Concordance of Lumipulse cerebrospinal fluid t-tau/Aβ42 ratio with amyloid PET status

Abstract

INTRODUCTION:

Cerebrospinal fluid (CSF) biomarkers can identify individuals with Alzheimer disease (AD) pathology (e.g., amyloid plaques, neurofibrillary tangles), but defined analyte cut-points using high-throughput automated assays are necessary for general clinical use.

METHODS:

CSF Aβ42, t-tau, and t-tau/Aβ42 were quantified by the LUMIPULSE platform in two test cohorts (A/B: Eisai BAN2401–201/MISSION AD E2609–301/302, n=138; C: Knight AD Research Center, n=198), and receiver operating characteristic (ROC) curve analyses defined cut-points corresponding best to amyloid determinations using PET imaging. The best-performing cut-point was then validated as a predictor of amyloid status in an independent cohort (D: MISSION AD E2609–301/302; n=240).

RESULTS:

Virtually identical t-tau/Aβ42 cut-points (~0.54) performed best in both test cohorts and with similar accuracy (areas under ROC curve [AUC] (A/B: 0.95; C: 0.94). The cut-point yielded an overall percent agreement with amyloid PET of 85.0% in validation cohort D.

DISCUSSION:

LUMIPULSE CSF biomarker measures with validated cut-points have clinical utility in identifying AD pathology.

1. Introduction

Alzheimer disease (AD) is characterized by neuronal and synaptic degeneration accompanied by intracellular aggregates of hyperphosphorylated tau (neurofibrillary tangles) and extracellular amyloid plaques primarily comprised of amyloid β peptide (Aβ42) [1]. The clinical symptoms of AD develop insidiously and progress slowly, most commonly starting with memory impairment followed by deterioration in other cognitive skills and behavioral functions, resulting in progressive dementia with a gradual loss of ability to perform activities of daily living [2].

Biofluid biomarkers and imaging methodologies can be used to identify individuals with AD pathology at the asymptomatic stage and throughout the disease continuum. Abnormalities in amyloid-related biomarkers have been shown to precede graded increases of tau-related neuronal injury markers during the asymptomatic period, with further accumulation of amyloid and increasing neuronal injury continuing in relation to the onset of cognitive symptoms [3–5]. Consequently, biomarkers of AD pathology are increasingly being used in trials of potential disease-modifying therapies for purposes of identification of eligible patients, evaluating therapeutic target engagement and determining clinical trial outcome [6–9]. Amyloid positron emission tomography (PET) is a U.S. Food and Drug Administration (FDA)-cleared methodology which confirms that lack of neuritic plaques is inconsistent with a neuropathologic AD diagnosis [10]. Since amyloid PET tracers have been shown to have high agreement with histopathologic amyloid aggregates [11–13], we use amyloid PET results in the present study as the standard for establishing a cerebrospinal fluid (CSF) biomarker cut-point for amyloid positivity. The development of CSF biomarker cut-points may be useful to support the biomarker requirements defined by the National Institute on Aging and Alzheimer’s Association (NIA-AA) A-T-(N) Research Framework, where AD is defined by its underlying pathologic processes (Amyloid-Tau-Neurodegeneration) that can be documented by postmortem examination or in vivo biomarkers [14].

At present, there are no FDA-cleared CSF biomarker assays for use in clinical routine. However, “research use only” CSF biomarkers are increasingly employed to support inclusion in clinical trials, e.g. to determine eligibility for presence of amyloid [15,16]. Indeed, there is an emerging use of CSF biomarker measures in parallel with amyloid PET to establish amyloid positivity when screening subjects for clinical trials, based on the finding that there is a high concordance between those measures [17–21]. This necessitates that there is a robustness in the CSF biomarker assay and platform performance and, most importantly, a well-defined cut point for amyloid positivity [22]. Across the field, cut-points for identification of amyloid positivity associated with a diagnosis of AD and/or mild cognitive impairment (MCI) due to AD have been shown to be impacted by factors such as subject demographics, sample collection and handling procedures as well as assay platform used for analyses [23]. Thus, universal use of CSF biomarkers has proven challenging. In the present study, the best-performing CSF biomarker cut-point(s) for amyloid PET positivity was tested and defined using the new fully automated Lumipulse® G β-Amyloid 1–42 (Aβ42) and Lumipulse G Total-Tau (t-tau) assays (Fujirebio, Malvern, PA). To develop a CSF cut-point where context of use would be concordance with amyloid PET tracer positivity, CSF samples from a total of 336 individuals across three separate cohorts were analyzed independently by the Lumipulse Aβ42 and t-tau assays. Cohorts A and B included baseline data from two clinical AD trials sponsored by Eisai (n=95 and 43, respectively), whereas Cohort C (n=198) included data from participants from the Knight Alzheimer’s Disease Research Center at Washington University in St. Louis. These cohorts were considered the test cohorts with respect to establishing a CSF cut-point(s). The utility of the selected most robust biomarker cut-point defined in Cohorts A-C was then evaluated as a predictor of amyloid PET positivity in an independent validation cohort from an Eisai clinical trial enrolling early AD patients (Cohort D, n=240).

2. Materials and methods

2.1. Subjects

Cohort A/B (Test Set):

Baseline CSF samples from 138 subjects used in the test sample set were selected from two different AD clinical trials involving more than 30 sites. Cohort A samples originated from the completed BAN2401- 201 (anti-Aβ protofibril antibody) clinical trial, while Cohort B samples were from the ongoing MISSION AD E2609–301/302 (BACE inhibitor) trials. In these trials, amyloid PET and/or CSF could be used at baseline to support trial eligibility, which required amyloid burden positivity (Table 1). Both Cohort A/B trials screened for subjects with early AD (including MCI due to AD – intermediate likelihood and mild AD dementia). Prior to the CSF biomarker cut-point being established via Lumipulse as described herein, subjects using CSF to define amyloid positivity required a sample to be analyzed at the Biomarker Core of the Alzheimer’s Disease Neuroimaging Initiative (ADNI) using an Aβ42 cutoff established on the INNO-BIA AlzBio3 assay (<250pg/mL)[24]. Overall trial(s) enrollment criteria are listed in the Supplemental Table. The combined Cohort A/B in the present study was comprised of individuals who had both CSF samples and amyloid PET data obtained within the screening window (~60–80 days) available for analysis. All trial-related documents were reviewed and approved by an Institutional Review Board (IRB) or Independent Ethics Committee (IEC) constituted and functioning in accordance with International Conference on Harmonization (ICH) E6 (Good Clinical Practice [GCP]) (https://ichgcp.net/) and any local regulations. All subjects participating in the trials provided written informed consent, and subject screening and consent was performed under GCP regulations.

Table 1.

COHORT	A	B	C	D
Test/Validation	Test (combined with B)	Test (combined with A)	Test	Validation
Study	BAN2401–201	E2609–301/302	Washington University Knight ADRC	E2609–301/302
N	95	43	198	240
PET positivity (%)	80%	26%	25%	27%
PET tracer	Florbetapir	Florbetaben Florbetapir	Pittsburgh Compound B (PIB)	Florbetaben Florbetapir Flutametamol

BAN2401-201 (https://clinicaltrials.gov/ct2/show/NCT01767311)

E2609-301 (https://clinicaltrials.gov/ct2/show/NCT02956486)

E2609-302 (https://clinicaltrials.gov/ct2/show/NCT03036280)

Florbetapir (Amyvid®); florbetaben (Neuraseq®); flutametamol (Vizamyl®)

Cohort C (Test Set):

Participants were community-dwelling research volunteers enrolled in cohort studies of normal aging and dementia at the Knight Alzheimer’s Disease Research Center (ADRC) at Washington University in St. Louis. Participants had no neurological, psychiatric, or systemic medical illness that might compromise longitudinal study participation, nor was any medical contraindication to lumbar puncture (LP) or PET present. All procedures were approved by the Washington University Human Research Protection Office (HRPO), and written informed consent was obtained from each participant. For inclusion in the study, participants had to undergo a clinical assessment, CSF collection and amyloid PET within a 365-day period. Samples in the present study were from a cohort pre-selected by amyloid PET status to create a 75% PET−negative (PET−) and 25% PET−positive (PET+) sample population (Table 1). Cohort C (n=198) was pre-selected in this manner so that the study population was enriched for possible CSF and PET discordance (e.g. amyloid PET− and CSF Aβ42+). The cohort was analyzed in prior studies[25,26] and was selected without regard to other variables (e.g., cognitive status, age, sex, APOE genotype, etc.).

Cohort D (Validation Set):

The sample set (n=240) used to validate the cut-point (established in Cohorts A/B/C) was comprised of baseline CSF collected from subjects enrolled in the MISSION AD E2609–301/302 trial (Table 1). The subjects of Cohort D also had baseline amyloid PET imaging within 3 months of CSF collection. All CSF and amyloid PET trial baseline assessments in this cohort occurred within three weeks of one another. Samples were obtained prior to randomization into the actual MISSION AD clinical trials. There was no overlap between the subjects in Cohorts B and D, although they were screened for the same clinical trial. The demographic information for all subjects in Cohorts A/B, C and D are summarized in Table 2.

Table 2.

	TEST Cohort A/B			TEST Cohort C			VALIDATION Cohort D
DEMOGRAPHICS	TOTAL	PET−	PET+	TOTAL	PET−	PET+	TOTAL	PET−	PET+
N	130	49	81	187	138	49	233	168	65
Age, y, mean±SD	68.8 ± 7.7	67.0 ± 8.8	69.9 ± 6.9	66.9 ± 9.5	64.9 ± 9.4	72.6 ± 7.2	70.1 ± 7.6	69.5 ± 7.6	71.7 ± 7.5
Female, n (%)	67 (52%)	27 (55%)	40 (49%)	111 (59%)	91 (66%)	20 (41%)	117 (50%)	92 (55%)	25 (38%)
APOE e4, n (%)	72 (55%)	16 (33%)	56 (69%)	72 (39%)	44 (32%)	38 (76%)	71 (30%)	30 (18%)	41 (63%)
Race, n (%)
Caucasian	123 (95%)	44 (90%)	79 (98%)	168 (90%)	123 (89%)	45 (92%)	198 (85%)	141 (84%)	57 (88%)
African American	7 (5%)	5 (10%)	2 (2%)	19 (10%)	15 (11%)	4 (8%)	6 (3%)	4 (2%)	2 (3%)
Japanese	0	0	0	0	0	0	22 (9%)	19 (11%)	3 (5%)
Other Asian	0	0	0	0	0	0	3 (1%)	1 (<1%)	2 (3%)
Other	0	0	0	0	0	0	1 (<1%)	0	1 (2%)
Missing	0	0	0	0	0	0	3 (1%)	3 (2%)	0
Ethnicity, n (%)
Non-Hispanic	106 (82%)	28 (57%)	78 (96%)	186 (99%)	137 (99%)	49 (100%)	115 (49%)	76 (45%)	39 (60%)
Hispanic	24 (18%)	21 (43%)	3 (4%)	1 (<1%)	1 (<1%)	0 (0%)	115 (49%)	89 (53%)	26 (40%)
Missing	0	0	0	0	0	0	3 (1%)	3 (2%)	0
CDR 0/0.5/1/2, n	0/120/10/0	0/46/3/0	0/74/7/0	165/18/3/1	131/7/0/0	34/11/3/1	0/228/0/0	0/163/0/0	0/65/0/0
CDR>0, n (%)	130 (100%)	49 (100%)	81 (100%)	22 (12%)	7 (5%)	15 (31%)	228 (98%)	163 (97%)	65 (100%)
MMSE, mean±SD	25.9 ± 2.1	26.3 ± 2.0	25.7 ± 2.2	28.8 ± 1.9	29.1 ± 1.2	28.0 ± 3.0	26.6 ± 1.6	26.7 ± 1.5	26.5 ± 1.7
CSF BIOMARKERS	TOTAL	PET−	PET+	TOTAL	PET−	PET+	TOTAL	PET−	PET+
Aβ42, pg/mL, mean±SD	745 ± 395	1099 ± 392	532 ± 192	735 ± 263	815 ± 249	510 ± 149	857 ± 362	968 ± 350	572 ± 200
t-tau, pg/mL, mean±SD	538 ± 307	329 ± 158	665 ± 306	382 ± 234	324 ± 185	548 ± 278	418 ± 240	345 ± 182	607 ± 270
t-tau/Aβ42, mean±SD	0.95 ± 0.67	0.35 ± 0.29	1.31 ± 0.58	0.61 ± 0.47	0.42 ± 0.27	1.12 ± 0.52	0.65 ± 0.62	0.44 ± 0.47	1.19 ± 0.64

Abbreviations: Aβ42, β-amyloid peptide 1–42; APOE, apolipoprotein E; CDR, clinical dementia rating; Cohort A/B, from Eisai clinical trials; Cohort C, from the Washington University Knight ADRC; CSF, cerebrospinal fluid; MMSE, Mini-Mental State Examination; PET−, amyloid-negative by positron emission tomography; PET+, amyloid-positive by positron emission tomography; t-tau, total tau.

In Cohort D (at the time of study data freeze): 1) Three PET− participants are missing data for age, sex, APOE genotype, race, ethnicity, CDR, and MMSE; 2) An additional two PET− participants are missing CDR; and 3) One PET+ and one PET− participant are missing APOE genotype. Each of these participants are included in the denominator to calculate the percentages in each factor.

2.2. Amyloid PET

Cohort A/B/D:

Amyloid PET screening for trial eligibility was performed across multiple imaging centers using 18F-florbetapir (Amyvid®) in Cohort A, mostly ¹⁸F-florbetaben (Neuraseq®) in Cohort B, and 18F-florbetapir (Amyvid®), ¹⁸F-florbetaben (Neuraseq®) and ¹⁸F-flutametamol (Vizamyl®) in Cohort D (Table 1). Detailed information for each PET tracer is summarized in the Supplemental Methods.

In brief, visual read for amyloid PET positivity (PET+ or PET−) was performed based on the individual PET tracer labels. PET scans from study imaging centers were transferred to a centralized database and thereafter evaluated by trained and formally qualified readers. The visual reads required agreement between two independent readers applying tracers’ manufacturers’ standard procedures. Regions listed for visual analysis and scan times differ slightly across PET tracers.

Cohort C:

Participants underwent PET imaging with the research tracer ¹¹C-Pittsburgh Compound B (PIB) [27,28] within 12 months of CSF collection. PET imaging was performed with a Siemens 962 HR 1 ECAT PET or Biograph 40 scanner (Siemens/CTI, Knoxville, KY). Structural magnetic resonance imaging (MRI) using MPRAGE T1-weighted images was also acquired, and images were processed using FreeSurfer (http://freesurfer.net/) to derive cortical and subcortical regions of interest (ROI) used in the PET processing. Regional PIB values were converted to standardized uptake value ratios (SUVRs) using cerebellar gray as a reference and partial volume correction using a regional spread function approach [29,30]. Amyloid deposition was summarized using the average measure across seven bilateral brain regions known to be associated with AD (i.e., left and right lateral orbitofrontal, medial orbitofrontal, precuneus, rostral middle frontal, superior frontal, superior temporal, and middle temporal cortices). PIB is a non-FDA cleared PET tracer routinely utilized in the research setting. Since there is no guidance on how to perform visual reads on PIB PET scans, SUVR was used for all analyses, with amyloid positivity defined as a mean cortical SUVR > 1.42 [31].

2.3. CSF collection

Cohorts A/B and D:

CSF samples were obtained under fasting conditions, or at 2–4 hours post meal. Samples (12 mL, single vial) were collected by gravity drip lumbar puncture (LP) using atraumatic Sprotte needles (22/24 gauge). To reduce variability in the preanalytical processing, low protein binding collection tubes (13 mL, #62.515.028, Sarstedt), aliquot tubes (2.0 mL, #022431102, Eppendorf), and pipette tips (# 14200T, Sorenson) were supplied to and used by all trial sites. Samples were stored at −70°C after centrifugation (2000 rpm, 10 min, room temperature) and aliquoting (0.5 mL), prior to shipment on dry ice to an Eisai assigned Contract Research Organization (CRO) for analysis.

Cohort C:

CSF from research volunteers was collected under standard operating procedures (SOPs). Participants underwent LP at 8 am, after overnight fasting. Twenty to 30 mL of CSF was collected in a 50 mL polypropylene tube (Corning #05-538-60, Fisher Scientific) via gravity drip using an atraumatic Sprotte needle (22 gauge). The sample was gently inverted to disrupt potential gradient effects and centrifuged at low speed (2000 × g, 5 min, 4°C) to pellet any cellular debris. Following transfer of all but the bottom ~0.5mL to a second 50 mL tube, CSF was aliquoted (0.5 mL) into 2.0 mL polypropylene tubes (#72.694.416, Sarstedt) and stored at −80°C until analysis [17].

2.4. Measurement of CSF AD biomarkers using Lumipulse G β-Amyloid 1–42 and Lumipulse G Total-Tau

The LUMIPULSE G 1200 instrument from Fujirebio was used for analysis of all CSF samples. Lumipulse uses single analyte, ready-to-use, immunoreaction cartridges with a throughput of 120 tests per hour on this instrument. Each cartridge generates quantitative results for a single analyte within approximately 30 minutes. The Aβ42 and total Tau cartridges, along with their calibrators and kit quality control (QC) samples, were used in the present study. The Lumipulse G β-Amyloid 1–42 and Lumipulse G total Tau assays have been developed using established monoclonal antibodies that are also used in the INNOTEST® assays (Fujirebio, enzyme-linked immunosorbant assays). In the present work, all Aβ42 results have been aligned with the recently certified reference materials as released by the Reference Materials Working Group of the International Federation of Clinical Chemistry (IFCC) [32]. Assay values that fell outside the lower and upper levels of quantitation (LLoQ and ULoQ, respectively) for each analyte were considered to be inaccurate for use and were, therefore, omitted from the statistical analyses. Acceptable ranges were 9–2335 pg/mL for Aβ42 and 141–1919 pg/mL for t-tau. In all study cohorts, one pristine aliquot (i.e., one freeze-thaw cycle) was used to determine CSF biomarker concentrations (i.e., Aβ42 and t-tau). Samples in the test cohorts (Cohorts A/B/C) were evaluated using a single assay lot number (but different lot numbers among the three cohorts), while different assay lot numbers were used for samples in the validation cohort (Cohort D). P-tau181 was not evaluated since the assay was not available for research use in the US at the time of the study.

Analysis of the test set samples (using a single aliquot) from Cohort A/B took place at Fujirebio (Ghent, Belgium), with each cohort tested on a single, yet separate, day. Cohort C samples were analyzed by the Biomarker Core laboratory of the Knight ADRC over three days, with Aβ42 and t-tau analyzed from separate pristine aliquots. Analysis of the validation samples (Cohort D) were performed by the Eisai contracted CRO laboratory (Covance, IN) that was actively analyzing samples from subjects for enrollment in the Phase 3 MISSION AD BACE inhibitor clinical trials, with samples assayed over a period of several months. For cohorts analyzed over more than one day, control and QC samples were analyzed each day in order to ensure assay reproducibility.

2.5. Statistics

For the test cohorts (A/B and C), receiver operating characteristic (ROC) curve analyses were performed to determine the cut-points for Aβ42, t-tau, and the t-tau/Aβ42 ratio that best distinguished amyloid PET+ from PET−individuals. Since PIB is a research tool and considered a “non-reference” standard, in the present work concordance between CSF analyses and PIB amyloid PET was evaluated by percent agreement, not sensitivity or specificity as discussed in prior studies [33]. Positive percent agreement (PPA) was defined as the percentage of individuals considered PET+ who were also positive by a CSF biomarker measure. Negative percent agreement (NPA) was defined as the percentage of individuals considered PET−who were also negative by a CSF biomarker measure. Overall percent agreement (OPA) was defined as the sum of the PET+ individuals who were positive by a CSF biomarker measure and the PET−individuals who were negative by a CSF biomarker measure, divided by the cohort size; thus providing an estimate of accuracy. The CSF biomarker single analyte value (or ratio) with the highest Youden index (PPA + NPA - 1) was identified as the cut-point value. The cut-point for the best performing biomarker in the test cohorts was then used to evaluate performance (PPA, NPA, and OPA) in the validation cohort (D). Analyses were performed using Analyse-it v.5.11 (https://analyse-it.com/).

3. Results

3.1. Characteristics of the test cohorts

Table 2 shows the demographics and summary biomarker data of the test cohorts (Cohorts A/B and C) in which the CSF Lumipulse data were within the defined acceptance ranges for both Aβ42 and t-tau. All Aβ42 concentrations in Cohort A/B were within the assay’s measurement range, whereas eight samples fell outside the range for t-tau (n=6 <141 pg/mL; n=2 >1919 pg/mL), thus providing a dataset of 130 samples evaluated (out of the 138 initially selected). One sample in Cohort C fell outside the range for Aβ42 (>2335 pg/mL), and 10 samples fell outside the range for t-tau (n=9 <141 pg/mL; n=1 >1919 pg/mL), thus, data was generated from 187 samples (out of the 198 initially selected). In both cohorts, the amyloid PET+ group was older and enriched for subjects that showed APOE ε4 positivity. In addition, when comparing between the cohorts, the majority (88%; 165/187) of participants in Cohort C were asymptomatic (cognitively normal, CDR = 0), whereas all the participants in the clinical trial cohort (Cohort A/B) were all symptomatic (CDR > 0) at baseline, in line with the trial design. Also, the mean levels of CSF Aβ42 were lower in the PET+ compared to the PET−groups in both cohorts, whereas levels of t-tau and the ratio of t-tau/Aβ42 were higher (Table 2).

3.2. CSF biomarker cut-points for distinguishing amyloid PET status

The distribution of analyte values in the two PET groups in the test cohorts are shown in Figure 1. Receiver operating characteristic analyses using the optimal Youden Index were performed to determine the cut-points of Aβ42, t-tau, and the t-tau/Aβ42 ratio that best distinguished amyloid PET status in each cohort. In both cohorts, CSF Aβ42 was better at identifying true positives (PET+) than true negatives (PET−) (i.e., PPA > NPA) (Figure 1A, D), whereas the opposite was true for CSF t-tau (i.e., NPA > PPA) (Figure 1B, E). Notably, the ratio of t-tau/Aβ42 performed best (defined by OPA) at discriminating the amyloid PET groups compared to each analyte alone (Figure 1C, F). Importantly, the t-tau/Aβ42 cut-points defining amyloid PET positivity were virtually identical in each cohort (>0.53 for Cohort A/B and >0.54 for Cohort C), as were the areas under the ROC curve (AUC) (0.95 [95% CI=0.91–1.00] for Cohort A/B and 0.94 [95% CI=0.91–0.97] for Cohort C) for distinguishing amyloid PET status (Table 3).

Figure 1. Analyte distribution and receiver operating characteristic (ROC) curves for CSF biomarkers compared to PIB binding.

Histograms (top panels) and ROC curves (bottom panels) as a function of amyloid PET status are shown for Aβ42 (A, D), t-tau (B, E), and the t-tau/Aβ42 ratio (C, F). Data for Cohort A/B are shown in A-C and those for Cohort C in D-F. The positive percent agreement (PPA), negative percent agreement (NPA), and overall percent agreement (OPA) with amyloid PET status are indicated, with the red dashed line and associated number in the histograms identifying the best-performing CSF analyte cut-points. The accompanying area under the ROC curves (AUC) for each analyte is shown in the panels below each histogram. 95% Confidence Intervals are included in parentheses.

Table 3.

SOURCE	BIOMARKER	CUT-POINT	AUC 95% CI)	PPA (95% CI)	NPA (95% CI)	OPA (95% CI)
Test Cohort A/B (n=130)	Aβ42	< 818	0.92 (0.88 – 0.97)	98.8 (93.3 – 100.0)	75.5 (61.1 – 86.7)	90.0 (83.5 – 94.6)
	t-tau	> 456	0.87 (0.80 – 0.93)	74.1 (63.1 – 83.1)	89.8 (77.8 – 96.6)	80.0 (72.1 – 86.5)
	t-tau/Aβ42 ratio	> 0.534	0.95 (0.91 – 1.00)	97.5 (91.4 – 99.7)	89.8 (77.8 – 96.6)	94.6 (89.2 – 97.8)
Test Cohort C (n=187)	Aβ42	< 732	0.86 (0.80 – 0.91)	95.9 (86.0 – 99.5)	63.0 (54.4 – 71.1)	71.7 (64.6 – 78.0)
	t-tau	> 410	0.82 (0.75 – 0.88)	73.5 (58.9 – 85.1)	80.4 (72.8 – 86.7)	78.6 (72.0 – 84.3)
	t-tau/Aβ42 ratio	> 0.541	0.94 (0.91 – 0.97)	93.9 (83.1 – 98.7)	83.3 (76.1 – 89.1)	86.1 (80.3 – 90.7)
Validation Cohort D (n=233)	t-tau/Aβ42 ratio	> 0.54	NA	87.7 (77.2 – 94.5)	83.9 (77.5 – 89.1)	85.0 (79.7 – 89.3)

Abbreviations: Aβ42, β-amyloid peptide 1–42; AUC, area under the receiver operating characteristic curve; Cohort A/B, from Eisai clinical trials; Cohort C, from Washington University Knight ADRC; Cohort D, from Eisai clinical trial; NA, not applicable; NPA, negative percent agreement; OPA, overall percent agreement; PPA, positive percent agreement; t-tau, total tau.

*Samples included have Aβ42 between 9 – 2335 pg/mL and t-tau between 141 – 1919 pg/mL (lower and upper levels of quantitation, respectively)

3.3. Validation of the defined CSF t-tau/Aβ42 cut-point

In order to provide biomarker validation, the CSF t-tau/Aβ42 cut-point defined in the two test cohorts (>0.54) was used to prospectively evaluate its performance for identifying amyloid PET status at baseline in an independent cohort from an ongoing clinical trial. The demographics and biomarker profile of this validation cohort (Cohort D) are shown in Table 2. The PET+ group was slightly older than the PET−group and was enriched in APOE ε4 positivity. By trial design, all individuals were symptomatic, with a baseline CDR of 0.5. Although the majority of participants were Caucasian (~85%), the validation cohort was racially more diverse than the two test cohorts (Cohort A/B; C) which were 95% and 90% Caucasian, respectively. One sample in Cohort D fell outside the measurement range for Aβ42 (>2335 pg/mL), and six samples fell outside the range for t-tau (n=5 <141 pg/mL; n=1 >1919 pg/mL). Similar to the test cohorts, the mean level of CSF Aβ42 was lower in the PET+ versus the PET−group, and levels of t-tau and the t-tau/Aβ42 ratio were higher. Application of the test cohort-derived t-tau/Aβ42 ratio of 0.54 as the cut-point value yielded an OPA of 85.0% in the validation cohort (D), virtually identical to that in Cohort C (86.1%), but lower than that for Cohort A/B (94.6%) (Table 3).

4. Discussion

Alzheimer disease is a debilitating disease that is primarily characterized by Aβ plaques and hyperphosphorylated tau-containing neurofibrillary tangles [1]. Autopsy series have shown that approximately 30% of AD cases are misdiagnosed, an observation that can be attributed to the fact there are many alternative causes of dementia in the elderly. Further, it is known that AD diagnosis is largely based on subjective assessments, such as evaluations of neuropsychological test performance and cognitive status [34]. In recent years, CSF immunoassays and amyloid PET have shown promise as AD biomarkers that may be able to identify individuals that are on a pathological trajectory [18]. The recent releases of the NIA-AA A-T-(N) Research Framework and the Alzheimer’s Association’s Appropriate Use Criteria for Lumbar Puncture and CSF testing have reinforced the positive impact that biomarker knowledge could have on both AD research and clinical practice [14,35].

In the present study, we sought to establish CSF biomarker cut-points (Aβ42, t-tau, and the tau/Aβ42 ratio) on the LUMIPULSE, a novel assay platform, that best distinguished amyloid PET status in two distinct cohorts (Cohort A/B, a cohort combining sample sets from two clinical trials and Cohort C, a research participant cohort). As described above, the t-tau/Aβ42 ratio demonstrated the best performance in both test cohorts relative to either Aβ42 or t-tau alone. This cut-point was then validated as a predictor of amyloid PET positivity in an independent trial cohort (Cohort D), which yielded good overall agreement with amyloid PET status (85%). Taken together, these findings are in alignment with recent research studies that have demonstrated that the use of CSF biomarker ratios improves concordance to amyloid PET over using single biomarkers alone [20,25,26,36]. Of particular relevance is the present finding of a better performance of the cut-point based on tau/Aβ42, since previous studies, in similar populations, have also shown a higher concordance of this ratio to amyloid PET when using other platforms for CSF biomarkers analysis [20,21,37].

Although improved correspondence of CSF biomarker ratios has been established, the current study presents some striking findings. Notably, in test Cohorts A/B and C, nearly identical tau/Aβ42 ratio cut-points of 0.534 and 0.541, respectively, were observed. This observation was surprising given that Cohort A/B was a clinical trial study cohort, while Cohort C was an academic research study cohort. As such, CSF samples in these cohorts were collected under two distinct CSF sample handling procedures that have differences in tube type and sample handling procedures (e.g., aliquoting, sample storage/transport), which are known pre-analytical variables that can significantly impact Aβ42 concentrations [38]. In addition, different amyloid PET tracers were used in Cohorts A/B (Florbetapir, Florbetaben) and Cohort C (PIB). Florbetapir and Florbetaben are FDA-cleared PET tracers evaluated by visual reads, while PIB is a research PET tracer that generally uses the SUVR approach to qualify the presence of amyloid. Although each tracer can be used to determine aspects of amyloid burden, they differ in their underlying method of assessment. Visual reads evaluate differences between white matter signal versus cortical gray in the cortex, whereas SUVR analysis evaluates uptake in ROIs at a defined time (static) over uptake in a defined reference region expected to be free of amyloid. The comparison of both methods for evaluating a subject provide information on amyloid burden that are highly correlated, but differences exist because they are not measuring the same output. Given these cohort-specific differences the t-tau/Aβ42 results in Test Cohorts A-C suggest that automated assays such as the Lumipulse may aid in standardization when establishing CSF biomarker cut-points in individuals with and at risk for developing dementia due to AD.

When evaluating overall performance, the t-tau/Aβ42 ratio performed quite well across both the test and validation cohorts with respect to OPA (Cohort A/B: 94.6%, Cohort C: 86.1%, Cohort D: 85.0%). The relatively higher OPA of the t-tau/Aβ42 cut-point for Cohort A/B compared to Cohorts C and D may reflect the more advanced disease severity in Cohort A/B; specifically, 62% percent of individuals in Cohort A/B were PET+, whereas only 26% and 28% were PET+ in Cohorts C and D, respectively. Further analysis at a subject level for Cohorts C and D revealed ~12% (n=24/187 in Cohort C and n=27/233 in Cohort D) had discordance demonstrating a positive CSF and a negative PET readout (e.g. CSF+/PET−). Conversely, 3 out of 187 subjects (1.6%) and 8 out of 233 subjects (3.4%) in Cohorts C and D, respectively, exhibited a negative CSF result and positive PET readout. This discordance was only observed in ~4% of Cohort A/B. These findings provide additional support for the notion that amyloid plaques can be detected via low levels of CSF Aβ42 prior to being detectable by PET [31,39]. Such a scenario could contribute to the lower OPAs of ~ 85% in Cohorts C and D compared to Cohort A/B (~95%). Another possible cause for this difference in t-tau/Aβ42 cut-point performance may be cohort racial diversity. Cohorts A/B and C were primarily Caucasian (95 and 90%, respectively), similar to other well-characterized AD biomarker studies [40], while Cohort D was much more racially diverse, with representation from African-American, Hispanic, and Japanese sub-populations. This observation is interesting given that recent studies have suggested that analyses of AD molecular biomarkers may need to be adjusted for race [41]. Thus, the higher accuracy observed in Cohorts A/B and C could be a result of more homogeneous study populations.

Taken together, the present study demonstrates that CSF biomarkers measured on an automated platform have utility for establishing and validating biomarker cut-points in both the research cohort study and clinical trials. At present, the only gold standard to confirm AD pathology is post-mortem autopsy; thus, CSF biomarkers such as the t-tau/Aβ42 ratio would provide a more accessible, real-time alternative to evaluate the underlying pathophysiology in symptomatic cohorts (e.g. MCI, dementia). Further, CSF biomarkers may prove useful in identifying asymptomatic individuals with preclinical AD, affording these individuals the opportunity to enroll in AD clinical trials or make appropriate lifestyle changes. As the AD community moves closer to the reality of disease-modifying drugs, it will be vital for automated CSF biomarkers to gain FDA approval for use, so they can be paired with these therapeutics as they become available.

Research in Context.

Alzheimer disease (AD) is a multifactorial neurological disorder that has proven very difficult to diagnose, especially at early stages. Cerebrospinal fluid (CSF) biomarkers have shown promise in identifying individuals exhibiting brain pathology consistent with AD (e.g. amyloid plaques). In the present study, the authors established CSF biomarker (Aβ42, t-tau, and t-tau/Aβ42 ratio) cut-points for amyloid positivity in separate clinical trial and research study cohorts using the LUMIPULSE automated chemiluminescence enzyme immunoassay platform. A consistent t-tau/Aβ42 ratio cut-point best distinguished positive amyloid burden (determined by positron emission tomography [PET]) despite differences in cohort characteristics (e.g. clinical trial versus research cohorts, primarily asymptomatic versus symptomatic) and sample processing. Importantly, this identical cut-point was then validated in an independent cohort. As the field moves closer to the reality of AD disease-modifying drugs, it will be vital for CSF biomarkers to gain FDA approval to support therapeutic treatments as they become available.

Abbreviations:

Aβ42

amyloid-β peptide 1–42

AD

Alzheimer disease

APOE

apolipoprotein E

AUC

area under the ROC curve

CDR

Clinical Dementia Rating

CRO

Clinical Research Organization

CSF

cerebrospinal fluid

FDA

U.S. Food and Drug Administration

GCP

Good Clinical Practice

HRPO

Human Research Protection Office

ICH

International Conference on Harmonization

IEC

Independent Ethics Committee

IFC

International Federation of Clinical Chemistry

IRB

Institutional Review Board

LLoQ

lower level of quantitation

LP

lumbar puncture

MCI

mild cognitive impairment

NIA-AA

National Institute on Aging and Alzheimer’s Association

NPA

negative percent agreement

OPA

overall percent agreement

PET

positron emission tomography

PIB

Pittsburgh Compound B

PPA

positive percent agreement

QC

quality control

ROC

receiver operating characteristic

ROI

region of interest

SOP

standard operating procedure

SUVR

standardized uptake value ratio

t-tau

total tau

ULoQ

upper level of quantitation