Study population
The data used in the present study were obtained from the SILCODE. SILCODE is a longitudinal study performed at Xuanwu Hospital of Capital Medical University, Beijing, China in April 2017, which aimed at identifying biomarkers related to the early diagnosis of AD by collecting multimodal neuroimaging data from the SCD population. This study was registered at ClinicalTrials.gov (number NCT03370744). And a detailed study protocol has been described elsewhere [18,19]. In summary, subjects participating in the SILCODE were asked to complete a series of clinical evaluation including medical history, physical and neurological examinations, neuropsychological tests, laboratory tests, and brain magnetic resonance imaging (MRI) scanning and as well as optional [18F] florbetapir (AV-45) positron emission tomography (Aβ-PET) or [18F] fluorodeoxyglucose (FDG) positron emission tomography (FDG-PET) separately by a GE Signa integrated PET/MRI system (Germany) at baseline. The same examinations will be performed except for PET scans during the 15-month follow-up.
All eligible participants aged between 60 and 80 years old were Mandarin- speaking, had at least 6 years of education, and were right-handed. The individuals meeting the following conditions could be diagnosed as SCD: (1) with self-reported memory complaints; (2) failure to meet criteria for mild cognitive decline (MCI) [20], dementia due to AD established by the National Institute on Aging Alzheimer’s Association workgroups (NIA-AA) [21], and any other disorders or conditon that may cause cognitive impairment, such as stroke, traumatic brain injury or gas poisoning.
A total of 72 SCD subjects with Aβ-PET data available were selected for this study. With the use of pre-established cutoff value of 1.18 applied to the global AV45-PET standardized uptake value ratio (SUVR) [22], participants were subdivided into two groups: Aβ- SCD subjects (SUVR≤1.18, n=43) and Aβ+ SCD subjects (SUVR>1.18, n=29). For details see figure 1 for a flowchart.
Clinical and neuropsychological assessments
A paper case report form (CRF) was used to record demographic features (for example, name, age, gender, education, occupation, etc.), medical history, biochemical examination, a battery of neuropsychological tests and clinical diagnosis at the baseline and at different visits. In this study, we focus on 4 cognitive domains: episodic memory (Auditory Verbal Learning Test-Huashan version long-delayed free recall [AVLT-H-N5] and recognition [AVLT-H-N7]) [23], language (Animal Fluency Test [AFT]; Boston Naming Test [BNT]) [24,25], speed/executive function (Shape Trailing Test A [STT-A] and B [STT-B]) [26], global cognition (Montreal Cognitive Assessment-Basic [MoCA-B]) [27]. Details on neuropsychological tests published previously [19].
APOE genotyping
For each participant, a fasting blood sample was drawn in the department of laboratory, Xuanwu Hospital at baseline. A part of this blood sample was used for analysis of the level of blood glucose, blood lipids, anti-syphilis, homocysteine, folic acid, vitamin B12, thyroid hormone, hemoglobin, blood coagulation and the other part was used to determine the Apolipoprotein E (ApoE) gene polymorphism status. Details of APOE genotyping have been published elsewhere [19]. In this study, APOE genotype was dichotomized into SCD individuals with 1 or 2 copies of the ε4 allele (APOE ε4 carriers) and those without any copies of the ε4 allele (APOE ε4 non-carriers).
AV45-PET Analysis
PET images were preprocessed with statistical parametric mapping (SPM12, available at https://www.fil.ion.ucl.ac.uk/spm/software/spm12/) software in MATLAB (version R2014a; MathWorks, Natick, MA, United States). For subjects with T1 scans, their PET images were registered to the corresponding structural MRI image. The structural MRI images were segmented into gray matter, white matter and cerebrospinal fluid tissue probability maps, in which nonlinear transformation parameters were obtained. Then, the non-linear transformation parameters were used to normalize the registered PET images into the Montreal Neurological Institute (MNI) stereotactic template. For subjects without T1 scans, their PET images were spatially normalized into MNI brain space with the TPM atlas. Then all PET images were resampled into 3 × 3 × 3 mm3 voxels. Finally, normalized PET images were smoothed by an isotropic Gaussian smoothing kernel with the full-width at half maximum (FWHM) of 8 × 8 × 8 mm3 to improve the signal-to-noise ratio. PET images were subsequently scaled to the whole cerebellum to get the SUVR maps. Then, mean SUVR maps of each group were obtained for visualization by MRIcroN (available at https://www.nitrc.org/projects/mricron) to show the difference in Aβ level between the two groups.
Statistical analyses
First, continuous variables were tested for normality using the Shapiro-Wilk test. Normally distributed variables were reported as the mean ± standard deviation, while not normally distributed variables were described as median ± interquartile range. Categorical variables were expressed in absolute numbers and percentiles. Statistical differences between the two groups were analysed using t test, Mann Whitney U test or Chi-square test. Second, explorative analysis probed the association between education and score of neuropsychological tests including AVLT-H-N5, AVLT-H-N7, STT-A, STT-B, AFT, BNT and MoCA-B. Third, multiple linear regression analysis was used to quantify the influence of education on cognitive function with the score of neuropsychological tests as the dependent variable and education as independent variables. Age, education, gender were entered as independent variables. Regression diagnostics were performed to ensure the assumptions for linear regression were met. Residuals were normally distributed. Durbin-Watson test statistics indicated independence of observations and heteroscedasticity was in conformance with test assumptions (results not shown). All statistical analyses were performed in R (R version 3.6.3). The significant level was set at p< 0.05 (2-sided).