Participants
This study consists of combined data from 2 Amsterdam UMC sub-studies of the European Medical Information Framework for Alzheimer’s Disease (EMIF-AD): the EMIF-AD 90+ study and the EMIF-AD PreclinAD cohort. The 90+ study consists of cognitively healthy and cognitively impaired subjects, aged ≥90 or over. For extensive recruitment information we refer to our set-up paper of this study by Legdeur et al [17]. The PreclinAD cohort [18] is a cohort consisting of cognitively healthy participants (monozygotic twins) aged ≥60, recruited from the Netherlands Twin Registry [19]. The studies adhered the Tenets of the Declaration of Helsinki and written informed consent was obtained from all participants. The Medical Ethics Committee of the Amsterdam UMC approved both studies.
For complete in- and exclusion criteria of the EMIF-AD 90+ study, we refer to our set-up paper of this study by Legdeur et al [17]. In short, inclusion criteria for the cognitively normal group of the 90+ study were: age ≥90 years and cognitively healthy This group is referred to as ‘healthy nonagenarian’ group.
Inclusion criteria for individuals with cognitive impairment (CI) of the 90+ study were: a diagnosis of amnestic Mild Cognitive Impairment (aMCI) [20] or a diagnosis of probable or possible AD [21]. As during the study we had difficulties identifying subjects of 90 years and older with aMCI or probable or possible AD, we broadened the inclusion criteria in this group to subjects older than 85 years. Six individuals from this group were aged 85-90 years. This group is referred to as ‘CI nonagenarian’ group.
Inclusion criteria for the PreclinAD study were: age ≥60 years, monozygosity and cognitively healthy. For complete in-/exclusion criteria we refer to the set-up paper of this study by Konijnenberg et al [18]. This group is referred to as the ‘control’ group.
From the total 298 participants included in the cohorts, 51 (17.1%) participants were excluded for both the OCT and SIVA analyses, but these were not necessarily the same participants, although there was a high overlap. For the OCT analyses, 9 were excluded due to low quality scans/failed imaging and 42 due to ophthalmological pathology. For the SIVA analyses, 24 were excluded due to low quality images/failed imaging and 27 due to ophthalmological pathology. Interfering ophthalmological pathology consisted mostly of glaucoma and (severe) AMD. Additional file 1 shows the reasons for exclusion in more detail, categorized per group (control, heathy nonagenarian and CI nonagenarian). Although the two study populations (i.e. for OCT and SIVA analyses) were slightly different from each other in terms of included individuals, they were very similar in their demographics, and statistical analyses revealed no significant differences. As such, their demographic information was reported as one combined group.
Medical history
Data about the medical and family history and medication use, in particular on the presence of diabetes mellitus, hypertension and coronary disease, were collected through a structured interview, in combination with information provided by the study partner (if available), general practitioner and/or medical specialist.
Ophthalmological examination
All participants underwent the following ophthalmological examinations: best corrected visual acuity, intra-ocular pressure, refraction data, slit lamp examination, indirect fundoscopy, fundus photography and OCT. Controls received tropicamide 0.5% to enable these examinations, nonagenarians both tropicamide 0.5% and phenylephrine 5% (as mydriasis was harder to achieve in these very aged patients). If a nonagenarian suffered from coronary stenosis, only tropicamide was given, due to the slight risk of phenylephrine inducing a coronary spasm. All photographs/OCT images were assessed by an experienced ophthalmologist (HTN or FDV) for unexpected pathology. Participants suffering from ophthalmological conditions severely interfering with the (neuro)retina or image quality were excluded from analyses (severe cataract, macular degeneration, glaucoma, diabetic retinopathy, vascular occlusions). Eyes with diseases considered to interfere with the OCT measurements excluded from OCT analyses could still be included in the SIVA analyses and vice versa (e.g. AMD with geographical atrophy interfered with OCT, but not fundus image analyses). This resulted in a slightly different study population for the OCT and SIVA analyses.
Optical Coherence Tomography
Using spectral domain OCT (Spectralis, Heidelberg), dense macular scans (49 B-scans) and axonal ring scans around the optic nerve head (ONH) were acquired. Total retinal thickness and individual layer thickness was measured in the macular region. The following individual retinal layers were analyzed: retinal nerve fiber layer (RNFL), ganglion cell layer (GCL) and inner plexiform layer (IPL). A distinction was made between the inner and outer macular ring according to the standard Early Treatment and Diabetic Retinopathy Study (ETDRS) macular grid (1-3mm around the fovea for inner ring and 3-6mm around the fovea for outer ring). For further details on the acquiring of OCT data we refer to our earlier paper by van de Kreeke et al [22].
Fundus photography and quantitative assessment of retinal vasculature
Digital fundus images were made of the fundus of both eyes in all participants (Topcon TRC 50DX type IA). All images were graded by a trained grader (JAvdK) using the Singapore I Vessel Assessment (SIVA) software (version 3.0, National University of Singapore, Singapore) [9-11]. The following 7 retinal vascular parameters were analyzed: central retinal artery equivalent (CRAE), central retinal vein equivalent (CRVE), arteriole–venular ratio (AVR), fractal dimension of the arteriolar network (FDa), fractal dimension of the venular network (FDv), curvature tortuosity of the arterioles (cTORTa) and curvature tortuosity of the venules (cTORTv). All values for retinal vascular parameters were measured within zone C (i.e. 0.5 – 2 disc diameters around the optic nerve head). For further information on the analyses of fundus images we refer to our earlier paper by van de Kreeke et al [23].
Statistical analysis
First, we compared group means of all ocular outcome measures of the healthy and CI nonagenarian groups using linear regression, corrected for age, sex and diabetes. Additionally, mean differences between both nonagenarian groups and younger healthy controls were obtained using Generalized Estimating Equations (GEE). GEE was used to correct for clustering in the data from twin pairs in de control group. It also allowed us to correct for confounders such as sex and a diagnosis of diabetes. We deliberately did not correct this analysis for age, to illustrate the differences based on aging effects. Curvature tortuosity (cTORT) values for arteries and veins were log-transformed to normalize their distribution. All statistical analyses were performed using SPSS (IBM, version 22).