Population-based blood screening for preclinical Alzheimer's disease in a British birth cohort at age 70

Abstract

Alzheimer's disease has a preclinical stage when cerebral amyloid-β deposition occurs before symptoms emerge, and when amyloid-β-targeted therapies may have maximum benefits. Existing amyloid-β status measurement techniques, including amyloid PET and CSF testing, are difficult to deploy at scale, so blood biomarkers are increasingly considered for screening. We compared three different blood-based techniques-liquid chromatography-mass spectrometry measures of plasma amyloid-β, and single molecule array (Simoa) measures of plasma amyloid-β and phospho-tau181-to detect cortical 18F-florbetapir amyloid PET positivity (defined as a standardized uptake value ratio of >0.61 between a predefined cortical region of interest and eroded subcortical white matter) in dementia-free members of Insight 46, a substudy of the population-based British 1946 birth cohort. We used logistic regression models with blood biomarkers as predictors of amyloid PET status, with or without age, sex and APOE ε4 carrier status as covariates. We generated receiver operating characteristics curves and quantified areas under the curves to compare the concordance of the different blood tests with amyloid PET. We determined blood test cut-off points using Youden's index, then estimated numbers needed to screen to obtain 100 amyloid PET-positive individuals. Of the 502 individuals assessed, 441 dementia-free individuals with complete data were included; 82 (18.6%) were amyloid PET-positive. The area under the curve for amyloid PET status using a base model comprising age, sex and APOE ε4 carrier status was 0.695 (95% confidence interval: 0.628-0.762). The two best-performing Simoa plasma biomarkers were amyloid-β42/40 (0.620; 0.548-0.691) and phospho-tau181 (0.707; 0.646-0.768), but neither outperformed the base model. Mass spectrometry plasma measures performed significantly better than any other measure (amyloid-β1-42/1-40: 0.817; 0.770-0.864 and amyloid-β composite: 0.820; 0.775-0.866). At a cut-off point of 0.095, mass spectrometry measures of amyloid-β1-42/1-40 detected amyloid PET positivity with 86.6% sensitivity and 71.9% specificity. Without screening, to obtain 100 PET-positive individuals from a population with similar amyloid PET positivity prevalence to Insight 46, 543 PET scans would need to be performed. Screening using age, sex and APOE ε4 status would require 940 individuals, of whom 266 would proceed to scan. Using mass spectrometry amyloid-β1-42/1-40 alone would reduce these numbers to 623 individuals and 243 individuals, respectively. Across a theoretical range of amyloid PET positivity prevalence of 10-50%, mass spectrometry measures of amyloid-β1-42/1-40 would consistently reduce the numbers proceeding to scans, with greater cost savings demonstrated at lower prevalence.

Keywords: Alzheimer’s disease; amyloid imaging; beta-amyloid; dementia; epidemiology; tau.

Figure 1

Formulae. (A) Calculations for number needed to screen to obtain 100 amyloid PET-positive individuals. Sensitivity and specificity are independent of prevalence, and are fixed once the cut-off point is chosen (e.g. at Youden’s index). If prevalence is specified, the first three simultaneous equations can be solved for B–D. The positive and negative predictive values of the plasma test and the number needed to screen can then be calculated by the following three equations. (B) Calculations for relative cost of the screening programme. These calculations are based on specified costs of an individual PET scan (x) and blood test (y), number needed to prescreen (NNS) with blood test and number proceeding to scan (NPS), to obtain a specified number of amyloid PET-positive individuals (n) in the context of a known estimated population prevalence of amyloid PET positivity (p). It is assumed that x and y include initial setup costs.

Figure 2

Recruitment flow chart. The chart shows an overview of Insight 46 recruitment from the MRC NSHD and summary of blood biomarker data available [modified with permission (James et al., 2018) under the terms of the Creative Commons Attribution 4.0 International Licence (http://creativecommons.org/licenses/by/4.0/)]. The derivation of the dementia free group (used for the main analyses) and the cognitively normal group (used in the sensitivity analysis for prediction of amyloid PET status) is shown. Aβ = amyloid-β.

Figure 3

Concordance of blood biomarkers with amyloid PET SUVR in dementia-free individuals in Insight 46 (n = 441). (A) ln Simoa plasma amyloid-β₄₂/amyloid-β₄₀. (B) ln Simoa plasma p-tau181. (C) ln LC-MS plasma amyloid-β_1-42/amyloid-β_1-40. (D) LC-MS plasma composite. Dashed vertical lines represent the SUVR cut-off point for PET positivity. Dashed horizontal lines represent the LC-MS biomarker cut-off points corresponding with Youden’s index derived by ROC analyses. Concordant classification by PET and plasma is represented by the blue area and discordant classification by the orange area on each graph. The non-log transformed cut-off points are 0.058 for Simoa plasma amyloid-β₄₂/amyloid-β₄₀, 10.8 pg/ml for Simoa plasma p-tau181, 0.095 for LC-MS plasma amyloid-β_1-42/amyloid-β_1-40 and −0.049 for LC-MS plasma composite. Aβ = amyloid-β; p-tau181 = phospho-tau181.

Figure 4

Hypothetical performance of the best-performing Simoa and LC-MS tests by prevalence of amyloid PET positivity. (A) Simoa biomarker alone. (B) LC-MS biomarker alone. (C) Simoa biomarker + age + sex + APOE ε4 carrier status. (D) LC-MS biomarker + age + sex + APOE ε4 carrier status. Lines were modelled by computing the positive (PPV) and negative predictive values (NPV) of the tests and the scan reduction afforded at specified values of amyloid PET positivity prevalence over the range 10–50%, in 5% intervals. Scan reduction is the percentage of scans saved relative to the number of scans that would be required without screening (calculated according to the specified prevalence of amyloid PET positivity). Solid lines show PPV, dashed lines show NPV and dotted lines show scan reduction. Aβ = amyloid-β; p-tau181 = phospho-tau181.

Figure 5

Relative costs of screening programmes using different blood tests. (A) Using different blood tests, at fixed prevalence of amyloid PET positivity. (B) Using LC-MS plasma amyloid-β_1-42/amyloid-β_1-40 over a range of 10–50% of theoretical prevalence of amyloid PET positivity. Aβ = amyloid-β; p-tau181 = phospho-tau181.