This study is embedded in the NORFACE project, which was founded in 2014 with the goal investigate retinal biomarkers of AD and analyze the relationship between retinal changes and different types of neurodegenerative disorders (Sánchez et al., 2018). Between February 2018 and March 2019, consecutive patients with a diagnosis of MCI (Petersen, 2004) evaluated at Ace Alzheimer Center Barcelona and who underwent, within 12 months, a lumbar puncture (LP) for the quantification of CSF core biomarkers for AD and an ophthalmological exam/OCT scan, were enrolled in the present study. Participants were recruited from three different sources: (1) the Memory Clinic, an outpatient diagnostic unit for individuals with cognitive decline referred by physicians from the Catalan Public Health System (Boada et al., 2014), (2) Fundació ACE’s Open House Initiative (Rodríguez-Gómez et al., 2015), a social program that assesses for free the cognitive status of individuals from the community without the need for a physician’s referral, and (3) the BIOFACE project, a research study of novel biomarkers in early onset MCI (Esteban-De Antonio et al., 2021). Inclusion criteria were: consensus-based clinical diagnosis of MCI (Petersen, 2004), age ≥50 years old, availability of APOE ε4 status, ability to complete the full ophthalmological exam and OCT scan and CSF core biomarkers for AD performed within 12 months of the OCT-A scan.

Further, a group of participants with subjective cognitive decline (SCD) from the Fundació ACE Healthy Brain Initiative (FACEHBI) cohort (Rodriguez-Gomez et al., 2017) with no objective impairment on formal cognitive testing and absence of brain amyloid uptake in a Florbetaben-PET scan (SCD Aβ-) were included as the control group (n = 83) in additional analyses. These SCD participants underwent the same cognitive testing, ophthalmological exam and OCT scan protocol than the MCI participants from the NORFACE cohort included in the main analyses.

Neurological, neuropsychological and social evaluations of the study participants were performed at Ace Alzheimer Center Barcelona. A consensus-based diagnosis of cognitive status was reached for each participant at the time of the study recruitment by a multidisciplinary team of professionals (Boada et al., 2014). Cognitive assessment consisted of the Spanish version of the Mini-Mental State Examination (MMSE) (Folstein et al., 1975; Blesa et al., 2001), the memory test of the Spanish version of the 7 min screening neurocognitive battery (del Ser Quijano et al., 2004), the Spanish version of the Neuropsychiatric Inventory Questionnaire (NPI-Q) (Cummings et al., 1994; Boada et al., 2005), the Geriatric Dementia Scale (GDS) (Reisberg et al., 1982), the Clinical Dementia Rating Score (CDR) (Morris, 1993), the Blessed Dementia Scale (Blessed et al., 1968), and a comprehensive neuropsychological battery of Fundació ACE (NBACE) (Alegret et al., 2012, 2013). Demographic information collected included age, gender and years of formal education. Past medical history collected included smoking habit (former or current), hypertension, diabetes mellitus, dyslipidemia, heart disease, and chronic obstructive pulmonary disease (COPD). This clinical information was obtained from participants’ medical records. MCI was defined using Petersen’s (Petersen, 2004) and the Cardiovascular health and cognition study criteria (Lopez et al., 2003).

Genomic DNA was extracted from peripheral blood using the commercially available Chemagic system (Perkin Elmer). APOE genotypes were extracted from the Axiom SP array (Thermo Fisher Scientific) (Moreno-Grau et al., 2019; de Rojas et al., 2021). Ace Alzheimer Center Barcelona has this variable as a standard in its assessment protocols. Alternatively, the APOE genotypes were determined using fluorogenic allele-specific oligonucleotide probes (TaqMan assay; Life Technologies, Spain) for rs7412 (Test ID: C____904973_10) and rs429358 (Test ID: C___3084793_20). For the TaqMan assays, PCR and real-time fluorescence measurements were carried out on a QuantStudio3 real-time PCR system (Thermo Fisher Scientific, Spain) using the TaqMan Universal Master Mix (ref: 4364341, Life Technologies, Spain) methodology according to manufacturer’s instructions. The polymerase chain reaction was performed as follows: first, a pre-read step for 30 s at 60°C, denaturation for 10 min at 95°C, followed by 40 cycles at 95°C for 15 s and 60°C for 1 min, and a post read stage for 30 s at 60°C. The genotype was determined using the Genotyping App for Thermo Fisher Scientific Cloud by clustering analysis. The laboratory technicians were blinded to other study variables. APOE ε4 status (presence of at least one ε4 allele) was reported as a covariate in the analyses.

Lumbar punctures were performed at Ace Alzheimer Center Barcelona by an experienced neurologist under fasting conditions. CSF was collected passively in two 10-ml polypropylene tubes (Sarstedt Ref 62.610.018) centrifuged (2000 × g 10 min at 4°C), aliquoted and stored in polypropylene tubes (Sarstedt Ref 72.694.007) at −80°C until its use. Time delay between CSF collection and storage was less than 2 h. The collection protocol follows the recommendations of the Alzheimer’s Biomarkers Standardization Initiative (Vanderstichele et al., 2012). The day of the analysis, one aliquot of 0.5 mL was thawed and used for the determination of Aβ1-42, Total Tau (T-tau) and p181-tau. Aβ and tau proteins were quantified by either the commercially available enzyme linked immunosorbent assays (ELISA) (Innotest, Fujirebio Europe) (n = 114) or the chemiluminescence enzyme immunoassay (CLEIA) using the Lumipulse G 600 II automatic platform (Fujirebio Inc.) (Leitão et al., 2019) (n = 30).

Using CSF biomarkers, participants were classified into three categories according to the AT(N) scheme (Jack et al., 2018): Normal AD biomarkers (A-T-N-), Alzheimer’s continuum (including A+T-N-, A+T+N-, A+T+N+ and A+T-N+) and Non-AD pathologic changes [Suspected non-Alzheimer pathology (SNAP), including A-T+N-, A-T-N+ and A-T+N+], where A refers to aggregated Aβ, T to aggregated tau and N to neurodegeneration or neuronal injury (Supplementary Table 1). Cut-offs from the Ace Alzheimer Center Barcelona CSF program were used to dichotomize each CSF biomarker into ± as follows: for ELISA, Aβ1-42 < 676 pg/ml for A, p181-tau > 58 pg/ml for T and t-tau > 367 pg/ml for N; for CLEIA, Aβ1-42 < 796 pg/ml for A; p181-tau > 54 pg/ml for T and t-tau > 412 pg/ml for N (Orellana et al., 2022).

In parallel to the cognitive assessment, study participants underwent a complete neuro-ophthalmological evaluation, which lasted about 20 min and was performed by an optometrist. The evaluation comprised: (1) a review of past ophthalmological diseases, treatments and surgeries, (2) monocular visual acuity assessment with the participants wearing their habitual correction for refractive error using a pinhole occluder and the Early Treatment of Diabetic Retinopathy Study (ETDRS) chart (Chew et al., 2009; Bokinni et al., 2015), (3) intraocular pressure (IOP) measurement by Icare tonometry (Pakrou et al., 2008), and (4) swept source (SS) OCT scan. More details can be found elsewhere (Marquié et al., 2022). The ophthalmologist and neurologists were blind to each other’s diagnosis.

Participants were imaged with a DRI OCT Triton–Swept Source (SS) OCT (Topcon Co. Tokyo, Japan). The OCT exam was completed in about 5–10 min, no pupil dilation was required, and both eyes were scanned separately. The OCT Angiography Ratio Analysis (OCTARA) processing software was used to analyzed the data. An automatic segmentation method was employed to obtain measures of the superficial vascular plexus and the quantification of VD, expressed as the % of area covered by vessels. VD measures were obtained in a 6 mm × 6 mm area centered in the fovea, and the central area (1 mm circle) was excluded from the analysis. The parafoveal area, defined by two concentrical rings measuring 1 and 3 mm in diameter, respectively, was subdivided into four quadrants: nasal, superior, temporal and inferior. More details can be found in a recent publication from the NORFACE cohort (Marquié et al., 2022). The retinal vascular network consists of two separate beds: the superficial vascular plexus, at the level of the ganglion cell layer, and the deep vascular plexus, at the level of the outer plexiform layer (Toussaint et al., 1961). The FAZ is the region in the center of the macula within the fovea that is devoid of vasculature. In our analysis we used data from macular VD in the superficial vascular plexus.

Only VD measures from the right eye were used for the analysis, as in previous papers from our group (Sánchez et al., 2018, 2020; Marquié et al., 2020). OCT-related exclusion criteria were the following: lack of collaboration in the neuro-ophthalmological exam or OCT scan, OCT data obtained only from the left eye, presence of OCT artifacts and diseases that could affect retinal measurements [e.g., open-angle glaucoma and other neuropathies maculopathies, prior retinal surgery, intraocular pressure (IOP) ≥24 mmHg, high myopia (< −6 Dp) or hyperopia (> + 6 Dp) and optic nerve congenital abnormalities].

This study and its informed consent were approved by the ethics committees of the Hospital Clínic i Provincial de Barcelona in accordance with Spanish biomedical laws (Law 14/2007, July 3rd, about biomedical research; Royal Decree 1716/2011, November 18th) and followed the recommendations of the Declaration of Helsinki. All participants signed an additional informed consent for the lumbar puncture procedure.

Data processing and analysis were carried out using R version 4.1.2 (R Core Team, n.d.).¹ All data were examined for normality, skew and restriction of range. All quantitative variables were normally distributed. Frequency analysis and measures of central tendency and dispersion were used to describe the demographic (age, sex, education) and clinical (hypertension, diabetes mellitus, dyslipidemia, heart disease, chronic obstructive pulmonary disease (COPD) and smoking) variables among the three AT(N) groups (Normal, Alzheimer, and SNAP). To summarize the distribution of these demographic and clinical variables among the three AT(N) groups, bivariate Analysis of Variance (ANOVA) and Pearson’s chi-squared tests were executed.

For the final multivariate model, we included all clinical variables (hypertension, diabetes mellitus, dyslipidemia, heart disease, COPD, and smoking) as adjusting factors. To identify which demographic variables should be additionally included as adjusting factors in the final model, a multinomial regression analysis was executed to determine their differential distribution among the three AT(N) groups. An analysis was performed for the demographic variables (age, sex, education, and APOE ε4 status) including all the clinical variables (hypertension, diabetes mellitus, dyslipidemia, heart disease, COPD and smoking habit) as adjusting factors. The Normal AT(N) group was considered the reference category. For these analyses, alpha level was set at p < 0.05.

The main analyses consisted of four multivariate regression analyses, one for every macular VD measure (nasal, superior, temporal, and inferior quadrants), including the three AT(N) groups (Normal, Alzheimer and SNAP) as discriminant factors and adjusting their effect by all six clinical variables and those demographic factors that showed any significant effect in the former multinomial regression analysis. The Normal AT(N) group was considered the reference category. Regression coefficients (the mean change in the outcome variable for one unit of change in the predictor variable while holding other predictors in the model constant), betas (the degree of change in the outcome variable for every one unit of change in the predictor variable) and t (assessing whether the beta coefficient is significantly different from zero) are reported.

The former four multivariate regression analysis were rerun without including the A+T-N- participants within the AT(N) Alzheimer group (amyloidosis alone without tauopathy or neurodegeneration, n = 9).

Additionally, the former four multivariate regression analyses were repeated including a group of participants with subjective cognitive decline (SCD) and absent brain amyloid uptake in a FBB-PET scan (SCD Aβ-) from the Fundació ACE Healthy Brain Initiative (FACEHBI) cohort (Rodriguez-Gomez et al., 2017) as the reference category (n = 83).

For the former multivariate regression analyses, alpha level was set up at p < 0.004 (0.05/12) after Bonferroni’s correction for multiple comparisons.

The association between individual CSF biomarkers (Aβ1-42, p181-tau and t-tau) and each of the four macular VD measurements (nasal, superior, temporal and inferior quadrants) was explored using separate partial correlations, including the same covariates. These analyses were performed separately for the ELISA and CLEIA groups and also for the whole cohort after a log-transformation of each CSF biomarker value. For these correlation analyses, alpha level was set at p < 0.004 (0.05/12), after Bonferroni’s correction for multiple comparisons.

To investigate whether a differential effect could be detected when considering sex, the previous four multivariate regression analyses were executed again, including now the interaction between AT(N) group and sex as the main factor of interest and the same covariates. For these analyses, alpha level was set at p < 0.0125 (0.05/4), after Bonferroni’s correction for multiple comparisons.

Data from 1,648 individuals with available clinical information and an OCT-A scan performed between January 2018 and March 2019 were initially reviewed (Figure 1). Several exclusion criteria were applied: lack of CSF AD core biomarkers data within 12 months of the OCT-A scan (n = 1363), not fulfilling the diagnostic criteria for MCI (n = 115), ophthalmological conditions that could interfere with OCT-A measurements (n = 23; n = 8 due to retinal surgery, n = 6 due to retinopathy, n = 5 due to open angle glaucoma, n = 3 due to IOP > 24 mmHg, n = 1 due to other reasons) and finally, lack of information on APOE ε4 status (n = 3).

The final sample consisted of 144 MCI individuals with available OCT-A scan and CSF AD core biomarkers performed within 12 months. Participants were classified into three groups according to the CSF AT(N) scheme (Jack et al., 2018): 66 Normal AT(N), 45 Alzheimer AT(N) and 33 SNAP AT(N). The demographic characteristics and past medical history of the cohort are displayed in Table 1.

The multinomial regression analysis exploring the distribution of age, sex, education and the APOE ε4 status among the AT(N) groups, using the clinical variables (hypertension, diabetes mellitus, dyslipidemia, heart disease, COPD, smoking habit) as adjusting factors, showed that age and the APOE ε4 status had a significant effect, so both were included as adjusting factors in the final analysis (Supplementary Table 2).

Table 2 depicts the contribution of age, the APOE ε4 status, hypertension, diabetes mellitus, dyslipidemia, heart disease, COPD and smoking habit as adjusting factors and the AT(N) groups to macular VD variance in each quadrant. Regression models revealed no significant effect of AT(N) groups on VD values (all, p ≥ 0.162) (Figure 2). Similarly, APOE ε4 status, hypertension, diabetes mellitus, heart disease, COPD and smoking habit did not show any significant effect on VD (all, p ≥ 0.124). Age showed an inverse relationship with VD in the superior quadrant (p = 0.041) and dyslipidemia an inverse relationship with VD in the nasal quadrant (p = 0.043), but both were no longer significant after correction for multiple comparisons (p > 0.004).

Raw and adjusted sector-specific macular VD measures across diagnostic groups are displayed in Supplementary Table 3.

Representative macular VD images from the superficial plexus for each AT(N) group are shown in Figure 3.

A sensitivity analysis was executed with the former multivariate regression analysis without including the A+T-N- participants within the Alzheimer AT(N) group (n = 9), showing similar results than the former, with no significant effect of AT(N) groups on VD values (Supplementary Table 4).

The former multivariate regression analysis were repeated using a cohort of 83 individuals with SCD with no evidence of brain amyloid deposition (negative FBB-PET scan) from the FACEHBI cohort as the reference category (SCD Aβ-), including age, sex, years of education, APOEε4 status, hypertension, diabetes mellitus, dyslipidemia, heart disease, COPD and smoking habit as covariates. In this case, regression models revealed significant differences in VD in several macular regions between groups (see Supplementary Tables 5, 6). The Alzheimer AT(N) group showed higher VD measures in the temporal quadrant (p = 0.005) compared to the SCD Aβ- group, but that was no longer significant after correction for multiple comparisons (p > 0.004). The SNAP AT(N) group showed higher VD measures in the nasal (p = 0.008), temporal (p = 0.002) and inferior (p = 0.004) quadrants compared to the SCD Aβ- group, but only the temporal and the inferior quadrants remained significant after correction for multiple comparisons (p < 0.004). Additionally, age was negatively associated with VD in the superior quadrant (p = 0.004).

The interaction between sex and AT(N) groups had no effect in differentiating VD measurements, adjusted by age, age, APOE ε4 status, hypertension, diabetes mellitus, dyslipidemia, heart disease, COPD and smoking habit (Supplementary Table 7). Thus, differences in macular VD among AT(N) groups were not significantly influenced by sex.

Cerebrospinal fluid Aβ1-42, p-tau and t-tau values from the whole cohort (n = 144) were log-transformed and correlated separately with the four macular VD measures. None of the three CSF biomarkers showed a significant correlation with any macular VD measurements, adjusted by age, APOE ε4 status, hypertension, diabetes mellitus, dyslipidemia, heart disease, COPD, smoking habit and CSF technique (all r < 0.13; p ≥ 0.133) (see Table 3). Additionally, the analyses were repeated using raw CSF data for the ELISA (n = 114) and CLEIA (n = 30) groups separately, showing again a lack of correlation of each CSF biomarker with macular VD measures (see Supplementary Tables 8, 9).