Plasma amyloid-β levels and prognosis in incident dementia cases of the 3-City Study

Abstract

Studies of plasma amyloid-β levels as potential biomarkers for incident Alzheimer’s disease (AD) have yielded contradictory results. We explored the associations between plasma Aβ_40, Aβ₄₂, truncated amyloid-β levels and prognosis of dementia in participants of the prospective 3C-Study. 120 aged individuals diagnosed with 2-year incident dementia were followed up for 7 years. The associations between amyloid-β plasma levels and baseline cognitive score, cognitive decline and death were examined. A higher level of baseline plasma amyloid-β was associated with worse cognitive status 2 years prior to incident dementia diagnosis. In incident AD patients, the association was only significant for Aβ₄₀ and Aβ_n-42. In the fast cognitive decliners group-especially in AD cases-a higher level of 5pg/ml of baseline Aβ_42, Aβ_n-42,Aβ_n-42/Aβ_n-40 and Aβ₄₂/Aβ₄₀ ratios were associated with a lower risk of fast cognitive decline based on the IST score. There was no association between peptide levels and mortality in demented subjects. When assayed at prodromal stage, plasma amyloid-β levels may be potentially useful markers of fast cognitive decline in individuals who subsequently become demented.

INTRODUCTION

It has been known for several years that low cerebrospinal fluid (CSF) levels of the 42-amino-acid amyloid-β fragment (Aβ₄₂) are strongly associated with current or future Alzheimer’s disease (AD) in patients with mild cognitive impairment (MCI) [1]. However, since CSF sampling is an invasive, time-consuming procedure, there has been increasing interest in establishing whether the plasma amyloid-β levels could also be relevant [2]. In this context, we and others have reported that a low plasma amyloid-β Aβ₄₂/Aβ₄₀ ratio is associated with an increased risk of dementia in two large, longitudinal studies [3, 4]. Even though the latter observation is still controversial and calls for further investigation, several lines of evidence suggest that plasma amyloid-β levels could be a marker of the short-term dementia risk. In particular: (i) the association of a low plasma Aβ₄₂/Aβ₄₀ ratio with AD risk is restricted to individuals diagnosed during the two years following the assay [3, 4]; (ii) Amyloid-β plasma levels are associated with the risk of conversion to dementia in MCI [5]; (iii) conversion to AD is accompanied by a concomitant decrease in the plasma Aβ₄₂/Aβ₄₀ ratio [6]. These various observations suggest that in addition to their potential as an indicator of short-term dementia risk, amyloid-β plasma levels may also be markers of disease progression and may therefore be useful in terms of prognosis. Even though the prognosis for incident dementia has already been linked to CSF Aβ₄₂ patterns, there are no data on the potential value of plasma amyloid-β levels as prognosis biomarkers. Hence, we decided to determine whether or not plasma amyloid-β levels are associated with the prognosis in dementia, measured as cognitive decline and mortality in incident demented cases from the French Three-City (3C) prospective cohort study of men and women aged 65 and over.

MATERIALS AND METHODS

The 3C study

The 3C Study is a population-based, prospective study of the relationship between vascular factors and dementia. It has been carried out in three French cities: Bordeaux, Montpellier and Dijon. Between January 1999 and March 2001, 9,294 non-institutionalized subjects aged 65 or over agreed to participate in the study. At inclusion, blood samples were obtained from 8,682 individuals. The baseline data collection included socio-demographic and lifestyle characteristics, symptoms and complaints, main chronic conditions, medication use and neuropsychological testing. These data were updated at each follow-up examination, performed in 2001–2003, 2003–2005, 2006–2008. Information on vital status and date of death was available until 2009 (Figure 1). The study protocol has been described in detail elsewhere [4, 7] and was approved by the Institutional Review Board at Kremlin-Bicêtre University Medical Center. All participants gave their written, informed consent to participation.

Figure 1.

Participant disposition for the Three City (3C) Study.

Evaluation of dementia

At baseline and at each follow-up examination, participants were screened for dementia in a three-step procedure. First, specially trained psychologists administered a battery of neuropsychological tests. Second, patients who screened positive for dementia on the basis of neuropsychological tests were examined by a neurologist in Dijon. In Montpellier and Bordeaux, all participants were examined at baseline. During follow-up, participants with suspected incident dementia (on the basis of their neuropsychological test results) were all examined by a neurologist. Lastly, an independent committee of neurologists reviewed all potential prevalent and incident cases of dementia in order to obtain a consensus on the condition’s diagnosis and etiology, according to the criteria given in the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV). Dementia was classified according to the National Institute of Neurological and Communication Disorders and Stroke/Alzheimer’s Disease and Related Disorders Association criteria for AD and the National Institute of Neurologic Disorders and Stroke criteria for vascular dementia [8, 9]. Subjects with a typical history of AD (progressive worsening of memory or other cognitive functions) and documented stroke were classified as having mixed dementia. Our analysis included incident dementia cases at the 2-year examination. None of the participants was demented at the start of the study, when the plasma samples were taken.

Amyloid-β peptide assays

Non-fasting plasma samples were collected at baseline in tubes containing sodium EDTA as an anticoagulant. Following centrifugation, plasma samples were aliquoted into polypropylene tubes, stored at −80°C and only thawed immediately prior to amyloid-β quantification. The analyses were performed in a single centralized laboratory in Lille (SS, LB). Plasma amyloid-β peptide levels were measured blind to cognitive status. The plasma amyloid-β peptide assay was performed using the INNO-BIA kit (Innogenetics, Ghent, Belgium), based on a multiplex xMAP technique with a LABScan-100 system (Luminex BV, The Netherlands). Aβ₄₀ and Aβ₄₂ (kit format A) and Aβ_n-40 and Aβ_n-42 (kit format B) were respectively determined. Amyloid-β peptide levels from each blood draw were measured in duplicate. The same plasma sample was measured in each assay and we obtained analogous coefficients of variations (CVs): 7% for Aβ₄₀ and 6% for Aβ₄₂. At this level, it is important to note that the affinities of antibodies used in formats A and B (3D6 and 4G8, respectively) for Aβ are different and this makes difficult the comparison of the absolute levels of the full-length forms with those of the truncated forms.

Cognitive functions

At baseline and during follow-up, the following cognitive tests were administered by trained staff: the Mini Mental State Examination (MMSE; theoretical score: 0–30), the Isaacs Set Test (IST, for testing overall fluency; total score in 4 categories at 15 seconds) and the Benton Visual Retention Test (BVRT, testing recall of previously presented designs; theoretical score: 0–15). The MCI status of each participant was defined at baseline of the recruitment. In the whole 3C population (n=9,294), 43.05% were defined as MCI (n = 3831 and 396 subjects with available data). The characteristics of the MCI subjects have been described elsewhere by our team [10]. The MCI subjects were defined according to the revised MCI criteria (MCI-R criteria) [11]. These criteria are those proposed by the Stockholm consensus group [12]. They stipulate: (i) presence of a cognitive complaint from either the subject and/or a family member; (ii) absence of dementia; (iii) change from normal functioning; (iv) decline in any area of cognitive functioning; (v) preserved overall general functioning but possibly with increasing difficulty in the performance of activities of daily living.

Statistical methods

Amyloid-β variables (Aβ₄₀, Aβ₄₂, Aβ_n-40, Aβ_n-42, Aβ₄₂/Aβ₄₀ and Aβ_n-42/Aβ_n-40) were examined as quantitative parameters. The multiple regression analyses used to probe associations between amyloid-β variables and cognitive score, have tested various known potential confounders: age (in years), gender, educational level (no schooling or primary school vs. secondary or university degree), study center, Apo E (presence or absence of the APOE ε4 allele), marital status (living alone, No/Yes), body mass index (BMI; normal value BMI<25 kg/m2 vs. overweight BMI≥25 kg/m2), alcohol consumption (No/Yes), smoking status (never-smoker vs. former or current smoker), self-reported history of vascular disease, hypertension (systolic/diastolic blood pressure ≥ 160 mmHg/≥95 mmHg, respectively, or current antihypertensive medication), hypercholesterolemia (fasting blood cholesterol >6.2 mmol/L or use of lipid-lowering medication), diabetes (fasting blood glucose ≥ 7.0 mmol/L or use of anti-diabetes treatment), depression (Center for Epidemiologic Studies Depression Scale (CES-D) score over 16 or current treatment for depression) and limitations in at least one of the Instrumental Activities of Daily Living (IADL).

In order to consider the longest cognitive follow-up period for each subject (Figure 1), the definition of cognitive decline was based on the difference between the last score noted during follow-up and the baseline score, divided by the length of follow-up (mean: 3.77 years, SD=1.88). Subjects were classified as “fast decliners” when they presented a decline equal to or greater than the upper quartile of the population as a whole (i.e. ≥ 2.13 points per year of follow-up for the MMSE score; ≥ 1.48 points/year for the BVRT score and ≥ 2.43 points/year for the IST score). Logistic regression models were used to determine the association between amyloid-β concentrations and the risk of fast cognitive decline, while controlling for potential confounders and the baseline cognitive score.

Associations between amyloid-β levels and the risk of mortality were determined by Cox proportional hazards regression, with age (in years) used for the time axis and left truncation at age of study entry.

Multivariate adjusted models included covariates that were associated with cognitive tests (MMSE, BVRT or IST) based cognitive decline or mortality (p< 0.15 in the univariate model). All analyses were performed first on the whole dementia population and then on probable or possible AD cases. The analyses were performed using SAS software (version 9.2, SAS Institute Inc., Cary, NC).

RESULTS

Characteristics of the subjects

During the first 2–year follow-up period, incident dementia was diagnosed in 133 subjects with available baseline plasma amyloid-β concentration data (Figure 1). Four individuals had aberrant plasma amyloid-β concentration value and 9 had missing data about confounding factors in the multivariate analysis. After exclusion of these 13 participants, the 120 analyzed cases of dementia were classified as follows: 81 cases of AD (38 possible and 43 probable), 23 cases of mixed/vascular dementia, 7 cases of dementia with Parkinson’s disease and 9 with other types of dementia (Lewy body dementia: n=4; dementia associated with multiple sclerosis, alcoholism, etc.: n=4; undefined dementia: n=1). The mean follow-up was 6.07 years (SD=2.09). Baseline cognitive tests scores of MMSE, IST and BVRT were available for 120, 117 and 116 subjects respectively. The mean follow-up concerning cognitive scores was 3.77 years (SD=1.88). During follow-up, 63 subjects deceased. The incident dementia population’s main baseline characteristics are summarized in Table 1. Mean age at baseline was 78.4 (SD=5.7) and 68 of the patients were females (56.7%). A primary school educational level was reported by 42.5% of subjects. High school and University educational level were noted in 17.5% of cases each. A history of CVD was reported in 37.5%. More than 50% of the subjects presented hypercholesterolemia (51.3%) and hypertension (57.5%). Overweight was observed in 36.2% and diabetes in 13.3%. The mean MMSE score was 25.6 (SD=1.99) at the baseline visit and 22.7 (SD=2.9) two years later (at time of diagnosis). The mean IST score was 24.8 (SD=5.7) at baseline and fell to 22.5 (SD=6.2) at the time of diagnosis. The BVRT score was 9.8 (SD=2.72) at baseline. 72.6% of the subjects were classified at baseline as having MCI and 26.7% had an IADL limitation. The mean values for all the amyloid-β levels are presented in Table 1. Levels were not correlated with age nor with sex in this sample (data not shown). Additionally, due to the link between albumin and plasma proteins levels, we evaluated the association between concentration of albumin and the risk of cognitive decline (IST decline (n=118): p=0.48, MMSE decline (n=122): p=0.84, Benton decline (n=115): p=0.23) and also between albumin and plasma amyloidβ-β (ββ₄₀: p=0.8; Aβ₄₂: p=0.2; Aβ_n-40: p=0.37; Aβ_n-42: p=0.86). No association was underlined. The total plasma proteins concentration was not available in our study.

Table 1.

Description of the 120 incident cases of dementia.

		Dementia
		n	(%)	m (SD)
Gender	Female	68	56.7
Age at baseline	(years)	120		78.4 (5.7)
Education	Primary	51	42.5
	Middle school	27	22.5
	High school	21	17.5
	University	21	17.5
Smoking status	Current or Former	44	36.7
Body Mass Index	Normal	63	54.3
	Overweight	42	36.2
	Obese	11	9.5
Hypercholesterolemia		61	51.3
Hypertension		69	57.5
History of CVD		45	37.5
Diabetes		16	13.3
Alcohol (g/day)	<1	22	18.3
	1–36 g	83	69.2
	>36 g	15	12.5
Living status	Alone	47	39.2
Apo E	At least 1 Apoε4 allele	44	36.7
IADL	At least 1	32	26.7
Depression	CES-D score <16 or antidepressants	53	44.2
MCI status		85	72.6
MMSE* (n=120)		120		25.6 (1.99)
BVRT* (n=116)		116		9.8 (2.72)
IST* (n=117)		117		24.8 (5.7)
Plasma Aβ levels
Aβ40	pg/ml	120		246.0 (56.39)
Aβ42	pg/ml	120		39.8 (10.55)
Aβn-40	pg/ml	120		269.4 (62.33)
Aβn-42	pg/ml	120		27.7 (8.32)
Aβ42/Aβ40 ratio		120		0.2 (0.05)
Aβn-42/Aβn-40 ratio		120		0.1 (0.03)

^*Data are presented for the 120 demented subjects, except for the BVRT (n=116) and the IST (n=117).

SD: Standard Deviation; CVD: Cardiovascular Disease; Apo E: Apo lipoprotein E; IADL: Instrumental Activities of Daily Living; CES-D: Center for Epidemiologic Studies Depression Scale; MCI: Mild Cognitive Impairment; MMSE: Mini Mental State Examination; BVRT: Benton Visual Retention Test; IST: Isaacs Set Test.

The ApoEε4+/ε4+ status was found on 3 of the 120 patients (2.5% of the total sample).

Cognition

Cognitive impairment (Table 2)

Table 2.

Plasma amyloid-β levels (for higher level of 5 pg/ml) and cognitive scores at baseline in incident dementia cases and Alzheimer’s disease patients. A multiple linear regression model including gender, age, educational level, center, alcohol consumption, hypertension, IADL, depression and a history of vascular disease.

	All dementia		Alzheimer
	β	p	β	p
	IST score
Aβ40	−0.13	0.005	−0.13	0.04
Aβ42	−0.60	0,03	−0.45	0.18
Aβn-40	−0.09	0.03	−0.07	0.19
Aβn-42	−0.79	0.02	−1.03	0.01
Aβ42/Aβ40	1.68	0.18	1.64	0.30
Aβn-42/Aβn-40	−1.01	0.55	−3.18	0.16
	BVRT score
Aβ40	−0.01	0.53	−0.02	0.39
Aβ42	0.16	0.20	0.04	0.79
Aβn-40	−0.002	0.90	0.01	0.57
Aβn-42	0.001	0.99	−0.17	0.33
Aβ42/Aβ40	0.97	0.10	0.66	0.30
Aβn-42/Aβn-40	−0.11	0.89	−1.54	0.09
	MMSE score
Aβ40	0.02	0.32	0.03	0.27
Aβ42	0.06	0.49	0.17	0.15
Aβn-40	−0.001	0.96	0.00003	1.00
Aβn-42	−0.01	0.95	0.10	0.51
Aβ42/Aβ40	−0.15	0.73	0.11	0.84
Aβn-42/Aβn-40	−0.10	0.89	0.36	0.66
IST	All dementia (n=117)		AD cases (n=80)
BVRT	All dementia (n=114)		AD cases (n=77)
MMSE	All dementia (n=117)		AD cases (n=80)

In the whole demented population (“All dementia” in the Table 2), high levels of Aβ₄₀ (β=−0.13; p=0.005), Aβ₄₂ (β=−0.60; p=0.03), Aβ_n-40 (β= −0.09; p=0.03) and Aβ_n-42 (β= −0.79; p=0.02) were all associated with a lower IST verbal fluency score at baseline (i.e. 2 years before incident dementia diagnosis), according to a multiple linear regression model that included gender, age, educational level, center, alcohol consumption, IADL limitations, hypertension, depression and a history of vascular disease (Table 2). None of the associations between plasma levels and the baseline MMSE score or BVRT score reached statistical significance in the whole demented population.

For AD patients (“Alzheimer” in Table 2), the same multiple linear regression model revealed that high levels of Aβ₄₀ and Aβ_n-42 were also correlated with a low IST score. In AD patients, there was no obvious relationship between the IST score on the one hand and the Aβ₄₂ level or the two peptide ratios on the other. None of the associations between plasma levels and the baseline MMSE score or BVRT score reached statistical significance in AD cases (Table 2).

Faster cognitive decline (Table 3)

Table 3.

The risk of fast cognitive decline in all dementia cases and in AD cases: the crude and adjusted odds ratios (ORs) and 95% confidence intervals (CI) associated with each 5 pg/ml increase in the various plasma amyloid-β peptide levels. The adjusting factors were gender, age, educational level, center, alcohol consumption, hypertension, IADL, depression, cognitive score at baseline and a history of vascular disease.

	All	dementia		AD
	Adjusted OR	95%CI	p	Adjusted OR	95%CI	p
Isaac Set Test a
Aβ40	0.97	0.92–1.03	0.32	1.02	0.92–1.13	0.76
Aβ42	0.70	0.51–0.96	0.03	0.35	0.15–0.81	0.01
Aβn-40	0.96	0.92–1.01	0.11	0.96	0.90–1.03	0.23
Aβn-42	0.82	0.56–1.20	0.31	0.29	0.11–0.75	0.01
Aβ42/Aβ40	0.36	0.08–1.54	0.17	0.002	0.001–0.18	0.008
Aβn-42/Aβn-40	1.43	0.24–8.53	0.70	0.02	0.001–1.01	0.05
Benton Visual Retention Test b
Aβ40	0.95	0.90–1.01	0.09	0.94	0.85–1.05	0.27
Aβ42	0.72	0.54–0.97	0.03	0.76	0.47–1.21	0.25
Aβn-40	0.99	0.95–1.04	0.68	1.02	0.95–1.10	0.59
Aβn-42	0.84	0.58–1.21	0.35	0.95	0.53–1.71	0.86
Aβ42/Aβ40	0.48	0.12–1.98	0.31	0.57	0.07–4.42	0.59
Aβn-42/Aβn-40	0.46	0.06–3.56	0.46	0.33	0.01–11.49	0.54
MMSE c
Aβ40	1.01	0.97–1.06	0.51	1.03	0.97–1.10	0.36
Aβ42	1.06	0.83–1.34	0.66	1.27	0.89–1.81	0.18
Aβn-40	1.01	0.97–1.05	0.51	1.02	0.96–1.07	0.56
Aβn-42	1.04	0.77–1.40	0.81	1.23	0.80–1.88	0.34
Aβ42/Aβ40	1.11	0.35–3.53	0.86	0.96	0.19–4.96	0.96
Aβn-42/Aβn-40	0.64	0.13–3.13	0.58	0.95	0.10–9.15	0.97

^a111 all dementia cases, including 28 fast decliners; 76 AD cases, including 14 fast decliners.

^b108 all dementia cases, including 27 fast decliners; 73 AD cases, including 12 fast decliners.

^c111 all dementia cases, including 29 fast decliners; 75 AD cases, including 15 fast decliners.

Of the demented subjects (“All dementia”) with cognitive follow-up data, 29, 28 and 27 were defined as fast decliners according to the MMSE, the IST and the BVRT scores, respectively. In a multivariate analysis, a higher level of 5pg/ml of baseline Aβ₄₂ level was associated with a decreased risk of cognitive decline for both the IST (odds ratio (OR) =0.70; 95% confidence interval (CI) [0.51–0.96], p=0.03) and BVRT scores (OR=0.72; 95%CI [0.54–0.97], p=0.03) (Table 3). None of the associations between plasma levels and faster cognitive decline according to the MMSE score reached statistical significance in the “all dementia” group.

For the 76 AD cases (including 15, 14 and 12 fast decliners, according to the MMSE, IST and BVRT scores, respectively), the risk of cognitive decline according to the IST decreased significantly with a higher level of 5 pg/ml of plasma Aβ₄₂ at baseline (OR= 0.35; 95%CI 0.15–0.81], p=0.01), Aβ_n-42 (OR= 0.29; 95%CI [0.11–0.75], p=0.01), and with a higher level of the two ratios Aβ₄₂/ Aβ₄₀ (OR= 0.002; 95%CI [0.001–0.18], p=0.008) and Aβ_n-42/ Aβ_n-40 (OR= 0.02; 95%CI [0.001–1.01], p=0.05) (Table 3). Associations with BVRT were not confirmed in AD patients. None of the associations between plasma levels and cognitive decline according to the MMSE score reached statistical significance in AD cases.

Mortality

During the follow-up period (mean: 6.07 years, SD=2.09), 63 demented patients (including 35 AD cases) died. In neither univariate nor multivariate analyses was the risk of death associated with plasma amyloid-β levels in the dementia patients or in AD cases (Supplementary Table 1).

DISCUSSION

To the best of our knowledge, this is the first report to show that plasma levels of some forms of truncated or non-truncated amyloid-β peptides and their ratios are associated with faster cognitive decline in non-demented subjects converting to dementia during a 2-year follow-up. We also found an association between an elevated plasma amyloid-β level and poorer cognitive status two years before the dementia diagnosis.

Our findings raise issues concerning the role of blood-based biomarkers in AD and prompt reexamination of the ways in which measuring amyloid-β and its truncated forms might be helpful, especially in terms of prognosis. Most of the data stay controversial in this aspect and few studies have examined potential relationships between prognosis and plasma amyloid-β levels. Interestingly, we observed a decreased risk of faster cognitive decline with higher plasma levels of Aβ₄₂, Aβ_n-42, and with higher Aβ₄₂/Aβ₄₀ and Aβ_n-42/Aβ_n-40 ratios in AD cases, whereas in the whole group of demented patients only Aβ₄₂ was associated with a faster cognitive decline as measured by the IST and also by the BVRT score. It should be noted that the association between amyloid-β and fast cognitive decline was the strongest evidenced for verbal fluency, and not significant with the BVRT nor the MMSE, probably because the IST is one of the domains known to be affected early in the course of AD as illustrated in the PAQUID study [13]. Moreover, the MMSE as a global neuropsychological test is not relevant for early AD. The first analysis of the Northern Manhattan community cohort determined that high plasma Aβ₄₂ levels were associated with faster cognitive decline-primarily in memory-in non-demented elders [14]. In line with our results Schupf et al. (2008) showed that declines in Aβ₄₂ levels or in the Aβ₄₂/Aβ₄₀ ratio are sensitive indicators of recent conversion to AD [5]. A similar association between the Aβ₄₂/Aβ₄₀ ratio and an imminent risk of cognitive decline was reported in the Mayo Rochester AD patient registry [15]. Our results are also in agreement with a very recent report from Yaffe et al., (2011) in which low plasma Aβ₄₂/Aβ₄₀ ratio in dementia-free, community-dwelling older adults appears to predict more rapid cognitive decline than in subjects with a high ratio [16].

The advantage of our study is that our design enabled us to examine plasma Aβ₄₂ and Aβ₄₀ levels and their truncated forms in the 2-year interval preceding diagnosis of dementia, which can be considered as corresponding (at least in part) to prodromal AD. With reference to Jack’s dynamic model, amyloid-β peptides may be pathological early in the course of the disease [17]. We found an association between an elevated plasma amyloid-β level and poorer cognitive status two years before the dementia diagnosis. Poorer verbal fluency was correlated with higher plasma levels of all forms of amyloid-β in incident dementia cases and only with Aβ₄₀ and truncated Aβ_n-42 levels in AD patients. These results were independent of various demographic, lifestyle and vascular factors. Cross-sectional [18–20] and longitudinal studies of plasma Aβ₄₂ levels in normal or MCI patients [1, 6] have generated conflicting results [21]. In the Northern Manhattan community cohort with repeated assays, authors indicated that plasma levels of Aβ₄₂ began to decrease soon after diagnosis of AD [22]. Although a higher risk of progression towards AD was observed in MCI patients with low levels of plasma Aβ₄₂ and a low Aβ₄₂/Aβ₄₀ ratio [3, 6, 18], the mean plasma Aβ₄₂ and Aβ₄₀ levels and amyloid-β peptide ratios have not been found to have predictive value in this context. Some articles have reported slightly higher Aβ₄₂ or Aβ₄₀ plasma levels in patients with AD than in healthy age-matched controls and that high level of plasma Aβ₄₂ and a large Aβ₄₂ /Aβ₄₀ ratio are risk factors for future AD [3, 4, 22–24].

There are several possible disparities between literature findings and our present results; they include different sample collecting, processing and plasma amyloid-β assay methods, lower amyloid-β levels in plasma than in CSF, circadian fluctuations in amyloid-β levels [25], and peripheral confounding factors (platelets, the binding of Aβ with albumin, cholesterol content and other proteins) [26–29]. First, due to the population-based study design, we do not have CSF or functional neuroimaging biomarkers such as Pittsburgh Compound B (PIB) or florbetapir positron emission tomography brain neuro-imaging to compare to plasma amyloid-β results. The prognosis in dementia has been linked to patterns of CSF biomarkers [30–33]. Although functional neuroimaging data in a pre-prodromal stage of the disease are still a matter of debate, the correlation between plasma Aβ isoforms and fibrillar brain amyloid-β load as measured by PIB retention is underlined [34]. Aβ₄₂ as well as the Aβ₄₂/Aβ₄₀ ratio in plasma were significantly lower in AD patients and inversely correlated with in vivo measures of fibrillar brain Aβ load as measured by PIB [35]. Concerning confounding factors, except for albumin which does not show any correlation in our study, the role of platelets cannot be ignored. Platelets represent a source of circulating Aβ and AβPP [27] and mostly contain Aβ ending at residue 40, with a small amount of Aβ₄₂, in both the quiescent and activated states [29]. It is known that, in the brain, Aβ peptides exist both within defined deposits (plaques and vascular) and as a diverse array of other forms including soluble, membrane associated and intracellular species that may play far more significant roles in the production of dementia than the molecules sequestered in extracellular plaques. Aβ peptides are generated outside of the central nervous system in appreciable quantities by the skeletal muscle, platelets and vascular walls [26, 36, 37] and in other non-neural tissues [38, 39]. These distinct reservoirs may allow for an active and dynamic interchange of Aβ peptides between the brain and periphery. Even if the efforts to validate plasma Aβ peptides as dementia biomarkers have been fraught with frustration, the critical role that circulating Aβ peptides play in AD pathology cannot be ignored. Moreover, our method is solid and we think that peripheral Aβ could be sufficient to define a model for disease progression as described also in previous studies [40] and as a marker of incipient AD. Further investigations are needed about brain Aβ turnover and metabolism and exact relationship between plasma and brain Aβ.

The analysis of the Aβ truncated forms has been insufficiently studied by other groups despite the important literature on the matter. The potential relevance of truncated amyloid-β peptides in AD has been underlined by Buée et al., who showed in early stages of AD, through post-mortem analysis of brain insoluble fractions, that these species represent more than 60% of all amyloid-β species in Alzheimer’s pathology [41]. These truncated species have been shown to induce learning impairments and neuronal apoptosis when injected in murine models [42]. Moreover, they are detectable in CSF and it has been suggested that their levels are relevant in the diagnosis of AD [43]. However, little is known about their relationship with cognitive decline; especially fast cognitive decline, giving to our study a major impact. Interestingly, distinctive non-linear relationships with cognitive scores and plasma biomarkers have been underlined in our study as previously reported with CSF biomarkers [30–32, 44–49]. The weight of CSF biomarkers is not the same in all stages of AD, probably because of the kinetics of CSF biomarkers during the course of cognitive impairments. CSF Aβ₄₀ levels showed a biphasic curve, whereas CSF Aβ₄₂ levels showed a monotonic but non-linear dependence on cognitive scores, with major variations up to the change point and then mild or no variations [44, 50].

We did not find any associations with mortality - another indicator of prognosis or more aggressive evolution of the disease but our sample size limits the power of this analysis. An association between high levels of Aβ₄₂ and mortality has been reported in AD patients [22] and Down’s syndrome patients [51].

Although our results concern a small number of incident cases and a single measurement of amyloid-β levels, their relevance is reinforced by the large size of the 3C study database. We controlled for factors known to be associated with the AD prognosis and/or plasma amyloid-β levels, such as age [18–20, 22] and cardiovascular factors [52]. Even though our cognitive evaluation was limited in comparison with the extensive battery used in clinical research, the IST score is known to be impacted early in the course of the disease. It is also easy to apply in routine clinical daily (especially by primary care physicians) and could be used to develop a preventive approach to dementia.

When assayed at a prodromal stage, plasma amyloid-β levels may be potentially useful markers of fast cognitive decline in individuals who subsequently become demented. Current results also lend to the potential utility of plasma amyloid-β as an indicator of disease progression. Our results need to be replicated in larger populations and with several assays over time per subject before any clinical recommendation can be made. Nevertheless, we believe that our present data raise intriguing questions about peripheral biomarkers of brain lesions in AD and how they relate to current and future cognitive status.