Participants
We enrolled 126 participants with normal cognition (NC) and 208 patients with cognitive impairment (AD and/or LBD). The NC participants lacked any subjective cognitive impairment symptoms or history of neurologic or psychiatric illnesses and underwent neurological and neuropsychological examination, brain magnetic resonance imaging (MRI), and APOE genotyping. Their neurological and neuropsychological findings were normal, and they lacked structural brain lesions. Eleven NC participants underwent 18F-Florbetaben (FBB)-PET, which did not indicate significant β-amyloid deposition.
The patients with AD and/or LBD underwent neurological examination, APOE genotyping, neuropsychological tests, 3T MRI, fluorodeoxyglucose (FDG)-PET, and FBB-PET scans at the dementia and movement clinics of Yonsei University Severance Hospital, Seoul, Korea, between April 2012 and May 2019. Using semi-structured questionnaires, caregivers evaluated the presence of clinical features of AD, including slow-progressive memory dysfunction, and those of LBD, including parkinsonism, rapid eye movement sleep behavior disorder, visual hallucinations, and cognitive fluctuation. Parkinsonism severity was assessed using the Movement Disorder Society Unified Parkinson’s Disease Rating Scale (UPDRS) motor score, with a score of >16 indicating moderate severity. The exclusion criteria were (1) pure vascular cognitive impairment; (2) other degenerative dementia causes, including frontotemporal dementia, corticobasal degeneration, and progressive supranuclear palsy; (3) drug-induced cognitive impairment; and (4) other adequate cognitive impairment causes, including epilepsy, psychiatric disorder, normal-pressure hydrocephalus, and structural brain lesions (e.g., tumor or hemorrhage).
All patients with AD dementia met the criteria for probable AD dementia with high levels of biomarker evidence [13], while all patients with mild cognitive impairment (MCI) due to AD met the criteria for a high likelihood of MCI due to AD based on the National Institute on Aging-Alzheimer’s Association workgroups guidelines for AD [23]. These patients were regarded as having “typical AD.” Specifically, all patients with typical AD had progressive memory problems with insidious onset, biomarker evidence of neuronal injury based on FDG-PET, and significant cerebral β-amyloid deposition confirmed by a global FBB standardized uptake value ratio (SUVR) of >1.478 [20].
All patients with Parkinson's disease (PD) satisfied the United Kingdom PD Brain Bank diagnostic criteria [24] and presented with decreased dopamine transporter uptake on 18F-fluorinated N-3 fluoropropyl-2-beta-carboxy-methoxy-3-beta-(4-iodophenyl) nortropane (FP-CIT)-PET. Patients with PD-MCI and PD dementia (PDD) met the Movement Disorder Society criteria for PD-MCI [25] and probable PDD [26], respectively. All patients with DLB fulfilled the 2017 revised criteria for probable DLB [14] and exhibited decreased dopamine transporter uptake on FP-CIT-PET. To identify early brain changes in patients with DLB, we included patients with MCI [27] who met all of the diagnostic criteria for probable DLB, except the presence of dementia. These patients were regarded as having LBD-related cognitive impairment (LBCI).
Based on clinical features and biomarker evidence, patients with mixed disease were categorized into the AD-dominant, LBD-dominant, and equally dominant mixed disease subgroups. Patients with LBCI who presented with β-amyloid deposition, but not memory problems, as their chief complaint, were considered to have LBD-dominant mixed disease or DLB with amyloid deposition. Conversely, patients with typical AD who presented with moderate/severe parkinsonism and abnormal FP-CIT-PET but lacked other LBD features, including cognitive fluctuation and visual hallucination, were considered to have AD-dominant mixed disease or LB variant of AD (LBVAD) [28]. If patients with typical AD satisfied the diagnostic criteria for DLB, they were regarded to have an equally dominant mixed disease or AD/DLB. Finally, there were 57, 32, 56, 21, and 42 patients with pure AD (PAD), LBVAD, pure LBD (PLBD), DLB with amyloid deposition (DLBA), and AD/DLB, respectively. Patients with PAD and LBVAD were considered to have typical AD; patients with PLBD and DLBA were considered to have typical LBD; and patients with AD/DLB were considered to have both typical AD and typical LBD.
AD-related cognitive impairment (ADCI) was defined to include patients with PAD, LBVAD, AD/DLB, and DLBA. Given the progressive nature of AD [29], we defined ADCI solely based on amyloid PET-positivity and differentiated it from typical AD that corresponds to the symptomatic AD stage. Similarly, LBCI was defined to include patients with LBVAD, AD/LBD, DLBA, and PLBD to differentiate it from typical LBD corresponding to LBD-dominant disease.
APOE genotyping
Genomic DNA was extracted from peripheral blood leukocytes using the DiaPlexQTM ApoE Genotyping Kit following the manufacturer’s instructions (SolGent co., Ltd.). Two single nucleotide polymorphisms (rs429358 for codon 112 and rs7412 for codon 158) in the APOE gene were genotyped using CFX 96 Real-time PCR system (Bio-Rad) following the manufacturer’s instructions.
MRI acquisition
All MRI scans were acquired using the same 3T MRI scanner (Philips Achieva; Philips Medical System, Best, The Netherlands) with a SENSE head coil (SENSE factor = 2). A high-resolution, T1-weighted MRI volume dataset was obtained for all participants with a three-dimensional T1-turbo field echo sequence configured with the following acquisition parameters: axial acquisition with a 224 × 224 matrix; 256 × 256 reconstructed matrix with 182 slices; 220 mm field of view; 0.98 × 0.98 × 1.2 mm3 voxels; echo time of 4.6 ms; repetition time of 9.6 ms; flip angle of 8°; and slice gap of 0 mm. Moreover, axial fluid-attenuated inversion recovery images were obtained to evaluate white matter hyperintensity (WMH) using the following parameters: matrix, 224 × 224; section thickness, 1 mm; echo time, 335 ms; repetition time, 8000 ms; and flip angle, 90°.
Regional WMH measurement and lacune counting
A visual rating scale of WMHs was modified from the Fazekas scale [30]. Periventricular WMH (PWMH) areas were classified as P1 (cap and band <5 mm), P2 (cap ≥5 mm or band <10 mm), and P3 (cap or band ≥10 mm); deep WMH (DWMH) areas were classified as D1 (maximum diameter of deep white matter lesion <10 mm), D2 (10 mm≤ lesion <25 mm), and D3 (lesion ≥25 mm). The number of lacunes was determined as previously described [31]. Manual ratings of WMHs and lacunes were performed by three blinded neurologists (J.H.J., K.B., S.H.J.).
Acquisition of 18F-FBB- and 18F-FP-CIT-PET imaging
PET scans were obtained using a Discovery 600 system (GE Healthcare, Milwaukee, WI). Doses of 185 MBq (5 mCi) of FP-CIT, 300 MBq (8 mCi) of FBB, and 4.1 MBq per body weight (kg) of FDG were intravenously injected during the procedure. At 90 min after the injection, images were acquired during a 20 min session after a computed tomography (CT) scan for attenuation correction. The parameters for the spiral CT scan were as follows: 0.8 s per rotation at 120 kVp, 10 mA, 3.75 mm slice thickness, 0.625 mm collimation, and 9.375 mm table feed per rotation. The images were reconstructed using the ordered subset expectation maximization algorithm with four iterations and 32 subsets. A Gaussian filter with 4 mm full-width at half maximum (256 × 256 matrix with 0.98 mm pixels and 0.98 mm slice thickness) was applied to reconstructed PET images.
Assessment of 18F-FBB- and 18F-FP-CIT-PET images
Quantitation of 18F-FBB-PET/CT images was based on surface-based PET analysis methods. First, we processed all T1-weighted MR images using the CIVET pipeline (http://mcin.ca/civet) to classify gray/white matter tissues and extract cortical surfaces. Subsequently, we co-registered the FBB-PET scans to individual T1-weighted images using rigid-body transformation. We performed partial volume correction within gray and white matter regions using the idSURF method [32]. Next, the corrected FBB values were normalized to the crus-I/II gray matter reference region, which yielded an SUVR. Finally, we extracted the global FBB SUVR value as the cortical volume-weighted average of the following cortical regions of interest: frontal, anterior/posterior cingulate, lateral parietal, and lateral temporal cortices. FP-CIT-PET was interpreted using visual ratings as previously described [33]. FBB-PET was regarded as amyloid-positive if the global FBB SUVR was >1.478, as described in a previous autopsy-validation study [20].
Quality assurance for image processing
All MR images and processing results were visually inspected by three researchers blinded to participant information (J.H.J., S.J., B.S.Y.) for quality assurance.
Neuropsychological evaluation
All participants were assessed using the standardized Seoul Neuropsychological Screening Battery, which assesses attention, language, visuospatial function, memory, and frontal/executive function [34, 35]. Standardized z scores were available for all scorable tests based on age- and education-matched norms. The following tests were performed for analyses: the digit span backward for the attention domain; Korean version of the Boston Naming Test for the language domain; copying item of the Rey–Osterrieth Complex Figure Test (RCFT) for the visuospatial domain; immediate recall, 20 min delayed recall, and recognition items of the RCFT and Seoul Verbal Learning Test for the memory domain; and semantic and phonemic Controlled Oral Word Association Test and Stroop color reading test for the frontal/executive domain. Moreover, global cognitive performance was assessed using the Korean version of the Mini-Mental State Examination (K-MMSE) and Clinical Dementia Rating Sum of Boxes (CDR-SOB).
Statistical analysis
An analysis of variance and the c2 test were performed for cross-group comparisons of clinical features. Logistic regression analyses were performed to evaluate the effect of APOE4 (carrier vs. non-carrier) on disease risk after controlling for age, sex, education, hypertension, diabetes mellitus, hyperlipidemia, DWMH, PWMH, and the lacune number. Model 1 analyses evaluated the APOE4 effect on the risk of each disease (PAD, LBVAD, AD/LBD, DLBA, PLBD, ADCI, LBCI, typical AD, or typical LBD) in a combined NC and each disease group. Model 2 analyses evaluated the effect of APOE4 on the risk of ADCI and LBCI in all participants after adjusting for LBCI and ADCI, respectively. Model 3 analyses evaluated the effect of APOE4 on the risk of typical AD and typical LBD in all participants after adjusting for typical LBD and typical AD, respectively.
To determine the effect of APOE4 on β-amyloid deposition, we used general linear models to investigate the independent and interactive effects of APOE4, typical AD, and typical LBD on the global and mean FBB SUVR in the frontal, temporal, parietal, and occipital cortices with the same covariates as the logistic regression analyses. Model 1 analyses evaluated the independent effects of APOE4, typical AD, and typical LBD. Model 2 analyses tested the significance of each pair of interaction terms, including APOE4 * typical AD, APOE4 * typical LBD, and typical AD * typical LBD. Model 3 analyses used APOE4, typical LBD, typical AD, and significant interaction terms from Model 2 as predictors.
To determine the effect of APOE4 on cognitive dysfunction, the effects of APOE4, typical AD, and typical LBD on composite cognitive scores were evaluated using general linear models with similar covariates. Model 1 analyses evaluated the independent effects of APOE4, typical AD, and typical LBD. Given the significant interaction effects of AD and LBD on neuropsychological test scores, the interaction term of typical AD * typical LBD was further included in Model 1 analyses [22]. Model 2 analyses tested the significance of interaction terms, including APOE4 * typical AD and APOE4 * typical LBD. Model 3 analyses used APOE4, typical LBD, typical AD, typical AD * typical LBD, and significant interaction terms from Model 2 as predictors. Statistical analyses were performed using SPSS software (version 23.0; IBM Corp., Armonk, NY, USA) and significance was set at p<0.05.
MATLAB-based SurfStat toolbox was used for statistical analyses of vertex-wise FBB uptake [36]. To identify regional β-amyloid deposition patterns associated with APOE4, typical AD, and typical LBD, the independent and interactive effects of APOE4, typical AD, and typical LBD on the vertex-wise FBB SUVR were investigated using general linear models after adjustment for same covariates as for global and lobar FBB SUVRs. Given the significant interaction effects of APOE4 * typical LBD and typical AD * typical LBD on the mean lobar FBB SUVR, they were included in the models.