In this study, we analyzed data of the Jidong Eye Cohort Study. Details of the design and methods have been previously described (Yang et al., 2020). Participants in this study were recruited from the Jidong communities (Tangshan city, northern China) from August 2019 to January 2020. In this study, we only entry the participants aged over 40 in this study, because the cognitive impairment occurrence in participants aged below 40 is rare (16/1380). A total of 1833 participants aged over 40 underwent both detailed ophthalmic examination and cognitive function testing (Song et al., 2020). The following participants were excluded from the study: 153 for ocular disorders and vision loss; 125 for suboptimal OCTA images. And no participants in our study had a history of dementia. The analyses ultimately included data of 1555 participants (Figure 1). Among these, 268 (17.2%) displayed poor performance on the Montreal Cognitive Assessment (MoCA) and were categorized into the cognitive impairment group. The remaining 1287 (82.8%) participants were included in the normal group.

This study followed the Declaration of Helsinki guidelines and was approved by the Ethics Committee of Staff Hospital of Jidong oil-field of Chinese National Petroleum (approval document 2018 YILUNZI 1). All participants provided written informed consent.

In this study, we obtained the basic information on the participants using laboratory tests, clinical examinations, and standardized questionnaires for age, sex, income, education level, current smoker status, and current drinker status (Zhang et al., 2013; Xia et al., 2020). Face-to-face interviews were performed by well-trained examiners. We divided the average monthly income into “<¥3000,” “¥3000–5000,” and “>¥5000.” We categorized the education levels as “illiteracy or primary,” “middle school,” and “college graduate or above.” In our study, hypertension was defined as systolic blood pressure ≥ 140 mmHg, or diastolic blood pressure ≥ 90 mmHg, or current use of antihypertension medication, or self-reported history of hypertension. Diabetes was defined as fasting blood glucose (FBG) ≥ 7.0 mmol/L, or current use of antidiabetic medication, or self-reported history of diabetes. Dyslipidemia was defined as low-density lipoprotein (LDL-C) ≥ 3.37mmol/L, or high-density lipoprotein (HDL-C) < 1.04mmol/L, or total cholesterol (TC) ≥ 5.18mmol/L, or triglyceride (TG) ≥ 1.7mmol/L, or current use of lipid-lowering therapy, or self-reported history of dyslipidemia (Chen et al., 2019).

We measured best-corrected visual acuity using standard logarithmic visual acuity charts at a distance of 5 m from participants. Vision loss in this study was defined as best-corrected visual acuity < 0.5 logMAR. We obtained the refractive status using an auto refractometer (KR800; Topcon; Tokyo, Japan). We calculated the refraction as the spherical equivalent (spherical value and half of the cylindrical value). We acquired the digital fundus photographs of both eyes using a 45° non-mydriatic fundus camera (CR2AF; Canon; Tokyo, Japan). At least two independent ophthalmologists identified ocular disorders (i.e., nerve fiber layer defects, macular degeneration, retinal vascular occlusion, and diabetic retinopathy) and vision loss. We excluded participants with ocular disorders and vision loss from the analyses.

We acquired OCTA images using a spectral-domain OCTA (RTVue XR Avanti with AngioVue; Optovue; Fremont, CA, United States). OCTA technology has been previously described in greater detail (Spaide et al., 2018). OCTA image acquisition was performed using a 3 × 3 mm² scan in the macula at a scan density of 304 × 304 A-scans with no pupillary dilation. We used the Early Treatment Diabetic Retinopathy Study grid as follows: two concentric circles with diameters of 1 and 3 mm divided into six quadrants (fovea, parafovea, superior, inferior, nasal, and temporal). The regionals of parafovea served as representatives of average RCP vessel density and average GCC thicknesses in this report due to the special structure of foveal area, although we also presented other regionals. The commercially available AngioAnalytics software (version 2017.1.0.155) automatically provided the vessel density (the proportion of the flow signals within the range of predefined area and depth) of superficial RCP (vessels between the internal limiting membrane and inner plexiform layer) and deep RCP (vessels between the inner and outer plexiform layer) in the macula. We also acquired the foveal avascular zone area (FAZ) and perimeter (PERIM), acircularity index (AI), vessel area density (FD300 AD), and length density (FD300 LD) within a 300 μm width ring surrounding the FAZ from the software automatically. The internally installed software automatically provided the GCC thickness (layer from the internal limiting membrane to the inner plexiform layer), as well.

The OCTA device included a three-dimensional projection artifact removal algorithm to reduce motion artifacts and increase image quality (Kraus et al., 2012). We only included the OCTA images with a signal strength index ≥ 8 and excluded the images with defects (i.e., significant motion artifacts, local weak signal, and incorrect segmentation) from the study. If available, we used the measurement of the right eye for each individual, whereas we used the left eye in the absence of valid data for the right eye (Ito et al., 2020; Ward et al., 2020).

To test for cognitive function, all participants underwent standardized neurological assessments by experienced neuropsychologists using the Montreal Cognitive Assessment (MoCA) and Mini-Mental State Examination. In this study, we use the MoCA to evaluate cognitive impairment as the MoCA is more sensitive and the specificity is not significantly different from others (Kasten et al., 2010; Dong et al., 2012). MoCA is a scale that evaluates visuospatial and executive functions, naming, language, attention, confrontation, memory, and abstraction (Nasreddine et al., 2005). Participants can obtain a maximum of 30 points; lower scores indicate lower cognitive function. If the participant had ≤12 years of education, we increased the total test score by one point. In this study, a score of ≥24 points was used to indicate normal cognitive function (Potocnik et al., 2020). We expressed cognitive function as a continuous MoCA score or a dichotomization of MoCA score as being cognitive impairment or normal.

We calculated the means and standard deviations for the continuous variables. We tested the differences in continuous variables between the groups using the t-test for normally distributed variables. We described the categorical variables using frequencies and percentages and examined them using the chi-square or Fisher’s exact test. Missing data were deemed as missing at random.

We performed multivariable models to estimate the association between vessel densities in the RCP and GCC thicknesses and cognitive impairment. Multivariable logistic regression model was used for binary cognitive impairment or normal MoCA score and multivariable linear regression model for continuous MoCA score. We adjusted all multivariable models for age, sex, educational status, income, refraction, hypertension, diabetes, and dyslipidemia. We also tested the interaction effects for the association of the RCP and GCC with cognitive impairment within the subgroups (sex, diabetes, hypertension, age) by including interaction terms in the adjusted model using all subjects in our study.

We measured the association as odds ratios (ORs) or Betas and 95% confidence intervals (CIs) per standard deviation decrease. We considered statistical significance as a P-value < 0.05. We performed all statistical analyses using the SAS software, version 9.4 (SAS Institute Inc., Cary, NC, United States).

A total of 1555 participants were included in the analyses. The mean age of the included participants was 52.3 (8.4) years, and 861 (55.4%) were women. Cognitive impairment was observed in 268 (17.2%) participants. Table 1 shows the baseline characteristics of the cognitive impairment (MoCA < 24) and normal (MoCA ≥ 24) groups.

Participants in the cognitive impairment group were older (P < 0.001) and less educated (P < 0.001), with lower incomes (P < 0.001), less myopia (P < 0.001), and more prevalent hypertension (P < 0.001), diabetes (P = 0.005), and dyslipidemia (P < 0.001). However, there were no significant differences between the two groups in terms of sex, current smoker and drinker status, their body mass index (BMI), and levels of TC, TG, HDL-C, and LDL-C.

Figure 2 and Table 2 show the RCP and GCC characteristics measured using OCTA on the baseline survey. We observed significant differences between the two groups in terms of superficial RCP vessel density (parafovea: P = 0.003; temporal: P = 0.02; superior: P = 0.002; nasal: P = 0.007; inferior: P = 0.006), deep RCP vessel density (parafovea: P < 0.001; temporal: P < 0.001; superior: P < 0.001; nasal: P < 0.001; inferior: P < 0.001), FD300 AD (P < 0.001), and FD300 LD (P < 0.001). There were no significant differences between the groups for FAZ, PERIM, and AI. For GCC thickness, there were significant differences between groups (parafovea: P = 0.01; temporal: P = 0.04; superior: P = 0.003; nasal: P = 0.02; inferior: P = 0.01). Figure 3 shows the distributions of the RCP vessel density and GCC thickness according to the MoCA score.

Cognitive impairment was respectively observed in 16.7% and 17.9% of female and male participants, 24.4% and 16.2% of patients with and without diabetes, and 21.7% and 14.6% of patients with and without hypertension, and 14.9% and 43.8% of participants aged <65 and ≥65 (Supplementary Figure 1).

We show the unadjusted and adjusted ORs of the RCP and GCC for cognitive impairment in Table 3. The adjusted OR with 95% CI for parafovea vessel density in deep RCP with cognitive impairment was 1.20 (1.03–1.39). Moreover, in regionals of temporal, superior, and inferior, the adjusted ORs with 95% CIs were 1.18 (1.01–1.37), 1.21 (1.05–1.40), and 1.17 (1.01–1.36), respectively. The adjusted ORs with 95% CIs for FD300 AD and FD300 LD with cognitive impairment were 1.23 (1.07–1.41) and 1.30 (1.13–1.49), respectively. The adjusted OR with 95%CI for thickness in the superior GCC with cognitive impairment was 1.16 (1.01–1.32).

In addition, we show the unadjusted and adjusted Betas of the RCP and GCC for cognitive function in Supplementary Table 1. The adjusted Betas with 95% CIs for each standard deviation decrease in vessel density of superficial and deep RCP (parafovea) with continuous MoCA score were –0.160 (–0.304 to –0.016) and –0.251 (–0.409 to –0.093), respectively. The adjusted Betas with 95% CIs for FD300 AD and FD300 LD with continuous MoCA score were –0.225 (–0.372 to –0.078) and –0.318 (–0.465 to –0.170), respectively. Further, the adjusted Beta with 95% CI for thickness in the parafovea GCC with continuous MoCA score was –0.205 (–0.351 to –0.059).

We observed significant interactions of diabetes with GCC for cognitive impairment (Table 4). The ORs of the GCC items in participants with or without diabetes for cognitive impairment were 1.77 vs. 1.07 (P for interaction = 0.02) for parafovea, 1.71 vs. 1.05 (P for interaction = 0.02) for temporal, 1.74 vs. 1.07 (P for interaction = 0.01) for superior, 1.75 vs. 1.06 (P for interaction = 0.02) for nasal, and 1.70 vs. 1.08 (P for interaction = 0.04) for inferior. We did not find significant interactions of the GCC with sex, hypertension, or age for cognitive impairment. In addition, no interactions between RCP and sex, diabetes, hypertension, or age were observed for cognitive impairment (Tables 4, 5).