Prevalence and Clinical Implications of a β-Amyloid–Negative, Tau-Positive Cerebrospinal Fluid Biomarker Profile in Alzheimer Disease

This cohort study assesses the memory clinic prevalence and prognosis of individuals with a β-amyloid–negative, tau-positive cerebrospinal fluid biomarker profile.


UGOT
The University of Gothenburg cohort is based on clinical laboratory routine data extracted from the local database at the neurochemistry laboratory at Sahlgrenska University Hospital, Mölndal, Sweden.In this study, we included all participants over the age of 50 with data on Aβ42/40 and p-tau from the same time of collection between 7 th of November 2019 and 18 th of January 2021.All participants had data on age and biomarker concentrations, while clinical information on diagnosis and cognitive function was not available, as memory and neurology clinics send their referrals without accompanying clinical data.However, they are encouraged to send referrals for these analyses only for individuals above the age of 50 where there is a suspicion of a cognitive disorder.

ADNI
Data used in the preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu), which is an ongoing longitudinal observational study.The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD.The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early Alzheimer's disease (AD).For up-to-date information, see www.adni-info.org.

WISC
The Wisconsin Registry for Alzheimer's Prevention (WRAP) is an ongoing longitudinal observational cohort study that was initiated in 2001.It is enriched with people with a parental history of probable AD dementia.Recruitment sources included memory clinics in which a parent was diagnosed or treated, limited radio and newspaper advertisements, community outreach events, and word of mouth.Participants generally meet the following inclusion criteria at study entry: age 40-65 years; fluent English speaker; visual and auditory acuity adequate for neuropsychological testing; good health with no diseases expected to interfere with study participation over time.Participants are excluded from enrollment if they have a prior diagnosis of dementia or evidence of dementia at baseline testing.
Longitudinal data collection for the Wisconsin Alzheimer's Disease Research Center (WI-ADRC) began in 2009 and is ongoing.Participants are recruited from Wisconsin Alzheimer's Institute-Affiliated Dementia Diagnostic Clinics, via community outreach events and lectures, media advertising, community partner organizations, and word of mouth.The cohort is enriched with middle-aged to older adult participants who have a parental history of probable AD and include participants across the clinical and biologic AD continuum.eMethods 2. CSF handling and assays in UGOT, ADNI and WISC.

UGOT
Cerebrospinal fluid (CSF) was collected at the referring center using standard procedures for lumbar punctures (LP) and was then collected in a 1.5 mL polypropylene low-binding tube, according to recent recommendations. 1 Next, the tube was transported at room temperature to the neurochemistry laboratory at Sahlgrenska University Hospital, Mölndal, Sweden within 24 hours of collection.Otherwise, the sample was partitioned in aliquots and transported frozen on dry ice, in accordance with international recommendations. 1 Previously published cut-offs were applied, determining an abnormal Aβ42/40 status when ≤0.072 using gaussian mixed modeling (GMM) and ≥50.2 pg/mL for p-tau using receiver operating characteristics (ROC) to best separate AD vs. controls. 2

ADNI
Up to date methods on CSF collection in ADNI can be found elsewhere (https://adni.loni.usc.edu/methods/).In ADNI, a liquid chromatography mass spectrometrybased assay was used to determine the CSF Aβ42/40 ratio, whereas a fully automated Elecsys ® assay (Roche Diagnostics International Ltd, Rotkreuz, Switzerland) was used to measure CSF p-tau.Briefly, the CSF handling was in accordance with recent international guidelines found elsewhere. 1Previously published ADNI cutoffs of ≤ 0.0138 (using GMM) and ≥24 pg/mL (using a method maximizing identification of progressors from MCI to dementia) were used for determining abnormality and CSF Aβ42/40 and p-tau, respectively. 3,4 ADNI, the baseline visit comprised both LP and cognitive testing for all individuals.

WISC
CSF samples are collected in the core WRAP and WI-ADRC studies as well as linked studies.A center-wide standard pre-analytical protocol was used to collect approximately 22 mL of CSF that is subsequently gently mixed to remove collection gradients, partitioned into 0.5-mL aliquots in -1.5-mL polypropylene tubes, and stored at −80°C.CSF Aβ42, Aβ40 (research use only), and p-tau were quantified using the fully automated Elecsys ® assays (Roche Diagnostics International Ltd, Rotkreuz, Switzerland).Abnormal status was determined as ≤ .046for Aβ42/Aβ40 (using ROC-based classification with Aβ PET as the diagnostic standard) and ≥ 24.8 pg/mL (+2 SD above Aβ42/Aβ40 negative participants) for p-tau.More details on methods used to determine positivity can be found elsewhere. 5F AT status was determined from the baseline LP, which was not uniformly performed at the first cognitive visit.Baseline cognitive testing occurred on average 2.1 (SD = 2.8) years prior to baseline LP (median = 0.47 years prior; interquartile range = 2.90 years prior -0.18 years prior; full range: 11.31 years prior -2.47 years after).The last cognitive assessment occurred on average 4.8 (SD = 2.9) years after baseline LP (median = 5.6 years after; interquartile range: 2.14 years after -7.02 years after; full range: 1.45 years prior -10.23 years after).The A-T+ group did not differ from other status groups on time interval between baseline LP and baseline or final cognitive assessments.eMethods 3. Imaging acquisition and pre-processing methods.

ADNI
A global cortical composite normalizing to cerebellar grey matter was used for Aβ-and tau-PET. 6,7The composite [18F]-florbetapir standard uptake value ratio (SUVr) value was generated for each participant by using the average SUVRs from the precuneus, prefrontal, orbitofrontal, parietal, temporal, anterior, and posterior cingulate cortices.A validated global composite was used for FDG-PET. 7Hippocampal volume was normalized to total intracranial volume. 8Further details on these widely used ADNI measures and its associated processing pipelines have been previously described. 7,8The [ 18 F]-flortaucipir tau-PET meta-ROI composite SUVR was generated for each participant by using the average SUVRs from the entorhinal, amygdala, fusiform, inferior and middle temporal cortices, as previously described. 6

WISC
[ 18 F]-MK6240 was used to quantify aggregated tau using previously established methods. 8riefly, the inferior cerebellum was used as reference, with SUVr calculated for the average uptake in the left and right entorhinal cortex.CSF AT status for the tau PET analysis was determined from the same LP as used for the cognitive trajectory analysis.That LP occurred on average 7.1 (SD = 2.9) years prior to tau PET.The time between tau-PET scan and LP did not differ between CSF AT groups.To enable longitudinal analyses of the cognitive trajectories in this study, cognitive tests were used to generate standardized preclinical Alzheimer's cognitive composite (PACC) z-scores.

ADNI
In ADNI, we modeled the PACC z-scores at each timepoint as a function of age as outcome.However, the composite is instead made up by the Alzheimer's Disease Assessment scale, which constitutes the Cognitive Subscale Delayed Word Recall, Logical Memory Delayed Recall, MMSE, and (log-transformed) Trail-Making Test B Time to Completion; thus, creating the modified PACC (mPACC). 9,10 ADNI, individuals with MCI had a mean [SD] follow up of 4.61 [2.83] years, whereas CU individuals were followed for 6.02 [3.55] years.

WISC
In the WISC sample, three cognitive tests were used (thus generating the PACC-3): Total Recall score from Rey auditory verbal learning test (RAVLT), Trail Making Test Part B, and Logical Memory IIa (from the Wechsler Memory Scale) 10 .Following implementation of the National Alzheimer's Coordinating Center (NACC) Uniform Data Set (UDS) version 3, the WI-ADRC switched to using Craft story as the measure of story recall, thus a published crosswalk 9 was used to estimate Logical Memory IIa for those observations. 11PACC-3 scores were standardized using the mean and standard deviation from the first cognitive assessment of cognitively unimpaired participants in the WRAP and WI-ADRC cohorts.
To increase the robustness of our results to potential analytical and classification errors, we performed a drift and a grey-scale analysis.The drift analysis examined the influence of measurement abnormalities on our prevalence results and investigated the analytical performance for the Lumipulse instrument over time.The time frame 071119-181021 investigated was then divided into three equal episodes (A, B and C) as stated in eFigure 1. Subsequently, the three episodes were all internally compared with paired t-tests.Then, a grey-scale analysis was performed, showing the potential effects the present cut-offs might have on the prevalence outcome.
To accommodate this, the cut-offs for p-tau and Aβ42/40 were increased and lowered 15 %, respectively.Individuals having values deviating up to 15% above the cut-off value for p-tau and below the cut-off value for Aβ42/40 were then excluded, whereafter the prevalence was computed again.The cut-offs used in the greyscale analysis were Aβ42/40 ≥0.61 and p-tau ≤58 ng/L.We thus increased the thresholds of pathology and decreased the sensitivity of the cut-offs.Subsequently, it was compared to the initial prevalence analysis.The results are found in eTable 1. eTable 1. Greyscale analysis in the UGOT dataset.

eFigure 1 .
Age-stratified prevalence of CSF AT profiles using CSF Aβ42 alone for "A".The colored dots represent the prevalence in % at each age (in years) of each biomarker category (blue, A-T-; orange, A+T-; red, A+T+; green, A-T+) based on the cut-offs used in clinical routine.The solid lines represent corresponding locally estimated scatterplot smoothing (LOESS) regression lines, with shaded areas indicating 95% CIs.This indicated that when using CSF Aβ42 alone, instead of the recommended CSF Aβ42/40 ratio, all three groups with abnormal biomarkers (A+T-, A+T+, and A-T+) present similar, nearly indistinct, trends for prevalence increases with increases in age.

eFigure 2 .
Drift analyses in the UGOT dataset.The graph displays group comparisons of (A) t-tau, (B) p-tau and (C) Aβ42/40*10 across time periods 1 (November 7 th , 2019 -July 1 st , 2020), 2 (July 2 nd , 2020 -February 23 rd , 2021), and 3 (February 24 th , 2021 -January 18 th , 2021) in the UGOT cohort.Group comparisons were performed using a one-way analysis of variance.A two-sided p-value <0.05 was considered significant.In panel A, concentrations of t-tau are presented on a log10 scale for visualization purposes.Also, for visualization purposes, three outliers with Aβ42/40*10 ratio > 2.6 was excluded from the graph but included in all statistical analyses.
Grey-scale analysis was performed so that cut-offs for p-tau and Aβ42/40 were increased and lowered 15 %, respectively.Individuals having values deviating up to 15% above the cut-off value for p-tau and below the cut-off value for Aβ42/40 were then excluded, whereafter the prevalence was computed again.The cut-offs used in the greyscale analysis were Aβ42/40 ≥0.61 and p-tau ≤58 ng/L, whereas the original cut-offs were ≤0.072 for Aβ42/40 and ≥50.2 pg/mL for p-tau.eTable2.Linear mixed effects models of baseline CSF AT profiles and longitudinal mPACC in CU and MCI individuals in ADNI.Linear mixed effects models of baseline CSF AT profiles and longitudinal PACC-3 in WISC.Sensitivity analysis using linear mixed effects models of baseline CSF AT profiles and longitudinal mPACC in CU and MCI individuals in ADNI using 15% higher p-tau cut-offs.Sensitivity analysis using linear mixed effects models of baseline CSF AT profiles and longitudinal PACC-3 in WISC using 15% higher p-tau cut-offs.Linear mixed effects models of baseline CSF AT profiles and longitudinal Aβ PET in CU and MCI individuals in ADNI.Linear mixed effects models of baseline CSF AT profiles and longitudinal FDG PET in CU and MCI individuals in ADNI.Linear mixed effects models of baseline CSF AT profiles and longitudinal MRI in CU and MCI individuals in ADNI.Sensitivity analysis using linear mixed effects models of baseline CSF AT profiles and longitudinal Aβ PET in CU and MCI individuals in ADNI using 15% higher p-tau cut-offs.Sensitivity analysis using linear mixed effects models of baseline CSF AT profiles and longitudinal FDG PET in CU and MCI individuals in ADNI using 15% higher p-tau cut-offs.Sensitivity analysis using linear mixed effects models of baseline CSF AT profiles and longitudinal MRI in CU and MCI individuals in ADNI using 15% higher ptau cut-offs.APOEε4 status and years of education as covariates were used in both cohorts.A+/-indicates CSF Aβ42/40 binary status, and T+/-indicates CSF p-tau181 binary status.Cross-sectional tau PET analysis in WISCThe A-T-, A+T-, A+T+ and A-T+ profiles constituted 179, 26, 16 and 6 individuals, respectively.Linear models, including age, sex, APOEε4 status and years of education as covariates were used in both cohorts.A+/-indicates CSF Aβ42/40 binary status, and T+/-indicates CSF p-tau181 binary status.