Combining Lumipulse p‐tau217 and Aβ42/40 as confirmatory tests for Aβ positivity prior to disease‐modifying therapy

Abstract

INTRODUCTION

For a blood‐based biomarker to be considered a confirmatory test for the detection of abnormal amyloid beta (Aβ) levels, the sensitivity and specificity must be equivalent to that of current cerebrospinal fluid tests.

METHODS

In the current study we assessed the ability of phosphorylated tau (p‐tau)217 and Aβ42/40 from the Lumipulse G p‐tau217 and β‐amyloid ratio (1‐42/1‐40) tests, individually and combined, to predict Aβ positron emission tomography status in two sub‐cohorts from the Australian Imaging, Biomarkers, and Lifestyle Study of Ageing.

RESULTS

Testing an Alzheimer's disease continuum cohort, the area under the curve (AUC), sensitivity, specificity, and accuracy for the p‐tau217/Aβ42 ratio reached 0.961, 93%, 92%, and 93%, respectively. Validation in an intention‐to‐treat cohort demonstrated similar AUC (0.959), with increased sensitivity (99%), decreased specificity (87%), and increased accuracy (95%). Dual cut‐offs generating balanced 95% sensitivity/specificity result in 93% accuracy.

DISCUSSION

Combinations of plasma p‐tau217 and Aβ42 demonstrate recommended performance, confirming the presence of Aβ positivity prior to selection for disease‐modifying therapies.

Highlights

• The phosphorylated tau (p‐tau)217/amyloid beta (Aβ)42 ratio had high performance to detect Aβ positron emission tomography (PET) status, with > 90% sensitivity, specificity, and accuracy.

• p‐tau217/Aβ42 ratio dual cut‐offs set at 95% sensitivity and specificity found 10% to 15% of participants in the intermediate zone.

• Cut‐offs derived for the intention‐to‐treat cohort meet confirmatory assay criteria for a disease‐modifying therapy and can be used in clinical settings.

1. BACKGROUND

Now that disease‐modifying therapies (DMTs) have been approved for the treatment of Alzheimer's disease (AD), it is imperative that the process by which a patient becomes eligible for DMTs is swift and accurate. Current eligibility for a DMT is based around the recruitment parameters, which were set for either lecanemab or donanemab trials. ¹ , ² , ³ Thus, cut‐offs for either triage or confirmatory blood tests need to be developed on study populations that also meet these criteria. Compared to the development of cerebrospinal fluid (CSF) biomarkers to detect AD proteinopathy, first acknowledged by the National Institute on Aging–Alzheimer's Association working group in 2011 ⁴ prior to US Food and Drug Administration (FDA) approval in 2023, ⁵ the development of leading blood‐based biomarkers (BBMs; amyloid beta [Aβ]42/40 and phosphorylated tau [p‐tau] variants) and their ability to predict disease status has rapidly progressed. However, significant gaps in clinical and analytical validity, a lack of understanding of BBM test characteristics, and insufficient evidence in under‐represented ethnic populations make it difficult for real‐world implementations. ⁶ , ⁷ Research using mass spectrometry BBMs in primary and secondary care has demonstrated strong performance of certain BBMs to predict Aβ status, ⁸ yet only moderate performance to predict dementia. ⁹ Most recently, the Fujirebio p‐tau217/Aβ42 ratio gained much interest with its FDA approval for early detection of Aβ plaques associated with AD. ¹⁰ However, there is still a gap in the guidance for prescribing physicians on what cut‐off is required for a confirmation of abnormal level of Aβ for each specified BBM within appropriate populations.

A consensus statement ¹¹ was recently published to define acceptable minimum performance levels for a chosen BBM to predict abnormal Aβ levels. In this work, if the BBM could match the performance of core CSF biomarkers to predict Aβ positivity (≈ 90% sensitivity and specificity), then it could be used as a confirmatory test, or if sensitivity was ≥ 90% and specificity only reached ≥ 85% for primary care and 75% to 85% in secondary care, it could be applied as a triaging test for identifying individuals unlikely to have high levels of Aβ or require additional testing to confirm Aβ abnormality. The authors demonstrate that these statistics and related predictive values vary by patient group given the prevalence of Aβ positivity, and as such, cut‐offs need to be developed to maximize the performance of each BBM.

The derivation of a single cut‐off for a BBM to be used as a confirmatory test needs to correctly classify patients with at least 90% sensitivity and specificity from a population with ≈ 50% brain Aβ+ prevalence. ¹¹ To reach higher values of sensitivity or specificity, the dual cut‐off approach has been recently implemented, setting sensitivity and specificity at a desired level for optimal performance. A dual cut‐off approach has also been recommended, partially in response to 5% to 20% of individuals in various cohorts having borderline levels of Aβ near a cut‐off. ¹¹ The application of dual cut‐offs inevitably creates an intermediate zone, whereby the proportion of patients with values residing between the two cut‐offs can become increasingly large when the overall performance of the test is lacking. A recommendation of < 15% to 20% has been suggested; however, this can still be difficult to reach, with one study ¹² demonstrating AUC values of > 0.90, yet still having > 20% of patients in the intermediate zone.

In the current study, we assess the performance of two plasma BBMs to predict Aβ positron emission tomography (PET) positivity in two separate sub‐cohorts of the Australian Imaging, Biomarkers, and Lifestyle (AIBL) Study of Aging. The first sub‐cohort was designed to assess performance of each BBM along the AD continuum, with specific comparisons defined for pre‐clinical AD as well as the complete cohort. The second sub‐cohort was designed to mimic the specific patient population that would be seen in primary/secondary care and meet the inclusion criteria for a DMT. In this study we focus on the performance of Lumipulse plasma p‐tau217 and Aβ42/40 assays, define the performance of single and ratio BBMs and combinations of BBMs, and devise specific cut‐offs for use in routine clinical testing.

2. METHODS

2.1. Population sample

Samples from two separate sub‐populations of the AIBL were selected (Figure S1 in supporting information). The AD continuum cohort (ADCC) was selected specifically to assess plasma biomarker levels across the AD continuum, including AIBL participants who were cognitively unimpaired and Aβ– (CU Aβ–, N = 75; ≈ 50% subjective memory complainers), cognitively unimpaired and Aβ+ (CU Aβ+, N = 48; 66% subjective memory complainers), mild cognitive impairment and Aβ+ (MCI Aβ+, N = 26), and AD Aβ+ (N = 48). The second cohort was designed to fit an “intention‐to‐treat cohort” (ITTC), thus matching the pre‐requisites for DMT with either donanemab (prescribed as kisunla ¹³ ) or lecanemab (prescribed as leqembi, ¹⁴ Clinical Dementia Rating 0.5–1, Mini‐Mental State Examination [MMSE] 22–28). It should be noted that while the package insert for both DMTs indicates suitability for patients with MCI or mild AD as per the clinical trial in which results were reported, ¹ , ¹⁵ eligibility requirements for both leqembi and kinsula may differ slightly per country where a DMT has been approved. For example, eligibility in the United States, European Union (EU), and Australia differs by the number of apolipoprotein E (APOE) ε4 alleles a patient carries (United States: any number, Australia and EU: 0/1 copies), thus presenting specific sub‐populations within the MCI/mild AD groups eligible for treatment. Participants in the ITTC in the present study included 65 MCI Aβ+, 61 AD Aβ+, 59 MCI Aβ–, and 15 AD Aβ–. Clinical classification within AIBL was based solely on neuropsychological test battery scores and the collective decision of the clinical panel, including a psychiatrist, neurologist, geriatrician, and neuropsychologist; biomarker testing was not used to aid in clinical decision making. All participants were randomly selected from the Aβ PET subgroups as per Figure S1, with no participants in both ADCC and ITTC groups.

RESEARCH IN CONTEXT

1. Systematic review: For this work, we performed a comprehensive assessment of the current literature on the use of blood‐based biomarkers (BBMs) to detect amyloid beta (Aβ) neuropathology. Recent publications have demonstrated high performance of BBMs in population studies, yet few have defined specific thresholds and confirmatory test performance that have utility in the clinic for disease‐modifying therapy (DMT) eligibility. To address this specifically, the current research focused on two BBMs and their performance to predict Aβ neuropathology in two separate sets of participants: one a disease continuum setting and two an intention‐to‐treat population.

2. Interpretation: Our findings demonstrate the use of the linear combination of phosphorylated tau (p‐tau)217 along with the Aβ42/40 ratio or the p‐tau217/Aβ42 ratio provides acceptable performance to be used as confirmatory biomarkers required for Aβ neuropathology DMT screening. Cut‐off values presented for the intention‐to‐treat populations are applicable, representing values for clinical use.

3. Future directions: When BBM data post‐DMT become available, the question of whether these cut‐offs can be used at a point where sufficient brain Aβ has been removed and treatment can be stopped needs to be addressed.

2.2. Biospecimen collection

Blood was collected from participants between 7:30 and 10:30 a.m. (after overnight fasting) into K3‐ethylenediaminetetraacetic acid tubes (7.5 mL S‐monovette 01.1605.008), with pre‐added prostaglandin E1 (33 ng/mL of whole blood, Sapphire Biosciences) to prevent platelet activation, a potential source of peripheral Aβ. Samples were centrifuged at room temperature at 200 × g for 10 minutes to collect platelet‐rich plasma, then centrifuged at 800 × g for 10 minutes, before being aliquoted into 0.5 mL vials (2D cryobankIT, NUN374088), and then snap frozen within 2 hours prior to storage in vapor phase liquid nitrogen (LN2). Plasma samples selected for both ADCC and ITTC were originally collected between the years 2016 and 2024. None had been removed from the LN2 storage prior to selection and delivery to the laboratory at Monogram Biosciences on dry ice.

2.3. Plasma assay details

Aliquots of AIBL plasma samples were shipped on dry ice from Australia to Labcorp‐Monogram (South San Francisco, USA) and stored at −80°C until use. Samples were analyzed from a single thaw on a Fujirebio Lumipulse G1200 instrument in the Labcorp‐Monogram clinical reference lab. Lumipulse assays were analytically validated to College of American Pathologists–Clinical Laboratory Improvement Amendments (CAP‐CLIA) guidelines and displayed inter‐assay precision of < 20% coefficient of variation (CV) and control accuracy within 20% relative error (RE) to expected values. The ADCC cohort was analyzed using the Fujirebio Lumipulse plasma Aβ42 assay (lot #3303), Aβ40 assay (lot #3033), and p‐tau217 assay (lot #4049). The ITTC cohort was analyzed separately at a later time (Plasma Aβ1‐42 lot #T6B5081, Aβ1‐40 lot #T4B5112, and p‐tau217 lot #D4C5066). Variability was evaluated, and bridging of the second reagent lots used for the ITTC cohort to the original reagent lots was performed by data analysis of 40 samples using both sets of assay reagent lots. Each cohort of ≈ 200 samples each was analyzed by two assay runs of 100 samples each over 2 days on the same instrument, allowing consistent and minimal platform time to maintain Aβ42 and Aβ40 stability (< 1–2 hours). Inter‐assay high and low control reproducibility CV between the two runs of 100 samples each were 1.6% and 0.2%, p‐tau217; 2.1% and 2.2%, Aβ1‐42; 0.1% and 1.8%, Aβ40; and 2.1% and 0.3%, Aβ42/40 ratio.

2.4. Plasma assay bridging

Given the two separate cohorts were measured ≈ 12 months apart, we ran a bridging study such that any differences in assay results could be normalized between cohorts. To do this, we re‐ran 40 samples from the ADCC at the same time as running the samples from the ITTC. Samples were split up into 10 from each of CU Aβ–, CU Aβ+, 10 MCI Aβ+, and 10 AD Aβ+. Deming regression equations, concordance correlations, and Bland–Altman plots were calculated using the newly assayed ADCC sample results and the previously assayed ADCC sample results. Systematic differences in the values between the two cohort measurements were normalized by linear transformation of the ITTC data to the original ADCC values.

2.5. PET imaging

All PET imaging was conducted within 12 months of the blood draw. Imaging for Aβ PET included a 20 minute acquisition and 50 minute post‐injection of either 11C‐Pittsburgh compound B (PiB), 18F‐NAV4694 (NAV), 18F‐Flutemetamol (FLUTE), or 18F‐florbetapir (FBP) intravenously. Spatial normalization was performed using CapAIBL, ¹⁶ with resulting scan information standardized to the Centiloid (CL) scale. ¹⁷ Delegation of participants to the ADCC aligned participants with a CL of < 15 as Aβ PET negative and ≥ 25 CL as Aβ PET positive to allow for the direct comparisons of biomarkers across the disease continuum. In the ITTC group, given this was a cognitively impaired population, Aβ PET positivity was classified as having a CL of ≥ 25.

2.6. Statistical methodology

Clinical and demographic comparisons between clinical groups were assessed using independent samples t tests or Mann–Whitney U tests (where appropriate) for quantitative data and chi‐squared for categorical data. Comparison of the marginal means between binary groups was performed using generalized linear models (GLMs), with Cohen D used for effect size estimation. Receiver operating characteristic (ROC) analyses were used for both individual biomarkers and model‐based assessment of biomarker (log transformed) and covariate combinations to calculate area under the curve (AUC) with 95% confidence intervals (CIs), sensitivity and specificity, negative predictive value (NPV) and positive predictive value (PPV), and accuracy. Cut‐offs were determined using either the Youden index or, for dual cut‐off purposes, setting either sensitivity or specificity at desired levels. For dual cut‐off distribution figures (Results and Supplementary Materials in supporting information), p‐tau217 and its ratio with Aβ42 were logged and scaled to approximate a normal distribution for plotting. Biomarker cut‐off values, based upon 95% sensitivity and specificity, were transformed in the same way. The transformed biomarker data were then used within a GLM to predict the binary Aβ PET status (0/1) to derive the Aβ probability across the range of biomarker values. Values for the linear combination of p‐tau217, Aβ42/40, age, sex, APOE ε4, and the composite score were not log transformed or scaled as values were fitted values from the GLM. Histogram bars and density plot lines were scaled to fit the same y axis as the Aβ probability (0–1). Comparison of ROC models between biomarkers alone to predict Aβ PET and between biomarkers alone and models including biomarkers and covariates was performed using the DeLong method. ¹⁸ p values were left uncorrected for multiple comparisons; however, where appropriate, p values that were greater than a Bonferroni‐adjusted alpha value (α = 0.002) but < 0.05 were discussed as being nominally significant. Multivariate selection was performed using the least absolute separation and shrinkage operator (LASSO) using all possible combinations of biomarkers (including ratios) and covariates. All analyses were performed using the R statistical environment (version 4.4.1). ¹⁹

3. RESULTS

3.1. Study sample characteristics

Within the ADCC, there were no significant differences in age or sex between Aβ PET groups (p > 0.05), while there were significant differences in all other clinical parameters tested (p < 0.0001, Table 1A). Aβ PET+ of the ADCC was 62%. In the CU group, 39% had preclinical AD (Aβ PET+), while 61% were Aβ PET–. In the ITTC, all participants were either MCI (62.5%) or had mild AD (37.5%). Of the MCI group, 52% were Aβ PET+, while in the AD group 80% were Aβ PET+, totaling 63% Aβ PET+. There were no significant differences in age and sex, while there were significant differences in other clinical parameters measured (Table 1B).

TABLE 1A.

Sample clinical and demographic characteristics for the ADCC.

Characteristic	Total sample	Aβ PET–	Aβ PET+	p value
N (%)	197	75 (38)	122 (62)
Sex male, N (%)	96 (49%)	37 (49)	59 (48)	0.89
Mean age, years (SD)	75.1 (7.3)	74.2 (6.1)	75.6 (7.9)	0.15
Mean Centiloid (SD)	56.7 (53)	−0.5 (4.6)	91.9 (35.4)	<0.0001
APOE ε4 carriage, N (%)	82 (42)	10 (13)	72 (59)	<0.0001
Median MMSE (MAD)	28 (3)	29 (1.5)	26 (4.4)	<0.0001
Median CDR‐SB score (MAD)	0 (0)	0 (0)	2 (3)	<0.0001
Mean PACC score (SD)	−0.7 (1.5)	0.4 (0.6)	−1.4 (1.5)	<0.0001
Diagnosis, CU N (%)	123 (62)	75 (61)	48 (39)
Diagnosis, MCI N (%)	26 (13)	0 (0)	26 (100)
Diagnosis, AD N (%)	48 (24)	0 (0)	48 (100)	<0.0001

Note: Cut‐off for Aβ PET+, ≥ 25 Centiloid.

Abbreviations: Aβ, amyloid beta; AD, Alzheimer's disease; ADCC, Alzheimer's disease continuum cohort; APOE, apolipoprotein E; CDR‐SB, Clinical Dementia Rating Sum of Boxes; CU, cognitively unimpaired; MAD, mean absolute deviation; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination; N, number of samples; PACC, Pre‐Clinical Alzheimer's Cognitive Composite; PET, positron emission tomography; SD, standard deviation.

TABLE 1B.

Sample clinical and demographic characteristics for the ITT cohort.

Characteristic	Total sample	Aβ PET–	Aβ PET+	p value
N (%)	200	74 (37)	126 (63)
Sex male, N (%)	112 (56)	45 (61)	67 (53)	0.29
Mean age, years (SD)	73.6 (8.4)	72.2 (9.3)	74.3 (7.8)	0.099
Mean Centiloid (SD)	64 (56)	−0.6 (5.8)	101.9 (32.5)	<0.0001
APOE ε4 carriage, N (%)	87 (44)	10 (14)	77 (61)	<0.0001
Median MMSE (MAD)	25.5 (3.7)	27 (3)	25 (3)	<0.0001
Median CDR‐SB score (MAD)	2 (2.2)	1 (0.7)	3 (2.2)	<0.0001
Mean PACC score (SD)	−2 (1.1)	−1.3 (1.1)	−2.5 (0.8)	<0.0001
Diagnosis MCI, N (%)	124 (62)	59 (48)	65 (52)
Diagnosis AD, N (%)	76 (38)	15 (20)	61 (80)	<0.0001

Note: Cut‐off for Aβ PET+, ≥ 25 Centiloid.

Abbreviations: Aβ, amyloid beta; AD, Alzheimer's disease; APOE, apolipoprotein E; CDR‐SB, Clinical Dementia Rating Sum of Boxes; CU, cognitively unimpaired; ITTC, intention‐to‐treat cohort; MAD, mean absolute deviation; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination; N, number of samples; PACC, Pre‐Clinical Alzheimer's Cognitive Composite; PET, positron emission tomography; SD, standard deviation.

3.2. Comparison and bridging of mean assay levels between ADCC and ITTC

Results of the bridging study are shown in Figure S2 in supporting information. Bland–Altman assessment plots saw the magnitude of bias range from an ≈ 0.7% decrease for Aβ40, to an ≈ 7% increase in Aβ42 and p‐tau217 from the ADCC to the ITTC assessments. Variation seen for each assay was consistent between ADCC and ITTC for each biomarker, with slight reductions in the standard deviation (SD) for p‐tau217 (ADCC: 0.43, ITTC: 0.39) and Aβ42 (ADCC: 5.48, ITTC: 5.12), and a slight increase in the SD for Aβ40 (ADCC: 56.49, ITTC: 60.28). The SD of the difference was smaller than 1 SD of the ADCC for each biomarker (Aβ40: 0.26 SD, Aβ42: 0.19 SD, and p‐tau217: 0.12 SD). Concordance correlations for all three assays (Aβ40, Aβ42, and p‐tau217) ranged between 0.93 and 0.99. Given the variance comparisons demonstrated a consistency across each assay, all plasma data from the ITTC were transformed using the Deming regression equations shown in Figure S2, such that all data from the ITTC are on the same scale as the ADCC for comparison.

3.3. Comparison of mean assay levels between Aβ PET groups

All BBMs in both ADCC and ITTC were significantly altered in Aβ PET+ groups (p < 0.001, Tables 2A and 2B, Figure 1). Cohen D values show significantly larger effect sizes for the Aβ42/40 ratio compared to Aβ42 alone (ADCC: p = 0.0001, ITTC: p < 0.0001); however, this was not the case for the p‐tau217/Aβ42 ratio relative to p‐tau217 (ADCC: p = 0.318, ITTC: p = 0.223). Box and whisker plots demonstrate very similar Aβ42, Aβ42/40, p‐tau217, and p‐tau217/Aβ42 levels and medians within the same diagnostic groups shared between the ADCC and ITTC (Figure 1), (Table 2B).

TABLE 2A.

Comparison of biomarker means between PET amyloid groups in the ADCC.

Biomarker	Mean (SD) PET– Aβ–	Mean (SD) PET– Aβ+	Cohen D	p value a	p value b
Aβ42	27.969 (5.29)	23.484 (4.637)	0.928	3.40E‐09	3.15E‐07
p‐tau217	0.154 (0.206)	0.596 (0.399)	2.17	1.21E‐32	2.53E‐09
Aβ42/40	0.092 (0.011)	0.076 (0.006)	1.81	2.61E‐21	9.11E‐11
p‐tau217/Aβ42	0.006 (0.009)	0.026 (0.019)	2.43	1.25E‐37	1.67E‐09

Note: Cohen D calculated on log transformed data. Cut‐off for Aβ PET+, ≥ 25 Centiloid.

Abbreviation: Aβ, amyloid beta; ADCC, Alzheimer's disease continuum cohort; APOE, apolipoprotein E; PET, positron emission tomography; p‐tau, phosphorylated tau; SD, standard deviation.

^a p value from comparison of means (log transformed biomarker data) without adjustment for age, sex, APOE ε4 allele status, and PET tracer.

^b p value from comparison of means (log transformed biomarker data) with adjustment for age, sex, APOE ε4 allele status, and PET tracer.

TABLE 2B.

Comparison of biomarker means between PET amyloid groups in the ITTC.

Biomarker	Mean (SD) PET– Aβ–	Mean (SD) PET– Aβ+	Cohen D	p value 1	p value 2
Aβ42	28.212 (5.569)	22.853 (3.926)	1.14	3.99E‐12	2.99E‐08
p‐tau217	0.168 (0.175)	0.663 (0.422)	2.36	8.81E‐34	5.47E‐11
Aβ42/40	0.09 (0.008)	0.074 (0.005)	2.41	2.43E‐32	2.21E‐10
p‐tau217/Aβ42	0.006 (0.009)	0.03 (0.02)	2.69	2.13E‐40	1.03E‐10

Note: Cohen D calculated on log transformed data. Cut‐off for Aβ PET+, ≥25 Centiloid.

Abbreviation: Aβ, amyloid‐beta; APOE, apolipoprotein E; ITTC, intention‐to‐treat cohort; PET, positron emission tomography; p‐tau, phosphorylated tau; SD, standard deviation.

¹ p value from comparison of means (log transformed biomarker data) without adjustment for age, sex, APOE ε4 allele status and PET tracer.

² p from comparison of means (log transformed biomarker data) with adjustment for age, sex, APOE ε4 allele status and PET tracer.

FIGURE 1.

Comparison of biomarker medians between Aβ PET groups within clinical classification. A–D, ADCC group. E–H, ITTC group. Cut‐off for Aβ PET+: ≥ 25 Centiloid; Aβ, amyloid beta; AD, Alzheimer's disease; ADDC, Alzheimer's disease continuum cohort; CU, cognitively unimpaired; ITTC, intention‐to‐treat cohort; MCI, mild cognitive impairment; NEG, positron emission tomography amyloid beta negative; PET, positron emission tomography; POS, positron emission tomography amyloid beta positive; p‐tau, phosphorylated tau.

3.4. Correlation between BBM and CL

Correlation between Aβ42/40 with CL showed a moderate inverse relationship (ADCC: Rho = −0.507, ITTC: Rho = −0.602, Figure 2B,F), with variance in the ratio reduced compared to Aβ42 alone. Agreement between Aβ42/40 and Aβ PET–/+ groups was strong, with only 4% and 2% false negative in the ADCC and ITTC, respectively, and 8% and 6% false positive values in the ADCC and ITTC, respectively. For p‐tau217 and its ratio with Aβ42, correlations were strong, with Rho values of 0.732 and 0.751, respectively, for the ADCC, and 0.723 and 0.753, respectively, for the ITTC. Using the p‐tau217/Aβ42 ratio compared to p‐tau217 alone reduced the proportion of false positives from 7% to 4% in the ADCC but limited change in the ITTC (4% to 5%), whereas false negatives decreased from 4% to 0% in the ITTC.

FIGURE 2.

Correlation between Aβ42/40, p‐tau217, and the p‐tau217/Aβ42 ratio and CL. Panels (A–D) ADCC; Panels (E–H) ITTC. Dashed vertical line representing CL cut‐off at 25. Dashed horizontal lines represent the Youden's Index from a ROC for prediction of PET‐Aβ (< 25 PET‐A‐, 25 and over PET‐Aβ+) per biomarker. Gray lines represent the simple regression line between CL and biomarker. Percentages represent the proportion of dots within each quadrant as separated by the dashed lines. Aβ, amyloid beta; AD, Alzheimer's disease; ADDC, Alzheimer's disease continuum cohort; CL, Centiloid; CU, cognitively unimpaired; ITTC, intention‐to‐treat cohort; MCI, mild cognitive impairment; p‐tau, phosphorylated tau; ROC, receiver operating characteristic.

3.5. Assay prediction in all participants and CU participants

In the ADCC, individual biomarker AUC values to predict Aβ PET were strongest for p‐tau217 and the p‐tau217/Aβ42 ratio (0.941 and 0.961, respectively, Figure 3A). For the ADCC, the AUC values for Aβ42 and the Aβ42/40 ratio were significantly lower than p‐tau217 (p = 0.0002 and p < 0.0001). In the ITTC (Figure 3B), again the AUC value for Aβ42 was significantly lower than for p‐tau217 (p = 0.0002); however, the AUC value for the Aβ42/40 ratio was not different from p‐tau217 (p = 0.843). In the ADCC, the p‐tau217/Aβ42 ratio, the linear combination of p‐tau217, Aβ42/40, age, sex, and APOE ε4, and the composite score (linear combination of age, sex, APOE ε4, Aβ42 + Aβ42/40 + p‐tau217) were significantly better (nominally, unadjusted) than using p‐tau217 alone to predict Aβ PET status (p = 0.011, 0.015, and 0.012, Table S1 in supporting information). Using data from the ITTC demonstrated similar results, with the composite score (linear combination of APOE ε4 + age + Aβ42 + p‐tau217 + Aβ42/40 + p‐tau217/Aβ42), the linear combination of Aβ42/40 with p‐tau217, age, sex, and APOE ε4, and the p‐tau217/Aβ42 ratio were all significantly better (nominally, unadjusted) at predicting Aβ PET status compared to p‐tau217 alone (p = 0.013, 0.010, and 0.009, respectively, Table S1). Adding age, sex, and APOE ε4 to biomarkers within either ADCC or ITTC to predict Aβ PET increased the AUC values for Aβ42 and the Aβ42/40 ratio only (Table S1).

FIGURE 3.

Receiver operating characteristic plots for individual biomarkers, their ratios, and a composite score in ADCC (A) and ITTC (B). ADCC composite, linear model including APOE ε4 + age + sex + Aβ42 + Aβ42/40 + p‐tau217. Base model, linear model including age, sex, and APOE ε4 allele status. ITTC composite, linear model including age, sex, APOE ε4 allele status, Aβ42 + Aβ42/40 + p‐tau217 + p‐tau217/Aβ42. Aβ, amyloid beta; ADCC, Alzheimer's disease continuum cohort; APOE, apolipoprotein E; AUC, area under the curve; CI, confidence interval; ITTC, intention‐to‐treat cohort; p‐tau, phosphorylated tau.

In the pre‐clinical AD subgroup, the Aβ42/40 ratio had a higher AUC to predict Aβ PET compared to the complete ADCC (pre‐clinical AUC: 0.906, complete ADCC group AUC: 0.893, Table S1 Figure S3 in supporting information), albeit this was not significant (p > 0.05). Adding age, sex, and APOE ε4 did not increase AUC values for any biomarker or combination thereof (p > 0.05). The p‐tau217/Aβ42 ratio, the linear combination of p‐tau217 and the Aβ42/40 ratio, and the composite score (linear combination of APOE ε4 + age + sex + Aβ42 + Aβ42/40 + p‐tau217 + p‐tau217/Aβ42) had higher AUC values than p‐tau217 alone (AUC = 0.943, 0.954, and 0.957, respectively; Table S1). However, upon statistical comparison, only the composite score and the linear combination AUC values were significantly greater (p = 0.020 and 0.031, respectively).

3.6. BBM sensitivity, specificity, and accuracy to predict Aβ PET status using a single cut‐off

Using data from the ADCC, the composite score, the linear combination of p‐tau217 and the Aβ42/40 ratio, age, sex, and APOE ε4, and the p‐tau217/Aβ42 ratio had the highest accuracy to predict Aβ PET status (≈ 93%–95%; Aβ+ prevalence 62%), which was higher than using p‐tau217 alone (≈ 90%, Table S2A in supporting information). Restricting to cognitively impaired only in the ITTC, the composite score had the highest accuracy (≈ 97%), followed closely by the p‐tau217/Aβ42 ratio and the linear combination of p‐tau217 with Aβ42/40, age, sex, and APOE ε4 (≈ 94.5% and ≈ 96.5%, Aβ+ prevalence 63%; Table S2B); all of which were significantly greater than p‐tau217 alone (91.5%). Sensitivity for the single p‐tau217 for the complete ADCC (95%, Table S2A) was higher compared to the same in the ITTC (93%, Table S2B), while specificity was increased in the ITTC (89%, Table S2B) compared to the ADCC (83%, Table S2A). Using the p‐tau217/Aβ42 ratio evened out the sensitivity‐to‐specificity ratio for the ADCC; sensitivity and specificity were 93% and 92%, respectively (Table S2A), but not for the ITTC; sensitivity and specificity were 99% and 86% respectively (Table S2B). Adding age, sex, and APOE ε4 increased specificity for both the composite score and the linear combination for the ITTC cohort (p‐tau217/Aβ42 ratio: specificity 86%, composite: specificity 95%, linear combination: specificity 95%, Table S2B). However, this was not the case across the whole ADCC, with specificity similar across the three markers (p‐tau217/Aβ42 ratio: 92%, composite: specificity 93%, linear combination: specificity 92%, Table S2A). In the pre‐clinical CU subgroup, the p‐tau217/Aβ42 ratio also had higher specificity (84%) and accuracy (89%; Aβ+ prevalence = 39%) than p‐tau217 alone (specificity 81%; accuracy 85%). The ratio was similar to the linear combination and composite score (accuracy = 89% and 92%, respectively; Table S2C).

3.7. Cut‐off values between the ADCC and the ITTC

Cut‐offs derived from the ITTC (Youden index, Table S3 in supporting information) were slightly different than those derived using the complete ADCC for each of the different BBMs, even after the linear transformation was used on the ITTC dataset. The main difference is an increased optimal p‐tau217 Youden cut‐off derived from the ITTC (0.253) relative to the ADCC (0.179). Using the cut‐offs derived from the ADCC on the data from the ITTC, we found the AUC value for the Aβ42/40 ratio was similar to that from the ADCC, while all other AUC values were decreased compared to derived AUCs from both ADCC and ITTC. The AUC performance, however, remained relatively high in the ITTC, ranging from 0.88 to 0.92 for the Aβ42/40 ratio, p‐tau217 and the p‐tau217/Aβ42 ratio.

3.8. Plasma p‐tau217, Aβ42/40, and the p‐tau217/Aβ42 ratio dual cut‐off model

Cut‐offs derived from p‐tau217 alone and in combination with Aβ42 or Aβ42/40 resulted in variable accuracies to detect Aβ PET status in the ADCC and ITTCs (90%–97%). The application of dual cut‐offs setting ≥ 90% assay sensitivity and specificity may mitigate variable single cut‐offs and further improve consistency and clinical performance near the cut‐off levels, so they were evaluated for single and combination biomarkers.

Sensitivity, specificity, PPV, NPV, and accuracies for the dual cut‐off results presented in this section are calculated from the complete data set, including those participants found to be in intermediate zones. Removal of participants who reside within intermediate zones, either from 90%, 92.5%, or 95% cut‐off values for sensitivity and specificity, and then recalculating sensitivity, specificity, PPV, NPV, and accuracies resulted in overestimated values, and as such, results are not presented here.

Table S2A,B shows the dual cut‐off results across sensitivities and specificities set at 90%, 92.5%, and 95% for each individual biomarker, biomarker ratio, and biomarker combination for the ADCC and ITTC. For Aβ42 and the Aβ42/40 ratio, setting the sensitivity and specificity at 90% produced the smallest proportion of participants in the intermediate zone (ADCC: Aβ42 ≈ 54%, Aβ42/40 24%; ITTC: Aβ42 ≈ 41%, Aβ42/40 10%). For p‐tau217 alone, intermediate zones for 90%, 92.5%, and 95% sensitivity and specificity contained 8%, 11%, and 17% of participants, respectively, from the ADCC, and 6%, 12%, and 23% of participants, respectively, from the ITTC. Using the p‐tau217/Aβ42 ratio in the ADCC, the sensitivity and specificity at Youden index were 93.4% and 92%, respectively; thus, all participants were classified as either negative or positive using 90% and 92.5% sensitivity and specificity, while the intermediate zone for 95% sensitivity and specificity contained only 9% of participants. Similar results were seen for the ITTC, with only 2% and 13% of participants in the intermediate zones using 92.5% and 95% cut‐offs for sensitivity and specificity. Using either the composite score or the linear combination of p‐tau217 and the Aβ42/40 ratio at the 95% sensitivity and specificity cut‐offs found only 10% of participants in the intermediate zone for the ADCC, and no intermediate zone participants for the ITTC (sensitivity and specificity at Youden index > 95%).

Figures 4 and 5 and S4 and S5 in supporting information demonstrate the low proportion of participants in the intermediate zone using the 95% levels for sensitivity and specificity for p‐tau217 (A and B) and the p‐tau217/Aβ42 ratio (C and D; Figure 4: ADCC, Figure 5: ITTC), and for the composite score (A and B) and the linear combination of p‐tau217, Aβ42/40, age, sex, and APOE ε4 (C and D; Figure S4: ADCC, Figure S5: ITTC). Cut‐off values aligned with sensitivity and specificity for p‐tau217 and its ratio with Aβ42 demonstrate low numbers of participants in the intermediate zone, with similar probabilities of being Aβ+ for both markers. Of interest, when assessing the same plots for the composite score and the linear combination of p‐tau217, Aβ42/40, age, sex, and APOE ε4 saw the participant predicted values pushed toward either 0 or 1. In the ADCC, assessment of the 14 participants in the intermediate zone saw all but one were APOE ε4 negative, with varying values around the p‐tau217 and Aβ42/40 thresholds (Figure S6 in supporting information). In the ITTC, at the chosen 95% sensitivity and specificity, there were no participants in the intermediate zone, as the sensitivity and specificity using Youden index were > 95%.

FIGURE 4.

ADCC: Density plot for p‐tau217 (A) and the p‐tau217/Aβ42 ratio (B), including lower and upper cut‐off values at the corresponding 95% sensitivity and specificity. Blue solid line corresponds to the density of values below the cut‐off. Red solid line corresponds to the density of values greater than the cut‐off. Blue dashed line (vertical, A, B, C, D) indicates lower cut‐off at 95% specificity and (horizontal) alignment at the lower cut‐off to the probability of being Aβ+ at the corresponding biomarker value. Red dashed line (vertical, A, B, C, D) indicates lower cut‐off at 95% sensitivity and (horizontal) alignment at the upper cut‐off to the probability of being Aβ+ from the corresponding biomarker value; (B and C) show sensitivity and specificity for the complete range of possible values of p‐tau217 and the p‐tau217/Aβ42 ratio. Blue dots/solid lines represent participant values less than the lower cut off. Red dots/solid lines represent participant values greater than the upper cut off. Cut‐off values (lower‐blue, upper‐red) are set using the 95% sensitivity and specificities. Aβ, amyloid beta; ADCC, Alzheimer's disease continuum cohort; p‐tau, phosphorylated tau.

FIGURE 5.

ITTC: Density plot for p‐tau217 (A) and the p‐tau217/Aβ42 ratio (B), including lower and upper cut‐off values at the corresponding 95% sensitivity and specificity. Blue solid line corresponds to the density of values below the cut‐off. Red solid line corresponds to the density of values greater than the cut‐off. Blue dashed line (vertical, A, B, C, D) indicates lower cut‐off at 95% specificity and (horizontal) alignment at the lower cut‐off to the probability of being Aβ+ at the corresponding biomarker value. Red dashed line (vertical, A, B, C, D) indicates lower cut‐off at 95% sensitivity and (horizontal) alignment at the upper cut‐off to the probability of being Aβ+ from the corresponding biomarker value; (B and C) show sensitivity and specificity for the complete range of possible values of p‐tau217 and the p‐tau217/Aβ42 ratio. Blue dots/solid lines represent participant values less than the lower cut off. Red dots/solid lines represent participant values greater than the upper cut off. Cut‐off values (lower‐blue, upper‐red) are set using the 95% sensitivity and specificities. Aβ, amyloid beta; ITTC, intention‐to‐treat cohort; p‐tau, phosphorylated tau.

Application of the 90%, 92.5%, and 95% dual cut‐offs for the most part did not provide large increases in PPV, NPV, and accuracy relative to single cut‐offs for the same biomarker compared across each cohort. There was a modest increase in PPV for p‐tau217 alone in the ADCC when dual cutoffs were applied (PPV = 93%–96%) relative to the single cutoff (90%), and smaller increases in PPV for the p‐tau217/Aβ42 ratio and combination models in both ADCC and ITTC, whereas accuracy ranged consistently from 93% to 95% (Aβ+ prevalence 62% and 63%, respectively; Table S2A,B). Accuracies for single cut‐offs applied to Aβ42/40 ratio, p‐tau217 alone, and combinations in the ADCC (88%–95%), or for relatively higher single cut‐offs for the combinations in the ITTC (92%–98%), either slightly decreased somewhat or remained the same when 90%, 92.5%, and 95% dual cut‐offs were applied (88%–95%).

Figure S7 and Table S2C in supporting information demonstrate the same analyses performed on the pre‐clinical AD subset from the ADCC. Aβ42/40 here performed better than in the complete ADCC, with only ≈ 17% of participants in the intermediate zone using the 90% sensitivity and specificity cut‐offs. This was similar to p‐tau217 at 90%, with ≈ 16% of participants in the intermediate zone at the same sensitivity and specificity cut‐offs. Using the p‐tau217/Aβ42 ratio or the composite score with 95% sensitivity and specificity cut‐offs also showed ≈ 16% in the intermediate zone, while the linear combination of p‐tau217, Aβ42/40, age, sex, and APOE ε4 allele status was slightly higher at ≈ 18% at the same sensitivity and specificity cut‐offs. Similar to the ADCC, in the CU subgroup PPV increased in p‐tau217 alone for all three dual cut‐off levels (PPV = 81%, 83%, and 87%, respectively) relative to the single cut‐off (76%). This was observed with all the biomarkers in the CU subgroup, achieving 86% to 89% PPV in the combination markers while accuracy remained nearly constant (≈89%–94%; Aβ+ prevalence = 39%; Table S2C). Relatively high NPVs for single cut‐offs applied to Aβ42/40 ratio, p‐tau217 alone, and combinations (94%–98%) either slightly decreased somewhat or remained the same when 90%, 92.5%, and 95% dual cut‐offs were applied (NPV = 93%–97%).

4. DISCUSSION

The accuracy of BBMs to detect abnormal Aβ levels has decidedly improved with the introduction of high‐performance immunoprecipitation‐mass spectrometry (IPMS) and ultra‐sensitive high‐throughput assays run on Quanterix, Lumipulse, and Elecsys machines, providing reproducible results across many cohorts. In the current research, we test specifically Lumipulse assays for Aβ42/40 and p‐tau217, and their combination either in a ratio, a linear model, or a composite score to predict Aβ PET positivity in two separate sample groups: (A) an ADCC, which also included a pre‐clinical comparison, and (B) an ITTC, designed specifically to match the patients who would be eligible for treatment upon presentation to the clinic.

Results from the pre‐clinical subgroup analysis of the ADCC demonstrate the utility of the Aβ42/40 ratio to predict Aβ PET status, having a similar AUC to p‐tau217, both at 0.906, albeit this was lower than the p‐tau217/Aβ42 ratio (AUC: 0.943); the linear combination of p‐tau217, Aβ42/40, age, sex, and APOE ε4 (AUC: 0.954); and the composite score (AUC: 0.945). Sensitivity and NPV values were much higher than specificity and PPV value for all markers in pre‐clinical AD, and maximum accuracy reached ≈ 93.5% for the linear combination of p‐tau217, Aβ42/42, age, sex, and APOE ε4, and 89.4% for the p‐tau217/Aβ42 ratio. With Aβ+ prevalence at 39% in this subgroup, AUC values seen here are slightly lower than those seen from the Wisconsin Registry for Alzheimer's Prevention and Alzheimer's Disease Neuroimaging Initiative (IPMS Aβ42/40 only) cohorts with ≈ 36% Aβ+ prevalence, ²⁰ , ²¹ and higher compared to that seen from the Alzheimer's and Families and Biomarkers for Identifying Neurodegenerative Disorders Early and Reliably cohorts with ≈ 27% Aβ+ prevalence. ²² , ²³

Our findings for the complete ADCC and the ITTC were remarkably similar in terms of AUC values, with the main difference being the increase in performance for the Aβ42/40 ratio in the ITTC. Of interest was the marked increase in the specificity for p‐tau217 alone in the ITTC (89%) compared to the ADCC (83%), with a small decrease in sensitivity in the ITTC (93%) compared to the ADCC (95%). Furthermore, while the AUC for the p‐tau217/Aβ42 ratio was almost the same in both cohorts, the sensitivity and specificity were quite different, being more even in the ADCC and more heavily weighted toward sensitivity in the ITTC. There are several reasons why this could be the case. First, the performance of Aβ42/40 to predict Aβ PET status was greater in the ITTC, adding to the possibility of detecting more “true positive” participants through the contribution of Aβ42 to the ratio. Second, given the ITTC was primarily designed as a cognitive impairment cohort (higher cognitive burden than that of the ADCC), it is possible that, even though the Aβ+ prevalence was the same, the p‐tau217 is reflecting a higher overall tau load among the Aβ+ participants. This is supported by others, with p‐tau217 being highly associated with both Aβ+ and tau load. ²⁴ , ²⁵

As a research tool, results from the ITTC show the combination of BBMs with age, sex, and APOE ε4 appears to provide the most robust estimates for accuracy, PPV, and NPV with Aβ PET status (Table S2B). These results are consistent with the clinical performance resulting from the combination of p‐tau217 and Aβ42/40 measurements by mass spectrometry and Lumipulse immunoassay, with or without adjustment for age, sex, and APOE ε4. ²⁶ Results across large and well‐characterized studies ¹² , ²⁷ , ²⁸ , ²⁹ have also demonstrated high AUC values and comparable accuracy for the p‐tau217/Aβ42 ratio; however, it remains inconclusive as to whether this ratio is better than using p‐tau217 alone. Results from Figdore et al. ¹² saw similar AUC values for p‐tau217; however, unlike the current study in which we saw nominal significance (p ≅ 0.01), the p‐tau217/Aβ42 ratio was not significantly better than p‐tau217. This disparity may be due to the large difference in sensitivity between results from Figdore et al. ¹² and the ITTC (92% vs. 99%, respectively). Collectively, results across these studies suggest that while keeping high accuracy, the p‐tau217/Aβ42 ratio leans toward higher sensitivity in a cognitively impaired cohort (ITTC, Figdore et al. ¹² ), and a more evenly balanced sensitivity and specificity in cohorts containing both CU and individuals with cognitive impairment (ADCC, Cousins et al. ²⁷ ).

In terms of utility of either individual BBMs, ratios of BBMs, linear combinations of BBMs, and composite scores to predict brain Aβ load with high accuracy, it was clear from the ADCC that using a biomarker‐only approach (no covariate model combinations), the p‐tau217/Aβ42 ratio had the lowest proportion of participants in the intermediate zone (≈ 9%) using dual cut‐offs (95%), similar to that seen in the recent study by Palmqvist et al. ²⁹ Using the composite score was not significantly better than the p‐tau217/Aβ42 ratio alone and resulted in a reduction of the intermediate zone by only 2%. When repeating this assessment in the ITTC, it was clear that using either the composite score or the linear combination of p‐tau217 and Aβ42/40 along with age, sex, and APOE ε4, had better performance, with no intermediate zones using the conservative 95% sensitivity and specificity cut‐offs (Figure S5), while the proportion in the intermediate zone for the p‐tau217/Aβ42 ratio was slightly higher in the ITTC compared to the ADCC (≈ 13%).

Determination of a BBM cut‐off to detect Aβ positivity, which can be used across different studies and populations, has been difficult. In our ADCC, the cut‐off for p‐tau217 was 0.1795, while in the ITTC, it was 0.2534. Using the same Lumipulse assay to predict Aβ PET positivity, other studies have published cut‐off values including 0.158 pg/mL, ²⁶ 0.18 pg/mL, ²⁴ 0.186 pg/mL, ³⁰ and 0.229 pg/mL. ¹² It is likely that both lab‐to‐lab variation in performing the assays and the study population design (CU/MCI/AD/Aβ+/Aβ–) all play a role in determination of the cut‐off for each study. With a single cut‐off most similar to the current work, all participants of the Mayo Clinic study were cognitively impaired, ⁴ while all other studies were a mixture across the entire AD continuum. Thus, the p‐tau217 cut‐off of 0.1795 from the ADCC is similar to that of the three studies mentioned here, ¹² , ²⁴ , ³⁰ with the lower cut‐off in the ADCC likely due to less cognitive impairment and a lower tau burden in this sub‐cohort.

The question remains, though, as to the stringency of the dual cut‐off approach, which may ultimately lie in the hands of the clinician. Using a single cut‐off approach and the linear combination of Aβ42/40 and p‐tau217, along with age, sex, and APOE ε4, can provide very robust estimates of Aβ PET status, at ≈ 97% accuracy and no intermediate zone (ITTC). Alternatively, application of a conservative dual cut‐off approach (95% sensitivity and specificity) to the p‐tau217/Aβ42 ratio provides comparable performance (≈ 93% accuracy), with an increase to 13% of participants in an intermediate zone, but may represent less complexity with fewer measurements and less analysis. Dual cut‐offs could be helpful to clinicians toward making a diagnosis or treatment decision when reviewing all clinical parameters of a patient.

Recently the Lumipulse p‐tau217/Aβ42 ratio test has also been approved by the FDA as the first in vitro diagnostic test for aiding in the diagnosis of AD, representing access to a widely available, scalable, and high‐performing clinical test. Results seen from the present work demonstrate strong performance of a BBM (or combination thereof) to detect a positive Aβ PET scan; however, clinical utility of such a blood test is appropriate while understanding that (1) although Aβ PET has been ranked highly as an appropriate use criteria for aiding in the diagnosis of AD, ³¹ it is not perfect and is open to measurement error, and (2) comparison of BBM results in the United States are specific to an FDA‐approved visual read test, which does not perfectly correlate to a positive Aβ PET scan based on CLs. ³²

The current study, like others, has its limitations. First, the ADCC and ITTC were run ≈ 12 months apart. While the bridging study results showed on average the p‐tau217 results were lower in the newly assayed samples (Figure S2), upon assessment of mean values for each cohort, both pre‐ and post‐transformation mean values were higher in the ITTC than in the ADCC. This in part might be due to a larger number of cognitively impaired individuals in the ITTC, similar to the cohort of Figdore et al. ¹² In the current study, there were a number of false positive results (four, Aβ PET–/p‐tau217+), which upon further investigation could not be explained by chronic kidney disease, medications, or other non–AD‐related illness. Moreover, the present cohort, while well characterized, is also relatively homogenous from Australia; further studies will be required to evaluate the performance and cut‐offs of these BBMs and their appropriate use as confirmatory tests for Aβ positivity in both more racially diverse and international populations.

In summary, the current work has demonstrated the use of the Lumipulse p‐tau217 and Aβ42/40 assays to have high accuracy to predict Aβ PET status in two sub‐cohorts of the AIBL study, across distinct disease stages and with expected Aβ+ prevalence. Results demonstrated here are similar to those previously seen with the notable addition that the p‐tau217/Aβ42 ratio; the linear combination of p‐tau217, Aβ42/40 along with age, sex, and APOE ε4; and the composite score all perform significantly better than p‐tau217 alone to predict Aβ PET status (albeit at a nominally significant, uncorrected level) and provide higher accuracies at a range of different sensitivity/specificity cut‐offs. Furthermore, the analysis of a pre‐clinical subgroup, using the combination of assays, also resulted in an acceptable 16% to 18% of participants in an intermediate zone set by 95% sensitivity and specificity and was found to be ≈ 89% to 94% accurate with lower Aβ+ prevalence. Collectively, our findings demonstrate that BBMs can work as well as CSF biomarkers here to predict the presence of high Aβ in the brain, providing the opportunity to use these tests routinely both in the clinic and in clinical trials to aid in the DMT journey.

CONFLICT OF INTEREST STATEMENT

Christopher J. Fowler has no conflicts of interest. Ahmed Chenna, Mintzu Lo, Youssouf Badal, Brandon Yee, Robert Martone, John Winslow, and Christos Petropoulos are employees of Labcorp. Ahmed Chenna, Mintzu Lo, Robert Martone, John Winslow, and Christos Petropoulos have stock options in Labcorp. Ahmed Chenna, Robert Martone and John Winslow have received travel support for attending meetings from Labcorp. John Winslow has issued patents but not related to this work. James D. Doecke has received funding from Roche. Christopher C. Rowe has received funding/support from Eisai Australia, Lilly Australia, Novo Nordisk, Cerveau, Australian Dementia Network, and Prothena. Stephanie R. Rainey‐Smith has received funding/support from Mature Adults Learning Association, CogSleep, Alzheimer's Association, Australian Imaging, Lifestyle study of aging. Simon Laws participates on the External Research Advisory board for the Centre for Precision Health, the Scientific Advisory Board for Cytox Ltd., and the External Advisory Board for the Perron Institute. Simon Laws sits on the Steering Management Committee for the Centre for Precision Health, and the Scientific Management Committee for AIBL. Colin L. Masters reports ad hoc consultancy engagements and scientific advice with Actinogen, Acumen, Alterity, Biogen, Eisai, Eli ‐Lilly, Roche, and Beckman Coulter. Ralph N. Martins reports grants from CAA Consortium, Biogen, and Alnylam.

GRANTS

Christos J Petropoulos is principal investigator on a BARDA grant and has numerous grants, none of which are related to this work. Christopher C. Rowe has received grants from Eisai, Roche, Australian National Health and Medical Research Council, and Enigma Australia. Stephanie R. Rainey‐Smith has received grants from the Australian National Health and Medical Research Council, Alzheimer's Association Australia, Alzheimer's Drug Discovery Foundation, Bright Focus Foundation. Ralph N. Martins reports grants from CAA Consortium, Biogen, and Alnylam.

CONSENT STATEMENT

All participants gave their written informed consent for use of their biological material and clinical data for research purposes. Ethical approval was granted by the ethical committee of each participating center.

Supporting information

Supporting Information

Supporting Information

Doecke JD, Chenna A, Lo M, et al. Combining Lumipulse p‐tau217 and Aβ42/40 as confirmatory tests for Aβ positivity prior to disease‐modifying therapy. Alzheimer's Dement. 2025;21:e70707. 10.1002/alz.70707