Printer Friendly

Unbiased Approach to Counteract Upward Drift in Cerebrospinal Fluid Amyloid-[beta] 1-42 Analysis Results.

Amyloid plaque deposition in the brain is a pathological hallmark in Alzheimer disease (AD) [6]. Amyloid-[beta] 1-42 (A[beta] 1-42) concentrations in cerebrospinal fluid (CSF) can be used as a biomarker for the presence of abnormal amyloid aggregation in vivo. Reduced CSF A[beta] 1-42 concentrations have been robustly associated with the clinical dementia syndrome (1, 2), are related with postmortem-determined amyloid plaque burden (3, 4), and are associated in nondemented persons with a substantially increased risk to develop dementia in the future (5, 6). Consequently, CSF markers for amyloid abnormality are now part of established and widely used research criteria for AD across the cognitive spectrum (7-9). Clinical use of CSF A[beta] 1-42 concentrations, however, seems hampered by preanalytical and analytical variability in CSF, which complicates the development of a universal cutpoint to determine amyloid abnormality (10, 11). A further problematic observation is that A[beta] 1-42 values as measured with Innotest ELISA, a broadly used assay, seem to have increased gradually over the past 2 decades (12-14), which is also reflected by increasing cutpoints over the years reported in several cohorts (12, 15-17). One approach to remove drift effects is to reanalyze CSF samples in a single ELISA batch (13, 14), suggesting that the observed drift is not related to changes in patient characteristic over time but rather to changes in ELISA kits. However, this approach is infeasible with growing cohorts, as well as costly, thus impeding high quality research on CSF markers over and across cohorts.

We propose an alternative approach to statistically correct for drifts in A[beta] 1-42 values over time by taking as a starting point the observation that amyloid measurements show a bimodal distribution, which has been robustly observed in CSF and positron emission tomography (PET) data (12, 18-20). Bimodal distributions reflect the presence of a normal and a pathological population, and a cutpoint for amyloid abnormality can be determined based on the intersection of these 2 distributions (12, 18). We hypothesized that bimodal distributions can be identified independently of absolute values of amyloid, and that drift effects can be removed in an unbiased way by aligning cutpoints across time.

Our aim was to develop a statistical approach to correct for upward drift in A[beta] 1-42 concentrations over time by determining year-specific cutpoints based on Gaussian mixture modeling and aligning these over time. We validated our approach in 3 ways: (a) by comparing drift-corrected values with remeasured A[beta] 1-42 concentrations that were available for a subset of individuals with AD-type dementia, (b) by comparing drift-corrected values with visual reads of PET for amyloid abnormality that were available for a subset of individuals, and (c) by evaluating whether our approach improved detection of clinical AD-type dementia compared with uncorrected data.

Material and Methods


Study participants who visited our memory clinic between 2000 and 2016 were selected from the Amsterdam Dementia Cohort when they had a CSF sample available. Most participants received standardized dementia screening during their first visit as described before (21). Diagnosis for each participant was made in a multidisciplinary meeting according to established criteria as described previously (21), and we labeled individuals as having AD-type dementia, mild cognitive impairment (MCI), or subjective cognitive decline (SCD) accordingly. Other participants were labeled as having a non-AD type of dementia when they were diagnosed with frontotemporal dementia (n = 305), dementia with Lewy bodies (n = 146), or vascular dementia (n = 76). Individuals who received a diagnosis unrelated to neurodegenerative disease (n = 594) or when their diagnosis could not be established at the first visit (n = 335) were labeled as "other." The VU University Medical Center (VUmc) ethical review board approved the study. All participants provided written informed consent to use their data for research purposes.


CSF samples were obtained by lumber puncture using a 25-gauge needle and syringe between the L3/L4, L4/L5, or L5/S1 intervertebral space, collected in polypropylene tubes, and processed as previously described (21). A[beta] 1-42 concentrations were determined with sandwich ELISAs [Innotest, fi-AMYLOID (1-42), Fujirebio (formerly Innogenetics)] at the neurochemistry laboratory of the Department of Clinical Chemistry of VUmc.


A subset of individuals had visual amyloid PET scans available within a year of CSF sampling [obtained, on average, within 1.7 (2.4) months]. All PET scans were labeled as amyloid positive or negative based on visual reading by an experienced nuclear medicine physician (BvB). Amyloid PET imaging was conducted using [sup.11]C-PiB (n = 177), [sup.18]F-flutemetamol (n = 112), [sup.18]F-florbetaben (n = 300), and [sup.18]F-florbetapir (n = 13). PET methodology has been previously described in detail (22-24), except for [sup.18]F-florbetapir. [sup.18]F-florbetapir images were made on a Gemini TF-64 PET/CT scanner (Philips Medical Systems) or a Philips Ingenuity TF PET/CT scanner and consisted of 2 X 10-min frames acquired 50 to 70 min after bolus net injection of approximately 370 MBq [sup.18]F-florbetapir.


Changes in cohort composition over time were assessed with linear regression for continuous variables age and Mini-Mental State Examination (MMSE) score, with Pearson [chi square] trend tests for nominal variables diagnostic group, sex, and apolipoprotein E (APOE) [7] [epsilon]4 genotype, and time as an ordered categorical variable. Changes in CSF A[beta] 1-42 concentrations over time were first determined for the total cohort with linear regression. If cohort composition (i.e., diagnostic group, age, sex, or APOE [epsilon]4 genotype) showed time dependencies, we performed additional analyses to test whether drift effects could be explained by such variables. We calculated year-specific cutpoints to determine amyloid status with Gaussian mixture modeling (12, 18). First, we determined the number of distributions providing the best fit on the data with the R function "boot.comp." Second, for each year we determined a cutpoint as the value where 2 fitted normal distributions intersected. Only a limited number of CSF samples were available for the period 2000 to 2002, and so we grouped these together to ensure reliable cutpoint estimation for these early years. We removed drift effects by aligning year-specific cutpoints, an unbiased method that keeps intact all other information present in the data. Per year, a scaling factor was determined based on the difference between the reference cutpoint obtained from the most recent period when cutpoints remained stable (2015-2016), the upper limit of abnormal values, and the lower limit of normal values. Scaling factors were added to the raw normal and abnormal A[beta] 1-42 values to align year-specific cutpoints to the reference. For a subset of AD-type dementia individuals, A[beta] 1-42 estimates were available that were remeasured using the same batch of ELISA kits, which removed drift effects (13). We used these new measurements to validate whether our approach accurately removed time dependencies from CSF data by studying whether the difference between corrected and remeasured values was related to the time that CSF was obtained. Furthermore, we compared drift-corrected values with uncorrected data in terms of concordance with visual amyloid PET read for the total cohort and per year. Concordance was defined as the percentage of individuals who were classified identically using PET and CSF methods. Finally, we compared uncorrected and corrected data in terms of sensitivity-specificity for clinical AD-type dementia vs all other individuals for the total cohort and per year. For uncorrected data, we used cutpoints we previously determined over the years using different methods: 550 pg/mL (based on clinical discrimination in 2010) (15), 640 pg/mL (based on optimization of concordance with amyloid PET in 2014) (16), and 680 pg/mL (based on Gaussian mixture modeling in 2017) (12). Concordance, sensitivity, and specificity estimates were considered statistically different from each other when an estimate was not contained in the 95% CI of another estimate.



Over the past 2 decades, CSF A[beta] 1-42 measurements were available for 4397 participants (Table 1 and Fig. 1A). The majority of these individuals had AD-type dementia (31%), followed by SCD (18%), non-AD dementia (16%), and MCI (14%), with an additional 21% labeled as "other."

Cohort composition showed changes over time in the relative proportions of diagnostic groups [[chi square] (4) = 12.82; P = 0.01; Fig. 1B], with the proportion of individuals with SCD increasing over time (Z = 3.23; P = 0.001), and the group "other" showing a trend to decrease (Z = -1.69; P = 0.09). Proportions of AD-type dementia, non-AD dementia, and MCI remained stable over time (all P > 0.05). Furthermore, the age of the total cohort decreased over time [fi (SE) = -0.12 (0.04); P < 0.001; Fig. 1C]. MMSE scores, sex distributions, and the proportion of APOEe4 carriers remained stable over the years (all P > 0.05; Fig. 1, D-F).


The total cohort showed increases in A[beta] 1-42 values of 22.07 pg/mL per year (SE = 1.13; P < 0.001). Repeating analyses additionally correcting for age did not change these results [[beta] (SE) = 21.07 (1.09); P < 0.001], suggesting that higher A[beta] 1-42 values in more recent years were not explained by an increasingly younger age of the cohort. The slope of the drift showed a dependency on diagnosis ([P.sub.interaction] < 0.001), with A[beta] 1-42 values increasing 50% less strongly for individuals with AD-type dementia, 13 pg/mL per year, compared with increases of 25 pg/mL per year observed in individuals with SCD, 27 pg/mL in those with MCI, 23 pg/mL in participants with non-AD dementia, and 25 pg/mL in individuals with an "other" diagnosis (Fig. 2A). The specificity of slope differences for AD-type dementia suggests that the rate of annual increases depended on amyloid concentrations, and this explanation was supported by exploratory analyses on data stratified based on the median, which showed a similar difference in slope [low concentrations: [beta] (SE) = 9.30 (0.57), P < 0.001; high concentrations: [beta] (SE) = 14.13 (1.14), P < 0.001].


Bimodal distributions provided the most optimal fit for each year of CSF A[beta] 1-42 concentrations (all P < 0.05). Year-specific cutpoints increased from 527 pg/mL for samples analyzed in 2002 to 807 pg/mL for samples analyzed in 2016, corresponding to annual increases of 19 pg/mL (P < 0.001; Fig. 2B). Using these year-specific cutpoints, we labeled 2156 individuals as having abnormal amyloid [1226 (57%) with AD-type dementia, 343 (16%) with MCI, 248 (11%) with non-AD dementia, 215 (10%) "other," and 124 (6%) with SCD]. Using these year-specific cutpoints as a reference, we compared the proportion of mislabeling of amyloid status when uncorrected data with a single cutpoint are used. At most, 680(15%) individuals were mislabeled when using a cutpoint of 550 pg/mL, and at least 282 (6%) individuals were mislabeled when using a cutpoint of 680 pg/mL compared with year-specific cutpoints (Table 2 and Fig. 2C).


Our approach of adding year-specific constant values to the data such that cutpoints based on bimodal distributions of amyloid concentrations aligned over time removed the drift effect [[beta] (SE) = 0.07 (0.59); P = 0.91; Fig. 2, D and E]. Gaussian mixture modeling on drift-corrected data of the total cohort resulted in a cutpoint of 813 (95% CI, 761-859).


We compared drift-corrected A[beta] 1-42 concentrations with remeasured CSF A[beta] 1-42 concentrations that were available for a subset of 148 individuals with AD-type dementia (initially measured between 2000 and 2014 and then remeasured in 2015) (13). As previously reported, the absolute differences between uncorrected and remeasured A[beta] 1-42 values increased over time [[beta] (SE) = 16.25 (2.27); P < 0.001]. In contrast, absolute differences between drift-corrected and remeasured A[beta] 1-42 concentrations were unrelated with time [[beta] (SE) = 0.33 (2.11); P = 0.88], which supports our approach successfully removing the drift.


A subset of 602 (14%) participants had a visual reading of amyloid PET available. Compared with the total cohort, this subset had 10% more individuals with AD-type dementia and 10% fewer individuals classified as "other" (see Table 1 in the Data Supplement that accompanies the online version of this article at http://www. Across this subset, drift-corrected CSF A[beta] 1-42 concentrations showed the highest concordance of 88% (95% CI, 85-90) with amyloid PET readings, which was higher than concordance estimates of uncorrected CSF A[beta] 1-42 values with amyloid PET readings (Table 2). When comparing concordance of amyloid PET readings with uncorrected and drift-corrected data per year, drift-corrected data showed consistently high concordance with amyloid PET compared with uncorrected data for the period between 2006 and 2016 (Fig. 3).


Across the total group, sensitivity and specificity estimates for classifying individuals as having AD-type dementia vs all other participants were largely comparable between corrected A[beta] 1-42 values with a cutpoint of 813 pg/mL and uncorrected A[beta] 1-42 values with a cutpoint of 680 pg/mL (Table 2). Compared with uncorrected data with other cutpoints (550 and 640 pg/mL), corrected A[beta] 1-42 values showed better sensitivity to detect clinical AD dementia at the cost of lower specificity. When comparing sensitivity and specificity estimates between uncorrected and drift-corrected data per year, the largest differences in sensitivity and specificity were observed in the most recent years when using a cutpoint of 550 pg/mL (Fig. 3). Drift-corrected values also showed improved sensitivity to detect AD-type dementia for the period 2014 to 2016 compared with uncorrected data with 640 and 680 pg/mL as cutpoints. In this period, specificity for drift-corrected values was worse compared with uncorrected data for all cutpoints. We further explored whether lower specificity observed in this period for drift-corrected values reflected an improved detection of predementia AD. To this end, we compared the percentage of MCI and SCD participants labeled as having abnormal amyloid determined based on corrected and uncorrected CSF A[beta] 1-42 concentrations with previously reported prevalence estimates (25). Using uncorrected CSF A[beta] 1-42 concentrations resulted in a lower than expected percentage of individuals with abnormal amyloid, regardless of the cutpoint used (Table 3), whereas amyloid status determined with corrected CSF values resulted in percentages of MCI and SCD participants with abnormal amyloid that were in line with expected prevalences.


Over a period of almost 2 decades, we observed a drift in CSF A[beta] 1-42 concentrations, with an average annual increase of roughly 20 pg/mL across the total cohort. Cutpoints to determine amyloid abnormality increased with a similar rate from 527 pg/mL in 2002 up to 807 pg/mL in 2016. During this period, the composition of the cohort showed some changes in the relative proportions of diagnosis and younger age over time. These changes in cohort composition, however, did not explain the increase in A[beta] 1-42 concentrations, suggesting that the drift is mainly caused by variability in ELISA kits over the years. Drift effects were less strong for individuals with AD-type dementia compared with other clinical diagnoses, indicating that the increase in A[beta] 1-42 values depended on A[beta] 1-42 concentrations. We demonstrated and validated that drift effects can be removed from the data by a simple scaling factor derived from year-dependent cutpoints for amyloid abnormality. This resulted in a uniform cutpoint value and avoided potential removal of patient information that might be relevant for disease and/or related to amyloid levels.

ELISAs detect soluble A[beta] 1-42, both in monomers and as complex with other proteins or oligomers. Many studies have demonstrated that CSF A[beta] 1-42 concentrations are reduced in individuals with AD, and this has resulted in the incorporation of CSF A[beta] 1-42 in research criteria for AD (7-9). Yet, precise causes for reduced A[beta] 1-42 concentrations continue to be uncertain. Postmortem studies comparing amyloid plaque load in the brain support the concept that reduced concentrations in A[beta] 1-42 reflect aggregation into plaques (3). However, it is unclear what explains the drift over time in A[beta] 1-42 concentrations as measured with ELISA, and why the magnitude of the drift depends on A[beta] 1-42 concentrations. If A[beta] 1-42 concentrations reflect only a sink of soluble A[beta] 1-42 into plaques, this leaves the question unexplained as to why drift rates would depend on whether amyloid levels are within normal or pathological ranges. The difference in drift rates we observed suggests that CSF samples contain additional differences in their composition that are associated with the presence of normal high or abnormal low A[beta] 1-42 concentrations. Previous studies have demonstrated that CSF of AD-type dementia patients contains large A[beta] 1-42 aggregates (26) and soluble oligomers (27-29) (although the latter are more difficult to detect) (30-33). A recent study showed that monomeric A[beta] 1-42 concentrations can be recovered when CSF samples containing A[beta] 1-42 aggregates are treated with a chemical to dissolve the aggregates (34). Possibly, during the past 17 years, changes in ELISA kits and/or calibration data have occurred that might be influenced by the presence of A[beta] 1-42 aggregates. Although the precise causes of increasing A[beta] 1-42 concentrations over time remain to be determined, our results implicate that when correcting this drift from historic data, A[beta] 1-42 concentrations need to be considered because these may influence the strength of the drift.

Compared with year-specific cutpoints for amyloid abnormality, the use of a single threshold on uncorrected data had minimal impact on the total cohort, with misclassification of amyloid abnormality for 6% to 15% of the individuals, depending on the cutpoint used. Impact in terms of classification for the clinical diagnosis of AD-type dementia was minimal for uncorrected data when a cutpoint of 680 pg/mL was used, showing similar sensitivity and specificity estimates in the total group of 4397 patients compared with corrected data with a cutpoint of 813 pg/mL. For the most recent period between 2014 and 2016, uncorrected A[beta] 1-42 concentrations for all cutpoints showed better specificity; however, this was associated with an underestimation of amyloid abnormality in those with MCI and SCD, whereas estimates based on corrected values were in line with those reported in the literature (25). Our results imply that when studying historic data for individuals from populations that less frequently show pathological amyloid, it is important to have a cutpoint that considers the drift to avoid selecting false-positive cases (i.e., participants selected from an earlier period) and false-negative cases (i.e., participants selected from more recent periods). Here we made use of the observation that the bimodal distributions of A[beta] 1-42 concentrations are independent of absolute values of A[beta] 1-42, and so by realigning these distributions, we could derive in an unbiased way a single cutpoint that can be used for all data.

A potential limitation of the present study is that we do not have pathological information available for all individuals, and so the possibility that our drift correction might still lead to misclassification of amyloid abnormality cannot be excluded. However, the validation analyses we performed support the idea that our approach effectively removed time effects when comparing with remeasured A[beta] 1-42 values. Moreover, the latter values demonstrated improved concordance with amyloid PET compared with uncorrected data. A strong aspect of our approach is that we determined cutpoints for amyloid abnormality using a Gaussian mixture modeling approach that does not rely on clinical outcome and so is unbiased (12, 18). With this approach, we avoided introducing dependencies in CSF data on clinical diagnosis. Another advantage of our approach is that it does not require regression modeling, which might remove too much information from amyloid concentrations that may potentially be related to time-dependent changes of (possibly yet to be discovered) factors in the cohort. An alternative, more costly approach would be to remeasure A[beta] 1-42 for each novel batch. Systematic differences also exist across different assays to assess A[beta] 1-42 in CSF; for this reason, ongoing international initiatives are developing novel measurement methods to reduce such variability with the aim to determine a universal cutpoint. Our approach provides the opportunity to standardize A8 1-42 concentrations across platforms and cohorts, which may enable identifying a universal cutpoint until better measurement methods exist. Once improved measurement methods have arrived, our approach will enable preserving the use of large, valuable historic data sets by calibrating these to new data.

In conclusion, we observed increasing A[beta] 1-42 CSF concentrations over time owing to changes in ELISA kits. These increases were doubled in individuals with high A[beta] 1-42 CSF concentrations, suggesting that drift over time depends on whether amyloid metabolism is normal or abnormal. Drift effects can be adjusted for by realigning year-specific cutpoints for amyloid abnormality, resulting in an unbiased and uniform cutpoint for the entire period. Potentially, this approach might also be useful to align CSF data and thereby reduce variability between multiple centers.

Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.

Authors' Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest:

Employment or Leadership: None declared.

Consultant or Advisory Role: C.E. Teunissen, Fujirebio.

Stock Ownership: None declared.

Honoraria: None declared.

Research Funding: The VUmc Alzheimer Center is part of the neurodegeneration research program of Neuroscience Amsterdam and is supported by Stichting Alzheimer Nederland and Stichting VUmc Fonds. The clinical database structure was developed with funding from Stichting Diorapthe. B.M. Tijms, the Memorabel grant program of ZonMW (number 733050506).

Expert Testimony: None declared.

Patents: None declared.

Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or final approval of manuscript.


(1.) Blennow K, Hampel H, Weiner M, Zetterberg H. Cerebrospinal fluid and plasma biomarkers in Alzheimer disease. Nat Rev Neurol 2010;6:131-44.

(2.) Duits FH, Teunissen CE, Bouwman FH, Visser PJ, Mattsson N, Zetterberg H, et al. The cerebrospinal fluid "Alzheimer profile": easily said, but what does it mean? Alzheimers Dement 2014;10:713-23.

(3.) Strozyk D, Blennow K, White LR, Launer LJ. CSF A8 42 levels correlate with amyloid-neuropathology in a population-based autopsy study. Neurology 2003;60: 652-6.

(4.) Tapiola T, Alafuzoff I, Herukka S-K, Parkkinen L, Hartikainen P, Soininen H, et al. Cerebrospinal fluid 8-amyloid 42 and tau proteins as biomarkers of Alzheimer-type pathologic changes in the brain. Arch Neurol 2009;66:382-9.

(5.) Vos SJB, Verhey F, Frolich L, Kornhuber J, Wiltfang J, Maier W, et al. Prevalence and prognosis of Alzheimer's disease at the mild cognitive impairment stage. Brain 2015;138:1327-38.

(6.) Toledo JB, Zetterberg H, van Harten AC, Glodzik L, Martinez-Lage P, Bocchio-Chiavetto L, et al. Alzheimer's disease cerebrospinal fluid biomarker in cognitively normal subjects. Brain 2015;138:2701-15.

(7.) Sperling RA, Aisen PS, Beckett LA, Bennett DA, Craft S, Fagan AM, et al. Toward defining the preclinical stages of Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011;7:280-92.

(8.) Albert MS, DeKosky ST, Dickson D, Dubois B, Feldman HH, Fox NC, et al. The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011;7: 270-9.

(9.) McKhann GM, Knopman DS, Chertkow H, Hyman BT, Jack CR, Kawas CH, et al. The diagnosis of dementia due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association work groups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011;7:263-9.

(10.) Vos SJB, Visser PJ, Verhey F, Aalten P, Knol D, Ramakers I, et al. Variability of CSF Alzheimer's disease biomarkers: implications for clinical practice. PLoS One 2014;9:e100784-9.

(11.) Mattsson N, Andreasson U, Persson S, Arai H, Batish SD, Bernardini S, et al. The Alzheimer's Association external quality control program for cerebrospinal fluid biomarkers. Alzheimers Dement 2011;7:386 -95.

(12.) Bertens D, Tijms BM, Scheltens P, Teunissen CE, Visser PJ. Unbiased estimates of cerebrospinal fluid 8-amyloid 1-42 cutoffs in a large memory clinic population. Alzheimer's Res Ther 2017;9:8.

(13.) Willemse EAJ, van Uffelen KWJ, van der Flier WM, Teunissen CE. Effect of long-term storage in biobanks on cerebrospinal fluid biomarker A[beta]1-42, T-tau, and P-tau values. Alzheimers Dement (Amst) 2017;8:4550.

(14.) Schindler SE, Sutphen CL, Teunissen C, McCue LM, Morris JC, Holtzman DM, et al. Upward drift in cerebrospinal fluid amyloid-[beta] 42 assay values for more than 10 years. Alzheimers Dement [Epub ahead of print 2017 Jul 12].

(15.) Mulder C, Verwey NA, Van Der Flier WM, Bouwman FH, KokA, van Elk EJ, et al. Amyloid- (1-42), total tau, and phosphorylated tau as cerebrospinal fluid biomarkers for the diagnosis of Alzheimer disease. Clin Chem 2010;56:248-53.

(16.) Zwan M, van Harten A, Ossenkoppele R, Bouwman F, Teunissen C, Adriaanse S, et al. Concordance between cerebrospinal fluid biomarkers and [11C]PIB PET in a memory clinic cohort. J Alzheimers Dis 2014;41: 801-7.

(17.) Zwan MD, Rinne JO, Hasselbalch SG, Nordberg A, Lleo A, Herukka S-K, et al. Use of amyloid-PET to determine cutpoints for CSF markers: a multicenter study. Neurology 2016;86:50-8.

(18.) De Meyer G, Shapiro F, Vanderstichele H, Vanmechelen E, Engelborghs S, De Deyn PP, et al. Diagnosis-independent Alzheimer disease biomarker signature in cognitively normal elderly people. Arch Neurol 2010;67:949-56.

(19.) Nordberg A, Carter SF, Rinne J, Drzezga A, Brooks DJ, Vandenberghe R, et al. A European multicentre PET study of fibrillar amyloid in Alzheimer's disease. Eur J Nucl Med Mol Imaging 2013;40:104-14.

(20.) Ewers M, Insel P, Jagust WJ, Shaw L, Trojanowski JJQ, Aisen P, et al. CSF biomarker and PIB-PET-derived beta-amyloid signature predicts metabolic, gray matter, and cognitive changes in nondemented subjects. Cereb Cortex 2012;22:1993-2004.

(21.) van der Flier WM, Pijnenburg YAL, Prins N, Lemstra AW, Bouwman FH, Teunissen CE, et al. Optimizing patient care and research: the Amsterdam dementia cohort. J Alzheimers Dis 2014;41:313-27.

(22.) Ossenkoppele R, Prins ND, Pijnenburg YAL, Lemstra AW, van der Flier WM, Adriaanse SF, et al. Impact of molecular imaging on the diagnostic process in a memory clinic. Alzheimers Dement 2013;9:414-21.

(23.) Zwan MD, Bouwman FH, Konijnenberg E, van der Flier WM, Lammertsma AA, Verhey FRJ, et al. Diagnostic impact of [(18)F]flutemetamol PET in early-onset dementia. Alzheimers Res Ther 2017;9:2.

(24.) de Wilde A, van Maurik IS, Kunneman M, Bouwman F, Zwan M, Willemse EAJ, et al. Alzheimer's biomarkers in daily practice (ABIDE) project: rationale and design. Alzheimers Dement (Amst)2017;6:143-51.

(25.) Jansen WJ, Ossenkoppele R, Knol DL, Tijms BM, Scheltens P, Verhey FRJ, et al. Prevalence of cerebral amyloid pathology in persons without dementia. JAMA 2015; 313:1924.

(26.) Pitschke M, Prior R, Haupt M, Riesner D. Detection of single amyloid 8-protein aggregates in the cerebrospinal fluid of Alzheimer's patients by fluorescence correlation spectroscopy. Nat Med 1998;4:832-4.

(27.) Holtta M, Hansson O, Andreasson U, Hertze J, Minthon L, Nagga K, et al. Evaluating amyloid-[beta] oligomers in cerebrospinal fluid as a biomarker for Alzheimer's disease. PLoS One 2013;8:e66381-8.

(28.) Wang-Dietrich L, Funke SA, Kuhbach K, Wang K, Besmehn A, Willbold S, et al. The amyloid-8 oligomer count in cerebrospinal fluid is a biomarker for Alzheimer's disease. J Alzheimers Dis 2013;34:985-94.

(29.) Englund H, Degerman Gunnarsson M, Brundin RM, Hedlund M, Kilander L, Lannfelt L, et al. Oligomerization partially explains the lowering of a beta 42 in Alzheimer's disease cerebrospinal fluid. Neurodegener Dis 2009;6:139-47.

(30.) Andreasen N, Blennow K. [beta]-Amyloid (A[beta]) protein in cerebrospinal fluid as a biomarker for Alzheimer's disease. Peptides 2002;23:1205-14.

(31.) Schuster J, Funke SA. Methods for the specific detection and quantitation of amyloid-beta] oligomers in cerebrospinal fluid. J Alzheimers Dis 2016;53:53-67.

(32.) Sancesario G. Increased detection of A[beta] oligomers in the cerebrospinal fluid of Alzheimer's disease: fact or artifact? J Mol Biomark Diagn 2012;03:1-2.

(33.) Bruggink KA, Jongbloed W, Biemans EALM, Veerhuis R, Claassen JAHR, Kuiperij HB, et al. Amyloid-beta oligomer detection by ELISA in cerebrospinal fluid and brain tissue. Anal Biochem 2013;433:112-20.

(34.) Janssen L, Sobott F, De Deyn PP, Van Dam D. Signal loss due to oligomerization in ELISA analysis of amyloid-beta can be recovered by a novel sample pre-treatment method. MethodsX 2015;2:112-23.

Betty M. Tijms, [1] * Eline A.J. Willemse, [2] Marissa D. Zwan, [1] Sandra D. Mulder, [1] Pieter Jelle Visser, [1,3] Bart N.M. van Berckel, [4] Wiesje M. van der Flier, [1,5] Philip Scheltens, [1] and Charlotte E. Teunissen [2]

[1] Alzheimer Center and Department of Neurology, VUmc, Amsterdam Neuroscience, Amsterdam, the Netherlands; [2] Neurochemistry Laboratory and Biobank, Department of Clinical Chemistry, VUmc, Amsterdam Neuroscience, Amsterdam, the Netherlands; [3] Department of Psychiatry and Neuropsychology, School for Mental Health and Neuroscience, Maastricht University, Maastricht, the Netherlands; [4] Department of Radiology and Nuclear Medicine, Amsterdam Neuroscience, VU University Medical Center, Amsterdam, the Netherlands; [5] Department of Epidemiology and Biostatistics, VUmc, Amsterdam Neuroscience, Amsterdam, the Netherlands.

* Address correspondence to this author at: Alzheimer Center and Department of Neurology, VU University Medical Center, PO Box 7057,1007 MB, Amsterdam, the Netherlands. Fax +31-204440816; e-mail

Received August 22,2017; accepted November 20,2017.

Previously published online at DOI: 10.1373/clinchem.2017.281055

[6] Nonstandard abbreviations: AD, Alzheimer's disease; A[beta] 1-42, amyloid-[beta] 1-42; CSF, cerebrospinal fluid; PET, positron emission tomography; MCI, mild cognitive impairment; SCD, subjective cognitive decline; VUmc, VU University Medical Center; MMSE, Mini-Mental State Examination.

[7] Human Genes: APOE, apolipoprotein E.

Caption: Fig. 1. Cohort characteristics plotted according to year of first visit.

Number of individuals (A); relative proportions of diagnostic groups (B); age (C); MMSE scores [data were missing for 179 (4%) individuals] (D); sex distribution (E); and proportion of APOE [epsilon]4 carriers [data missing for 383 (9%) individuals] (F).

Caption: Fig. 2. Dashed horizontal line throughout the figure shows the uncorrected cutpoint 680 pg/mL (12) unless specified otherwise.

A[beta] 1-42 concentrations increase overtime (A). Year-specific cutpoints determined with mixture modeling increase over time (B). The dotted lines reflect the 95% CIs of the cutpoints. Uncorrected Aft 1-42 concentrations with individuals colored blue when having abnormal amyloid or orange when having normal amyloid (C). Drift-corrected year-specific cutpoints (D). The dotted lines reflect the 95% CIs of the cutpoints. Drift-corrected A[beta] 1-42 concentrations with cutpoint of 813 pg/mL (reflected by dashed line) (E).

Caption: Fig. 3. Annual concordance (top), sensitivity (middle), and specificity (bottom) estimates for corrected and uncorrected A[beta]1-42 values with different cutpoints (c1 = drift corrected; c2 = uncorrected 550 pg/mL; c3 = uncorrected 640 pg/mL; c4 = uncorrected 680 pg/mL).

PET data were available from 2006 onward; owing to small sample sizes, years 2006 to 2008 were taken together to obtain reliable 95% CI. (a) P < 0.05 c2 and all other cutpoints; (b) P < 0.05 c2 and c4; (c) P < 0.05 c3 and c1; (d) P < 0.05 c3 and c1, c4; (e) P < 0.05 c4 and c1.
Table 1. Clinical and biological description of subjects
seen between 2000 and 2016 with available CSF
A[beta]-42 measurements.

                                     Total group
Characteristic                        (n = 4397)

Age, years [mean (SD)]               63.5 (9.1)
Female, n (%)                        1881 (43)
APOE [epsilon]4 carrier, n (%) (a)   1940 (44)
MMSE, median (IQR) (b)               27 (26-29)
Subject group, n (%)
  AD dementia                        1363 (31)
  Non-AD dementia                     710 (16)
  MCI                                 624 (14)
  SCD                                 771 (18)
  Other                               929 (21)

IQR, interquartile range.

(a) Data missing for 383 subjects.

(b) Data missing for 179 subjects.

Table 2. Comparison of uncorrected and drift-corrected values
for different cutpoints in terms of labeling individuals
according to amyloid abnormality, and sensitivity-specificity
analyses to detect AD dementia.

                                Amyloid status         Amyloid PET
                                                        (n = 602)

                               False       False
                            positive,    negative,
Cutpoint                       n (%)       n (%)      Concordance %

550 pg/mL uncorrected (a)     4 (0.1)     676 (15)    70 (66-74)
640 pg/mL uncorrected (b)    64 (1.5)     225 (5)     81 (78-84) (c)
680 pg/mL uncorrected (d)   149 (3)       133 (3)     84 (81-87) (c)
813 pg/mL drift corrected   Reference   Reference     88 (85-90) (f)

                                AD dementia vs
                              all other subjects

Cutpoint                          (95% CI)

550 pg/mL uncorrected (a)   0.66 (0.64-0.69)
640 pg/mL uncorrected (b)   0.85 (0.83-0.86)
680 pg/mL uncorrected (d)   0.89 (0.87-0.91) (e)
813 pg/mL drift corrected   0.90 (0.89-0.92) (e)

                                AD dementia vs
                              all other subjects

Cutpoint                         (95% CI)

550 pg/mL uncorrected (a)   0.81 (0.79-0.82)
640 pg/mL uncorrected (b)   0.72 (0.71-0.74)
680 pg/mL uncorrected (d)   0.68 (0.67-0.70) (e)
813 pg/mL drift corrected   0.69 (0.67-0.70) (e)

PET data were available for the period 2006 to 2016.

(a) From Mulder et al. (15).

(b) From Zwan et al. (16).

(c) Significantly different from uncorrected values with
cutpoints 550 pg/mL and 640 pg/mL.

(d) From Bertens et al. (12).

(e) Significantly different from 550 pg/mL and 640 pg/mL.

(f) Significantly different from 550 pg/mL, 640 pg/mL,
and 680 pg/mL.

Table 3. Percentage of SCD and MCI participants with abnormal
amyloid as determined with corrected and uncorrected CSF
amyloid values for different cutpoints and for the period
2014 to 2016.


                                                 Corrected   Uncorrected

Cognitive        Age, years                         813        550
status           [mean (SD)]    Reference (a)      pg/mL    pg/mL (b)

MCI (n = 142)       65(8)      42.8 (38.7-47.1)     44.4     13.4 (e)
SCD (n = 207)       59(8)      17.4 (11.6-25.2)     17.9     3.9 (e)



Cognitive           640         680
status           pg/mL (c)   pg/mL (d)

MCI (n = 142)     22.5 (e)    29.5 (e)
SCD (n = 207)     8.7 (e)     11.5 (e)

(a) From Jansen et al. (25).

(b) From Mulder et al. (15).

(c) From Zwan et al. (16).

(d) From Bertens et al.(12).

(e) Significantly different from reference prevalence.
COPYRIGHT 2018 American Association for Clinical Chemistry, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2018 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Proteomics and Protein Markers
Author:Tijms, Betty M.; Willemse, Eline A.J.; Zwan, Marissa D.; Mulder, Sandra D.; Visser, Pieter Jelle; va
Publication:Clinical Chemistry
Date:Mar 1, 2018
Previous Article:Leukocyte Counts Based on DNA Methylation at Individual Cytosines.
Next Article:Plasma Steroid Metabolome Profiling for Diagnosis and Subtyping Patients with Cushing Syndrome.

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters