Research Article: Conflicting cerebrospinal fluid biomarkers and progression to dementia due to Alzheimer’s disease

Date Published: December 9, 2016

Publisher: BioMed Central

Author(s): Panagiotis Alexopoulos, Lukas Werle, Jennifer Roesler, Nathalie Thierjung, Lena Sophie Gleixner, Igor Yakushev, Nikolaos Laskaris, Stefan Wagenpfeil, Philippos Gourzis, Alexander Kurz, Robert Perneczky.


According to new diagnostic guidelines for Alzheimer’s disease (AD), biomarkers enable estimation of the individual likelihood of underlying AD pathophysiology and the associated risk of progression to AD dementia for patients with mild cognitive impairment (MCI). Nonetheless, how conflicting biomarker constellations affect the progression risk is still elusive. The present study explored the impact of different cerebrospinal fluid (CSF) biomarker constellations on the progression risk of MCI patients.

A multicentre cohort of 469 patients with MCI and available CSF biomarker results and clinical follow-up data was considered. Biomarker values were categorized as positive for AD, negative or borderline. Progression risk differences between patients with different constellations of total Tau (t-Tau), phosphorylated Tau at threonine 181 (p-Tau) and amyloid-beta 1–42 (Aβ42) were studied. Group comparison analyses and Cox regression models were employed.

Patients with all biomarkers positive for AD (N = 145) had the highest hazard for progression to dementia due to AD, whilst patients with no positive biomarkers (N = 111) had the lowest. The risk of patients with only abnormal p-Tau and/or t-Tau (N = 49) or with positive Aβ42 in combination with positive t-Tau or p-Tau (N = 119) is significantly lower than that of patients with all biomarkers positive.

The risk of progression to dementia due to AD differs between patients with different CSF biomarker constellations.

The online version of this article (doi:10.1186/s13195-016-0220-z) contains supplementary material, which is available to authorized users.

Partial Text

An increasing body of evidence suggests that Alzheimer’s disease (AD) pathophysiology can be identified using biomarkers [1, 2]. AD is characterized by abnormal patterns in structural and functional imaging as well as by a pathological cerebrospinal fluid (CSF) signature [3]. The pathological CSF signature is defined by decreased CSF concentrations of the peptide amyloid-beta 1–42 (Aβ42) and increased levels of the proteins total Tau (t-Tau) and Tau phosphorylated at threonine 181 (p-Tau). It is of note that biomarkers reflect AD neuropathological changes with relatively high accuracy [4, 5]. In clinical practice, CSF biomarkers aid clinicians with decision-making, embody a key tool in the differential diagnosis especially of atypical dementia syndromes and increase diagnostic confidence [4, 6–9].

A total of 469 MCI patients out of 1729 ADNI participants with available baseline data fulfilled the inclusionary criteria of the study. APOE ε4 and sex distribution, as well as age and MMSE scores, significantly differed between the subgroups (Table 1). In particular, the MMSE scores of the MCINon+ subgroup were significantly higher compared with the scores of the MCIAll+ (p < 0.001) and MCIAβ+T+ (p < 0.01) subgroups. MMSE scores in the MCIAll+ subgroup were significantly lower in comparison with those of the MCIAβ+ (p = 0.02), MCIAβ+T+ (p = 0.03) and MCIT+ (p < 0.01) subgroups. Across the five studied MCI subgroups, approximately 45% of patients had conflicting CSF biomarker constellations. Figure 1, a graphical presentation of participants’ Aβ42, t-Tau and p-Tau CSF levels using NNMF, points to the high variability of the CSF biomarker findings in patients with MCI. Data were available from clinical follow-up visits conducted every 6 months up to 8 years after baseline. In total, 159 patients with MCI progressed to dementia due to AD. No patient progressed to any other form of dementia. The difference between the MCI subgroups in the proportions of patients who developed dementia due to AD within the follow-up period attained statistical significance (p < 0.001), whilst the duration of the follow-up period did not differ.Table 1Characteristics of the study sampleMCI subgroupp valueMCINon+MCIAβ+MCIAβ+T+MCIAll+MCIT+N1114511914549Age (years)71.29 (7.83)74.78 (6.59)74.25 (7.06)72.68 (7.40)71.57 (9.00)0.010Education (years)16.54 (2.70)16.29 (3.07)16.13 (2.77)15.97 (2.83)15.92 (2.86)0.468MMSE28.13 (1.70)27.64 (1.79)27.42 (1.88)26.93 (1.87)27.92 (1.78)<0.001Sex (male:female)67:4434:1178:4173:7229:200.020APOE ε4 carriers (%)21.6242.2262.1877.9326.53<.001CSF Aβ42 (pg/ml)232.59 (30.25)140.81 (26.14)132.81 (23.70)134.98 (20.96)232.93 (30.24)<0.001CSF Aβ42 negative/borderline/positive for AD87/24/00/0/450/0/1190/0/14539/10/0<0.001CSF p-Tau (pg/ml)18.63 (4.44)20.36 (4.45)43.66 (16.00)58.41 (2.53)41.77 (13.58)<0.001CSF p-Tau negative/borderline/positive for AD57/54/018/27/00/2/1170/0/1450/0/49<0.001CSF t-Tau (pg/ml)50.66 (18.12)55.53 (17.28)78.37 (18.06)158.09 (46.47)74.86 (34.40)<0.001CSF t-Tau negative/borderline/positive for AD106/5/042/3/059/58/20/0/14529/15/5<0.001Follow-up period (months)32.22 (23.34)32.53 (23.50)30.81 (22.34)29.96 (21.01)32.02 (11.64)0.350Dementia due to AD vs no dementia at follow-up14:9714:3143:7680:658:41<0.001Data presented as mean (standard deviation) or frequenciesAD Alzheimer’s disease, MCI mild cognitive impairment, APOE apolipoprotein E, MMSE Mini-Mental State Examination, CSF cerebrospinal fluid, Aβ42 amyloid-beta 1–42, p-Tau tau phosphorylated at threonine 181, t-Tau total tau, MCINon+ MCI without positive CSF biomarkers, MCIAβ+ MCI with positive Aβ42 and negative or borderline p-Tau and t-Tau, MCIAβ+T+ MCI with positive Aβ42 and positive t-Tau or p-Tau, MCIAll+ MCI with Aβ42 and both t-Tau and p-Tau positive, MCIT+ MCI with negative or borderline Aβ42 and at least p-Tau or t-Tau positiveFig. 1Condensed representation, as a 2D scatter plot, of the CSF biomarker values of the study participants. The ensemble of trivariate measurements of CSF Aβ42, p-Tau and t-Tau for all participants was analysed via NNMF and approximated by means of a bivariate data swarm that conveniently represents the total variation in the original data. Labels indicate the different groups and lend semantics to the plot. MCINon+ MCI without positive CSF biomarkers, MCIAβ+ MCI with positive Aβ42 and negative or borderline p-Tau and t-Tau, MCIAβ+T+ MCI with positive Aβ42 and positive t-Tau or p-Tau, MCIAll+ MCI with Aβ42 and both t-Tau and p-Tau positive, MCIT+ MCI with negative or borderline Aβ42 and at least p-Tau or t-Tau positive In line with a number of previous reports [17–19], but in contrast to others [8, 9, 34, 35], approximately half of the MCI cases in our study had conflicting CSF biomarker constellations. This discrepancy in the frequency of patients with conflicting CSF biomarker results could be possibly attributed to differences in study design. For instance, not all studies considered all three CSF AD biomarkers. In addition, past studies implemented a dichotomization strategy in interpreting biomarker findings, whilst in the present study biomarker values were categorized as positive, negative or borderline in line with the NIA-AA guidelines. Moreover, it is possible that academic, research centres recruit more patients with complex constellations of biomarker findings, whilst more patients with AD-typical CSF profiles and consequently more advanced neuropathology are recruited in non-academic, clinical settings. Interestingly, it has been shown that patients of a non-academic memory clinic suffered from more severe clinical symptoms in comparison with the patients of an academic memory clinic [36]. The present study provides a further piece of evidence for the prognostic differences between MCI subgroups with distinct neurochemical biomarker constellations. The study reveals significant differences between subgroups with conflicting biomarkers, on the one hand, and patients with all neurochemical biomarkers positive or non-positive (borderline or negative) for AD on the other. Even though our observations exclusively refer to neurochemical biomarkers and do not consider imaging markers, they point to the necessity of modifying/refining the NIA-AA algorithms for categorizing MCI.   Source: