduction in sample size, follow-up time, and number of incident AD/ADRD cases, respectively, in the cohort restricted to clinic-only participants compared to those seen either at home or in clinic. More importantly, analyses using data from clinic-only participants alone resulted in overestimation of the association of several factors—age, cerebrovascular disease, APOE, and several neuropathological findings—with AD/ADRD risk compared to the results for participants being seen at home or in clinic [46]. This highlights the issue of selection bias and consequently inaccurate conclusions in studies that only include research clinic data collection. Validation of clinic-based study results using community-based studies with home visit capacity is critical in dementia research.

Sensory impairment is one of the primary risk factors of interest for dementia development. Visual impairment may contribute to social isolation and decreased cognitive stimulation, which may exacerbate dementia risk [54 , 64]. Eye ACT participants evaluated at home were older and had significantly worse visual acuity and contrast sensitivity than those seen in clinic, potentially making them at higher dementia risk compared to those seen in clinic. While we did not find significant differences in CASI scores between participants evaluated in the research clinic versus at home in this pilot sample, the differences in risk factors between these groups highlight the need for the longitudinal follow-up.

Interestingly, while functional eye measurements (i.e. visual acuity and contrast sensitivity) were significantly different between clinic and at home groups, no obvious difference was noted in imaging measurements, except for a difference in total macular thickness in the better seeing eye between cohorts. Several studies have previously shown that functional ophthalmic measures are associated with cognitive impairment and dementia risks [65 –67], thus our findings may suggest that functional deficits could occur earlier than detectable imaging changes in AD/ADRD pathology development. Patient-dependent factors, such as cooperation and concentration level during data collection. can induce noise in measuring functional deficits, making later analysis and ability to draw meaningful insight from the data more difficult [68 –70]. However, the anticipated Eye ACT longitudinal data, which will follow both functional and structural imaging deficits, will provide more insights.

Several limitations exist with our study. Currently, the ACT study is >90% non-Hispanic white. However, ongoing initiatives of the parent ACT study include increasing sample size from 2000 to 3000 with targeted recruitment of racial and ethnic minority groups. Thus, Eye ACT is also increasing its recruitment goal to enroll a more diverse population, which will increase the generalizability of the study findings. The pilot data only include cross-sectional evaluations, but future biennial visits will allow longitudinal data analyses. Recruitment is ongoing with over 600 participants to date towards a goal of ≥80% of all eligible ACT participants. We only analyzed data from the most recent ACT data freeze date (03/2023) since more recent data acquired after the data freeze are not as complete due extensive quality control that happens at the time of the data freeze, including conversion of the CASI scores into IRT. Missing data are an important consideration. OCTA was not available at home visits initially but is now being added to home visits. To our best knowledge, this will be the first OCTA capture of older adults at home, allowing assessment of whether retinal vascular health significantly differs between those seen at home versus in clinic. Image quality can vary depending on participant factors such as small pupil size, tremor, and fatigue during study visits. We plan to identify non-random missing variables and adjust for bias using statistical methods such as IPW [62].

In summary, this study has identified potential sources of bias and is designed to address them, as well as other gaps in previous research. Home visit participants had significantly decreased visual function compared to research clinic participants. Given potentially high rates of dementia among home visit participants in the ACT study, continuing to recruit at-home participants is critical to remove selection bias in future ophthalmic biomarker and dementia risk association studies. With its novel combination of functional and imaging eye measurements, retrospective eye health data, study setting, and anticipated longitudinal follow-up, Eye ACT provides unique strengths to assess potential retinal biomarkers of dementia.

AUTHOR CONTRIBUTIONS

Cecilia Lee (Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Supervision; Writing – original draft; Writing – review & editing); Alina N. Ferguson (Writing – original draft; Writing – review & editing); Laura E. Gibbons (Formal analysis; Methodology; Writing – original draft; Writing – review & editing); Rod Walker (Formal analysis; Methodology; Writing – review & editing); Yu-Ru Su (Formal analysis; Methodology; Writing – review & editing); Chloe Krakauer (Formal analysis; Methodology; Writing – review & editing); Michael Brush (Methodology); Jason Kam (Methodology); Eric B. Larson (Conceptualization; Funding acquisition; Methodology; Writing – review & editing); David E. Arterburn (Conceptualization; Writing – review & editing); Paul K. Crane (Conceptualization; Funding acquisition; Methodology; Writing – review & editing).