Subjective cognitive decline domain improves accuracy of plasma Aβ 42 /Aβ 40 for preclinical Alzheimer's disease diagnosis: The SILCODE study.

To the Editor: Accumulating evidence has shown that the preclinical stage of Alzheimer's disease (AD) (i.e., asymptomatic amyloidosis) lasts for decades before the onset of cognitive symptoms, providing a large window for early intervention. Amyloid pathology, the earliest pathological change associated with AD, can be detected in vivo with cerebrospinal fluid (CSF) analysis or positron emission tomography (PET), and its presence is necessary for the diagnosis of preclinical AD (pre-AD). However, both PET scans and CSF analyses are expensive, hampering their use in large-scale screening. Thus, blood-based biomarkers are desirable alternatives, as they are cost-effective and not invasive. Plasma β-amyloid (Aβ) biomarkers, and particularly the Aβ42/Aβ40 ratio, are logical blood-based candidates, as recent studies have shown decreased Aβ42/Aβ40 ratios along the AD continuum.[1] Several studies have also demonstrated the predictive value of plasma Aβ in combination with other variables. However, magnetic resonance imaging (MRI)-based measures of gray matter atrophy are a neurodegenerative AD biomarker, as cortical atrophy can predict subsequent cognitive decline and AD conversion.[2] Recent data have also demonstrated an association between central amyloid and the memory and language domains of the subjective cognitive decline interview (SCD-I).[3] Thus, in this study, we aimed to (1) investigate the plasma Aβ alterations in pre-AD and validate the diagnostic efficiency of plasma Aβ for detecting brain amyloidosis in peripheral pre-AD; and (2) explore the additive effects of subjective cognitive decline (SCD) features and gray matter volume in the plasma Aβ model. A total of 88 right-handed Han Chinese subjects (57 normal control [NC] and 31 pre-AD) were selected from 617 subjects who were enrolled in the Sino Longitudinal Study on Cognitive Decline (SILCODE) study between March 2017 and September 2018 [Supplementary Figure 1, https://links.lww.com/CM9/B722]. All participants had complete clinical examinations, underwent neuropsychological screening and blood sample collections, and had PET and sMRI scans. The inclusion criteria for pre-AD were: (1) asymptomatic amyloidosis; (2) no typical/atypical AD symptoms; (3) normal neuropsychological test performance; and (4) not meeting mild cognitive impairment (MCI)/dementia criteria. NC included those with negative PET scans and normal neuropsychological test results. The exclusion criteria were: (1) severe depression/anxiety; (2) other cognitive decline-causing conditions; (3) history of psychosis or mental growth slowing; and (4) cognitive decline from brain injury. Subjective cognitive functions were evaluated using the SCD-I developed by the Multicenter German Center for Neurodegenerative Diseases (DZNE)-Longitudinal Cognitive Impairment and Dementia Study (DELCODE). It included information about all five SCD-plus features (decline in memory, onset time, concerns, comparison with others, and informant confirmation) in five domains (memory, language, attention, planning, and others). In the present study, we focused on the involved domains. Objective cognitive functions were evaluated using a battery of neuropsychological tests. The Auditory Verbal Learning Test (AVLT)-HuaShan version was used for memory assessment; the semantic Verbal Fluency Test and the Boston Naming Test were used to assess language functions; and the Shape Trails Tests-A and B were used to assess executive function. The Hamilton Depression Rating Scale (HAMD) and the Hamilton Anxiety Rating Scale (HAMA) were also administered to all participants. Blood sample collection, storage, and analysis of the plasma Aβ42/Aβ40 ratio and amyloid-PET (AV45-PET) and MRI acquisition parameters, processing methods, calculation and analysis of standard uptake value rate (SUVR) values are shown in Supplementary Materials, https://links.lww.com/CM9/B722. Demographics, neuropsychological characteristics, and SCD-I domains were compared using t-tests and chi-squared tests. General linear models (GLMs) were controlled for age, sex, education, and apolipoprotein E (ApoE) ε4 status to compare plasma Aβ and brain volumes. Parietal correlation analyses were adjusted for age, sex, and education assessed associations between plasma Aβ and amyloid deposition and gray matter volume. Logistic regression models were used to compare four classification models, and receiver operating characteristic (ROC) curves were used for the evaluation of performance. The Youden Index determined the plasma Aβ42/Aβ40 cutoff value. P <0.05 was considered to be statistically significant. The demographic and clinical characteristics of all 88 participants, as well as between-group differences, were summarized in Supplementary Table 1, https://links.lww.com/CM9/B722. The NC and pre-AD groups had no differences in age, sex, education, ApoE ε4 positivity, or neuropsychological tests. The prevalence and group differences in SCD domains were shown in Supplementary Table 2, https://links.lww.com/CM9/B722. Both groups reported complaints in memory, executive, and language domains, with pre-AD reporting more memory decline (90.3% [28/31] vs. 66.7% [38/57], P = 0.014) and less language decline (9.7% [3/31] vs. 35.1% [20/57], P = 0.010). No group differences were found in planning, attention, or other domains. ROC analysis only used in memory and language domains. Figure 1A showed the group differences in plasma Aβ levels and their associations with brain amyloid deposition. Compared to the NC group, the pre-AD group had significantly lower Aβ42 (GLM F = 14.827, P = 0.002; 12.51 ± 3.89 pg/mL in NC vs. 10.38 ± 2.46 pg/mL in pre-AD) and Aβ42/Aβ40 (GLM F = 9.104, P <0.001; 0.017 ± 0.0006 in NC vs. 0.014 ± 0.0005 in pre-AD) levels. No significant difference was found in plasma Aβ40 levels between two groups (t = 0.228, P = 0.710). Additionally, lower plasma Aβ42/Aβ40 ratios were associated with higher global SUVR (ρ = –0.228, P = 0.037) [Figure 1B]. Plasma Aβ42 was negatively associated with global SUVR at the trend level (ρ = –0.215, P = 0.050) [Figure 1B], but no correlation was found between plasma Aβ40 and SUVR (ρ = –0.008, P = 0.954).Figure 1: (A)Plasma Aβ42 level and plasma Aβ42/Aβ40 ratios in different groups. Plasma Aβ42 and Aβ40 levels were quantified by Mesoscale Diagnostics in 57 NC and 31 pre-AD patients. Analyses were adjusted for age, gender, education, and ApoE ε4 status.(B) Correlations between SUVR and Aβ42/Aβ40 ratios, plasma Aβ42 levels. (C) Global gray matter volume in different gruops. Analyses were adjusted for age, gender, and education.(D) Correlations between global gray matter volume and SUVR. (E) The ROC curves of different features diagnostic model. AD: Alzheimer's disease; ApoE: Apolipoprotein E; AUC: Area under the curve; NC: Normal control; pre-AD: Preclinical AD; ROC: Receiver operating characteristic; SCD: Subjective cognitive decline; SUVR: Standard uptake value rate.Gray matter volumes of the whole brain and the bilateral hippocampi were compared among the two groups after controlling for age, sex, and education. The pre-AD group showed significantly decreased global gray matter volumes compared to the NC group (t = –2.648, P = 0.010) [Figure 1C], but there were no differences in bilateral hippocampal volumes (left hippocampi: t = 1.361, P = 0.177, right hippocampi: t = 0.705, P = 0.483). Partial correlation analysis, which controlled for age, sex, and education, showed a significant, negative association between global gray matter volume and SUVR (ρ = –0.225, P = 0.039) [Figure 1D], while no association was found between hippocampal volumes and SUVR (ρ = 0.343, P >0.1). We trained logistic regression models to classify pre-AD and NC patients using Aβ characteristics as well as global gray matter volume in combination with the two selected SCD domains. Classification performance was evaluated using ROC curves [Figure 1E]. Global gray matter volume could discriminate between pre-AD and NC patients (area under the curve [AUC] = 0.642, 95% confidence interval [CI] = 0.520–0.765), as could plasma Aβ42/Aβ40 ratios (AUC = 0.762, 95% CI = 0.656–0.868). Youden's cut-offs for plasma Aβ42/Aβ40 ratios and gray matter volume were 0.0145 with a sensitivity of 71% and specificity of 79%, and 389.66 cm3 with a sensitivity of 61% and specificity of 67%. Using the cut-off value of 0.0145 for the plasma Aβ42/Aβ40 ratio, the number of patients receiving AV45-PET would be reduced by 57%. Combining the Aβ42/Aβ40 ratio with the SCD domains also significantly improved the classification performance (AUC = 0.835, 95% CI = 0.748–0.922), which was further slightly improved when combined with global gray matter volume (AUC = 0.839, 95% CI = 0.754–0.925). Our study suggested using memory and language SCD domains increased plasma Aβ42/Aβ40 ratio model accuracy in distinguishing pre-AD from NC. SCD-I, especially in memory and language domains, is linked to lower CSF Aβ42 levels and higher amyloid deposition, indicating its potential as an early AD indicator. SCD is associated with tauopathy independent of global Aβ burden, adding to the predictive power of plasma Aβ42/Aβ40 ratio. Our data validate plasma Aβ42/Aβ40 ratio as a pre-AD screening biomarker in Chinese patients (AUC = 0.762). Our findings are comparable with previous results from the FACEHBI study (AUC = 0.681)[4] and the SCIENCe project (AUC = 0.680).[5] Atrophy has been defined as a biomarker of neurodegeneration following amyloid pathology in AD. In line with previous studies, we found decreased global gray matter volume in pre-AD patients compared to controls. However, there were no significant differences found in bilateral hippocampal volumes between the two groups. Our previous study in individuals at risk for AD also showed reduced global surface areas and cortical volumes but no differences in network or regional scales in patients at risk for AD compared to NC. These findings indicate that changes in gray matter volume in the preclinical stages may be slight but extensive. Limitations include a small cross-sectional sample and exclusion of other diseases that may cause cognitive decline. Plasma Aβ levels may be linked to other comorbidities, and their ability to distinguish AD from other dementias requires further investigation. Pathologic tau biomarkers such as CSF levels of phosphorylated tau or tau PET were not examined. In conclusion, our results suggest that the plasma Aβ42/Aβ40 ratio could reflect central amyloid AD pathology in cognitively normal elderly Chinese subjects, especially when combined with SCD characteristics and neurodegenerative biomarkers. Funding This work was supported by grants from the National Natural Science Foundation of China (Nos. 82020108013 and 82001773), National Key Research and Development Program of China (No. 2022YFC24069004), Beijing Brain Initiative from Beijing Municipal Science & Technology Commission (Z201100005520018), and STI2030-Major Projects (No. 2022ZD0211800). Conflicts of interest None.