Participants

Clinical evaluation and procedures

Cognitive function was assessed at baseline using standard procedures. The Mini-Mental State Examination (MMSE) was used to assess global cognition [28]. Clinical Dementia Rating (CDR) scores were used to represent the severity of dementia [29, 30]. We also performed a memory test using the delayed recall of 12 items [31, 32], a modified Boston naming test [33], and a category verbal fluency test to assess short-term memory and language function [34]. Forward and backward digit-span tests were used to assess attention and working memory [35]. The NPI questionnaire was administered to assess the frequency and severity of BPSD [1, 2]. The NPI agitation/aggression, mood, and frontal symptoms subscales were used in this study [3]. The NPI agitation/aggression subscale includes agitation/aggression, disinhibition, irritability, and aberrant motor behavior items. The NPI mood subscale includes depression, anxiety, and irritability items. The NPI frontal subscale includes apathy, disinhibition, irritability, and euphoria items. Each subscale score was calculated by summing the scores for the severity of each item and then multiplying the summed score by the duration. Follow-up examinations for AD patients were performed at 1 and 2 years. At each follow-up, we re-evaluated patient cognition, BPSD, and disease severity using the MMSE, NPI, and CDR scores. All tests were administered by trained individuals in a clinical setting.

DNA and plasma clusterin level analyses

Blood samples were collected in tubes containing sodium EDTA as an anticoagulant. Following centrifugation, plasma was aliquoted into polypropylene tubes and stored at –80 °C until analysis. Genomic DNA was isolated from whole blood using the Gentra Puregene kit according to the manufacturer’s protocols (Qiagen, Hilden, Germany). The presence of the ε2, ε3, and ε4 alleles of the apolipoprotein gene (ApoE) was determined by assessing the sequences at two single-nucleotide polymorphisms (SNPs; rs429358 and rs7412) [36, 37]. An ApoE4 gene carrier was defined as having at least one allele containing the ε4 gene. The plasma Aβ peptide assay was performed using the INNO-BIA plasma Aβ forms assay (Innogenetics, Ghent, Belgium), which is based on the multiplex xMAP technique, with a LABScan-100 system (Luminex BV, Netherlands). The ratio of plasma Aβ1–42/Aβ1–40 was calculated as the plasma Aβ1–42 level divided by the plasma Aβ1–40 level. There was a significant difference in the ratio of plasma Aβ1–42/Aβ1–40 between patients with AD and controls [38]. Each sample was assessed in duplicate. The plasma clusterin levels were measured using a Human Clusterin Quantikine ELISA Kit (R&D Systems, Minneapolis, MN, USA) according to the manufacturer’s instructions. All samples from each participant were measured on the same plate to avoid interplate variation. The average intra-assay coefficient of variation based on duplicate samples was 1.2% for clusterin, 7.6% for Aβ1–40, and 9.1% for Aβ1–42. The average inter-assay coefficient of variation, as determined using three samples (low, medium, and high measurement range) that were analyzed on 4 different days, was 3.3% for clusterin, 12.0% for Aβ1–40, and 15.8% for Aβ1–42.

Statistical analysis

All statistical analyses were performed using SPSS (version 21.0). Continuous variables are expressed as the means ± standard deviations. Categorical variables are presented as numbers and ratios. Independent t tests were performed to compare the ages, years of education, BMI, and MMSE scores of patients. Chi-square tests were used to compare AD patients and controls in terms of gender and ApoE4 carrier status. The analysis of variance (ANOVA) test and Kruskal-Wallis test were used to compare MMSE and NPI subscale scores with different levels of AD severity. We classified baseline plasma clusterin and Aβ levels into tertiles (low, median, and high) to study mean changes in MMSE scores at the 1-year follow-up (MMSE at 1-year subtracted from MMSE at baseline). One-way ANOVA tests were used to compare group differences. To investigate the relationship between plasma biomarkers and MMSE and NPI subscale scores during longitudinal follow-up, the generalized estimating equation (GEE) method was used to analyze longitudinally correlated data. Baseline plasma clusterin levels and the plasma Aβ1–42/Aβ1–40 ratio in patients with AD were then stratified into tertiles (low, median, and high). Analyses were performed using the GEE method because it accounts for variations in time and identifies correlations among repeat measurements that are collected during a longitudinal study design. In addition, the MMSE and NPI subscale scores did not display a normal distribution. For the GEE analyses, we used a robust estimator for the covariance matrix and selected autoregressive (1) for the working correlation matrix. Age, years of education, gender, and ApoE4 gene status were used to adjust the MMSE and NPI subscale scores. Interactions between groups, time, and ApoE4 carrier status were also analyzed by GEE. Estimated β values with 95% confidence intervals (CIs) were calculated for all the blood biomarkers. Statistical significance was defined as p < 0.05.

Demographic and clinical data

In total, 322 AD patients (173 males and 149 females; mean age 80.4 ± 6.2 years; mean education 10.1 ± 4.4 years) and 88 controls (54 males and 34 females; mean age = 77.0 ± 6.2 years; mean education = 10.1 ± 5.1 years) were recruited. Controls were significantly younger than patients in the AD group (p < 0.01). There were no significant differences between the groups in terms of gender or years of education. MMSE scores differed significantly between patients with AD and controls (Table 1; p < 0.01). Compared with controls, a significantly higher proportion patients with AD were ApoE4 carriers (p < 0.01). The plasma Aβ1–40 levels were significantly higher in patients with AD compared with controls (Table 1; p = 0.02). The plasma Aβ1–42 and clusterin levels were similar between groups (p = 0.23 and 0.29, respectively). To study the differences in the plasma Aβ and clusterin levels between patients with AD and controls, regression analyses were performed to adjust for the effects of age, gender, ApoE4 carrier status, and years of education. Our results showed no significant differences between groups in the plasma Aβ1–40 or Aβ1–42, the Aβ1–42/Aβ1–40 ratio, or clusterin levels after correcting for multiple covariates (p = 0.15, 0.17, 0.69, and 0.29, respectively).

Baseline characteristics for the complete study groups and the missing data group

Ultimately, 244 AD patients (75.8%) completed the 1-year clinical follow-up; 166 AD patients (51.5%) completed the 2-year clinical follow-up. At the 1-year follow up, the baseline characteristics (age, gender, years of education, ApoE4 carrier status, MMSE scores, CDR stage, NPI subscale scores, plasma clusterin, and plasma Aβ levels) were similar in both groups. At the 2-year follow-up, patients who did not complete the study were significantly older than those who participated in follow-ups (81.2 ± 5.8 vs. 79.6 ± 6.5 years, p = 0.02) and had lower MMSE scores (17.2 ± 6.1 vs. 19.2 ± 5.2, p < 0.01). There was no significant difference in ApoE4 carrier status, CDR stage, NPI subscale scores, plasma clusterin level, or Aβ level between the groups.

Plasma biomarkers and dementia severity

Baseline plasma biomarkers and ApoE4 carrier status in AD

The Kruskal-Wallis test revealed that patients with severe AD were significantly more likely to carry the ApoE4 allele (Table 2; p < 0.01). Among the patients who were categorized as ApoE4 positive or negative, there was no significant difference in plasma levels of Aβ1–40, Aβ1–42, or the Aβ1–42/Aβ1–40 ratio. However, the plasma clusterin levels were significantly higher in the ApoE4-positive group (254.4 ± 33.8 μg/ml vs. 244.4 ± 40.2 μg/ml, p = 0.02). ApoE4 carrier status was used as a covariate in the longitudinal follow-up study.

Plasma clusterin and Aβ levels at baseline and the 1-year follow-up

Based on the baseline plasma clusterin levels, we stratified patients into high (mean = 289.4 ± 22.6 μg/ml), medium (mean = 246.9 ± 8.1 μg/ml), and low (mean = 209.3 ± 23.4 μg/ml) tertiles to determine whether there was an association with baseline MMSE scores by one-way ANOVA. There was a significant group difference in MMSE scores (mean MMSE scores were 17.5 ± 6.3 in the high tertile, 18.1 ± 5.6 in the medium tertile, and 19.4 ± 4.6 in the low tertile; p = 0.04). Post-hoc analysis showed a difference between the high and low tertiles (p < 0.05). We further studied mean changes in MMSE scores at the 1-year follow-up. A significant within-group difference was found by one-way ANOVA (mean changes in MMSE scores at 1 year were −2.5 ± 3.4 in the high tertile, −2.0 ± 4.5 in the medium tertile, and −0.9 ± 3.5 in the low tertile; p = 0.02). Post-hoc analysis showed a significant difference between the high and low tertiles (p = 0.008). This suggests an association between higher plasma clusterin levels and lower cognitive status in patients with AD. For the analysis of plasma Aβ, Aβ1–40, Aβ1–42, and the Aβ1–42/Aβ1–40 ratio, patients were stratified into tertiles and we compared the mean changes in the MMSE scores at the 1-year follow-up using a one-way ANOVA. There were no significant differences in mean changes in MMSE scores among the tertiles of Aβ1–40, Aβ1–42, or the Aβ1–42/Aβ1–40 ratio (p = 0.77, 0.15, and 0.43, respectively).

Plasma clusterin levels at the 2-year follow-up

Longitudinal data collected over 2 years were analyzed by GEE adjusted for age, years of education, gender, time, and ApoE4 carrier status. This revealed that the MMSE scores were significantly decreased in patients in the high tertile of plasma clusterin levels (average decrease in MMSE scores after 2 years was −2.09 points, 95% CI = −3.67 to −0.51, p = 0.01) compared with patients in the low tertile (Fig. 1). This MMSE difference was the same difference as observed at baseline. No significant interactions were detected between the tertiles of plasma clusterin and time or between ApoE4 status and time (p = 0.10 and p = 0.16, respectively). MMSE scores were significantly lower among patients carrying the ApoE4 allele than those without (−2.75, 95% CI = −4.11 to −1.39, p < 0.01). Because plasma clusterin was significantly higher in the ApoE4-positive group, we performed an interaction analysis. No significant interaction between clusterin tertile and ApoE4 carrier status was detected (p = 0.93; Table 3). The plasma clusterin levels showed no association with agitation/aggression, mood, or frontal subscale NPI scores.

	β	SE	95% CI	P value
Gender
Male	−0.32	0.83	(−1.95, 1.32)	0.71
Female	Referent
Clusterin level
High	−2.09	0.81	(−3.67, −0.51)	0.01
Medium	−0.87	0.71	(−2.27, 0.53)	0.22
Low	Referent
ApoE4 carrier
Positive	−2.75	0.69	(−4.11, −1.39)	0.00
Negative	Referent
Age	0.03	0.08	(−0.12, 0.17)	0.74
Education years	0.13	0.09	(−0.04, 0.30)	0.15
Time	−0.91	0.21	(−1.33, −0.49)	0.00

Adjusted for age, education years, gender, time, and ApoE4 carrier

Interaction in group*ApoE4 carrier: p value = 0.93 and group*time: p value = 0.21

AD Alzheimer’s disease, ApoE apolipoprotein E, CI confidence interval, MMSE Mini-Mental State Examination, SE standard error

Longitudinal analysis of plasma Aβ

To explore the predictive value of plasma Aβ for cognition and BPSD, we stratified patients by the plasma Aβ1–42/Aβ1–40 ratio into high (mean = 31.9 ± 10.2%), medium (mean = 22.6 ± 1.4%), or low (mean = 16.7 ± 2.5%) groups. After adjustment for age, years of education, gender, time, and ApoE4 carrier status, GEE analysis revealed no association between the plasma Aβ1–42/Aβ1–40 ratio and the decline in the MMSE score at 2 years (p = 0.12). GEE analysis showed that the plasma Aβ1–42/Aβ1–40 ratio and ApoE4 carrier status were significant predictors of the NPI agitation/aggression score (p = 0.04 and 0.009, respectively). GEE analysis also revealed a significant interaction between the plasma Aβ1–42/Aβ1–40 ratio and ApoE4 carrier status (p = 0.003). Next, we stratified patients with AD based on ApoE4 carrier status (135 ApoE4 carriers and 187 non-carriers). After 2 years, ApoE4 carriers in the high plasma Aβ1–42/Aβ1–40 ratio tertile showed a significant increase in the NPI agitation/aggression score compared with those in the low tertile (6.06, 95% CI = 1.20–10.62, p = 0.02; Table 4, Fig. 2a). No such association was found among ApoE4-negative patients (Fig. 2b). The plasma Aβ1–42/Aβ1–40 ratio was not a significant predictor of NPI-mood or NPI-frontal scores. GEE analysis revealed no significant association between plasma Aβ1–42 level and MMSE or NPI subscale scores at 2 years.

	β	SE	95% CI	P value
Gender
Male	0.16	2.25	(−4.25, 4.57)	0.94
Female	Ref
Ratio of Aβ1–42/Aβ1–40
High	6.06	2.48	(1.20, 10.62)	0.02
Medium	1.20	2.24	(−3.18, 5.59)	0.59
Low	Ref
Age	0.18	0.14	(−0.10, 0.46)	0.21
Education years	0.11	0.17	(−0.23, 0.45)	0.51
Time	−0.96	0.70	(−2.34, 0.42)	0.17

Adjusted for age, education years, gender, and time

Aβ amyloid beta peptide, AD Alzheimer’s disease, CI confidence interval, NPI Neuropsychiatric Inventory, SE standard error

Plasma clusterin levels and cognitive decline

At present there is no reliable biomarker for predicting the speed of AD progression [41, 42]. The plasma level of clusterin could be a prognostic marker for patients with AD, but the specific relationship between the clusterin level and cognitive decline remains to be determined. A previous study reported that higher concentrations of plasma clusterin were significantly associated with an accelerated cognitive decline in patients with AD after 1 year of follow-up [23]. In patients with MCI, higher baseline levels of plasma clusterin have been associated with lower rates of brain atrophy. The authors of that study proposed that increased clusterin levels could protect at-risk subjects; failure of this mechanism in patients with AD would account for a more rapid decline in cognition and brain structural changes [43]. Another study reported conflicting results, showing that higher clusterin levels were related to less cognitive decline in patients with AD [22]. The results from our 2-year follow-up showed that higher levels of plasma clusterin were not associated with a more rapid cognitive decline in AD, which indicates that plasma clusterin levels are related to cognitive state and disease severity in AD [19, 44].

Many confounding factors may contribute to plasma clusterin levels. Recent studies have shown that plasma clusterin levels can be affected by gender, obesity, systemic inflammation, or atherogenic components of the lipid profile in AD patients [45, 46]. Plasma clusterin levels also play a role in tissue remodeling and immune responses in the brain, which may be related to the speed of deterioration in AD patients [45–48]. Further research is necessary to explore a potential role for clusterin in response to tissue damage and how it could influence the progression of AD. In practice, interpreting the plasma clusterin level is required to determine the disease state of a subject.

Plasma Aβ predicts agitation/aggression symptoms in ApoE4 carriers

In our current work, we found that the plasma Aβ1–42/Aβ1–40 ratio at baseline could predict the increase in NPI agitation/aggression scores among ApoE4 carriers. Agitation and aggression symptoms are difficult to manage and are especially distressing for patients, caregivers, and medical staff members. In a previous study, the CSF level of Aβ1–42 was associated with agitation and irritability symptoms in patients with MCI [49]. β-Amyloid precursor protein, a precursor for Aβ, has also been linked to aggression symptoms in young children with autism [50]. The ApoE4 gene also contributes to aggression symptoms in patients with AD [51, 52]. Our findings suggest that plasma Aβ may act synergistically with the ApoE4 gene to aggravate agitation/aggression symptoms in patients with AD because the ApoE4-negative group did not show this association. Additional large, cohort studies with longitudinal follow-ups will be needed to confirm this finding.

Plasma Aβ and cognitive changes in AD

Our study found no significant differences in the plasma Aβ levels between patients with AD and controls after adjustment for multiple covariates, nor was there a longitudinal association with MMSE changes in patients with AD. Previous studies have shown increased plasma Aβ1–40 or Aβ1–42 levels in patients with AD [7, 53], but negative studies have also been reported [54, 55]. Longitudinal studies on the association between the plasma Aβ levels and cognitive changes have also shown both positive [11, 56] and negative results [4]. This variation in studies on plasma Aβ may be the result of different collection methods [57], assay techniques [58], and multiple confounding factors such as diabetes, hypertension, or medication use [59, 60]. A recent study demonstrated an association between chemically treated plasma Aβ and the deposition of brain amyloid, which may provide a potential blood-based biomarker for the study of AD [61].

Limitations

There are several limitations to the current study. Over the 2 years of follow-up, patients exhibited disease progression, cognitive decline, and an increased severity of neuropsychiatric symptoms, such as agitation and aggression [62]. This could be part of the natural course of the disease; MMSE and NPI symptoms associated with AD decline at different rates depending on the severity of AD. Although levels of plasma clusterin did not differ significantly among patients with different levels of AD severity, we cannot exclude such a possibility. Other confounding factors, such as depression and agitation symptoms, have been reported to increase the plasma clusterin levels, and the plasma amyloid levels may be affected by genetic variants and creatinine levels [47, 63, 64]. We acknowledge this limitation, and consider both plasma biomarkers to be predictors of clinical phenotype, rather than exhibiting a direct causal relationship. Second, we did not include other plasma biomarkers or imaging data in this study. CSF Aβ1–42 levels or brain amyloid scans could provide more reliable pathological markers than plasma biomarkers for the diagnosis of AD [65]. Measurements of structural changes in brain regions could be a robust marker, although it is more expensive than plasma biomarkers. Our previous cross-sectional study showed that the score for right-side posterior atrophy, as measured by brain MRI, was associated with NPI agitation/aggression symptoms [66]. A longitudinally repeated MRI measurement could be useful for linking plasma biomarkers with brain morphology and BPSD. Third, our drop-out rate at the 2-year follow-up was 49%, and the missing data group had an older age and lower MMSE scores than did the complete study group. Although there were no differences in plasma clusterin or Aβ levels between groups, we cannot exclude the possibility of bias caused by our missing data. A study with a larger sample size and a higher follow-up rate may improve our findings.