Predicting amyloid PET and tau PET stages with plasma biomarkers

Abstract

Staging the severity of Alzheimer's disease pathology using biomarkers is useful for therapeutic trials and clinical prognosis. Disease staging with amyloid and tau PET has face validity; however, this would be more practical with plasma biomarkers. Our objectives were, first, to examine approaches for staging amyloid and tau PET and, second, to examine prediction of amyloid and tau PET stages using plasma biomarkers. Participants (n = 1136) were enrolled in either the Mayo Clinic Study of Aging or the Alzheimer's Disease Research Center; had a concurrent amyloid PET, tau PET and blood draw; and met clinical criteria for cognitively unimpaired (n = 864), mild cognitive impairment (n = 148) or Alzheimer's clinical syndrome with dementia (n = 124). The latter two groups were combined into a cognitively impaired group (n = 272). We used multinomial regression models to estimate discrimination [concordance (C) statistics] among three amyloid PET stages (low, intermediate, high), four tau PET stages (Braak 0, 1-2, 3-4, 5-6) and a combined amyloid and tau PET stage (none/low versus intermediate/high severity) using plasma biomarkers as predictors separately within unimpaired and impaired individuals. Plasma analytes, p-tau181, Aβ1-42 and Aβ1-40 (analysed as the Aβ42/Aβ40 ratio), glial fibrillary acidic protein and neurofilament light chain were measured on the HD-X Simoa Quanterix platform. Plasma p-tau217 was also measured in a subset (n = 355) of cognitively unimpaired participants using the Lilly Meso Scale Discovery assay. Models with all Quanterix plasma analytes along with risk factors (age, sex and APOE) most often provided the best discrimination among amyloid PET stages (C = 0.78-0.82). Models with p-tau181 provided similar discrimination of tau PET stages to models with all four plasma analytes (C = 0.72-0.85 versus C = 0.73-0.86). Discriminating a PET proxy of intermediate/high from none/low Alzheimer's disease neuropathological change with all four Quanterix plasma analytes was excellent but not better than p-tau181 only (C = 0.88 versus 0.87 for unimpaired and C = 0.91 versus 0.90 for impaired). Lilly p-tau217 outperformed the Quanterix p-tau181 assay for discriminating high versus intermediate amyloid (C = 0.85 versus 0.74) but did not improve over a model with all Quanterix plasma analytes and risk factors (C = 0.85 versus 0.83). Plasma analytes along with risk factors can discriminate between amyloid and tau PET stages and between a PET surrogate for intermediate/high versus none/low neuropathological change with accuracy in the acceptable to excellent range. Combinations of plasma analytes are better than single analytes for many staging predictions with the exception that Quanterix p-tau181 alone usually performed equivalently to combinations of Quanterix analytes for tau PET discrimination.

Keywords: Alzheimer’s biomarkers; amyloid PET; plasma biomarkers; staging Alzheimer’s disease; tau PET.

Figure 1

Topographic staging of amyloid and tau PET. Box plots of meta-region of interest PET values by PET topographic stage for amyloid PET and tau PET.

Figure 2

Box plots of amyloid PET, tau PET and plasma biomarkers by amyloid magnitude stage (left) and tau PET topographic stage (right).

Figure 3

Discrimination between lower and higher PET stage pairs among CU participants for several models including plasma analyte(s) as predictors as well as a base model with risk variables. Concordance (95% confidence interval) estimates from multinomial regression models fit among CU participants are shown on the top row (forest plots) and ROC curves for the same models are shown in the bottom row. The colours in the forest plots in the top row for each model match the corresponding colours of the ROC curves in the bottom row. The numbers in the bottom right of each ROC plot represent the C statistic estimate from the corresponding forest plot above it. Separate models were fit for the amyloid PET magnitude stage outcome, the tau PET topographic stage outcome and the tau PET magnitude stage outcome. The columns represent different contrasts between the PET stages. Braak 3–4 and Braak 5–6 were combined due to small numbers in these groups among the CU participants. Models were compared to answer the following: (i) how well does each plasma analyte discriminate among amyloid PET and tau PET stages when added individually to a base model consisting of risk factors? and (ii) is discrimination between PET stages improved over the base model or models with individual analytes plus base when all plasma analytes were included in the model simultaneously? Footnotes below the top row show which comparisons were significantly different with P < 0.05.

Figure 4

Discrimination between lower versus higher PET stage pairs among CI participants for several models including plasma analyte(s) as predictors as well as a base model with risk variables. Concordance (95% confidence interval) estimates from multinomial regression models fit among CI participants are shown on the top row (forest plots) and ROC curves for the same models are shown in the bottom row. The colours in the forest plots in the top row for each model match the corresponding colours of the ROC curves in the bottom row. The numbers in the bottom right of each ROC plot represent the C statistic estimate from the corresponding forest plot above it. Separate models were fit for the amyloid PET magnitude stage outcome, the tau PET topographic stage outcome and the tau PET magnitude stage outcome. The columns represent different contrasts between the PET stages. Models were compared to answer the following: (i) how well does each plasma analyte discriminate among amyloid PET and tau PET stages when added individually to a base model consisting of risk factors? and (ii) is discrimination between PET stages improved over the base model or models with individual analytes plus base when all plasma analytes were included in the model simultaneously? Footnotes below the top row show which comparisons were significantly different with P < 0.05.

Figure 5

Discrimination between lower versus higher PET stage pairs among all participants (CU and CI) for several models including plasma analyte(s) as predictors as well as a base model with risk variables. Concordance (95% confidence interval) estimates from multinomial regression models fit among all participants are shown on the top row (forest plots) and ROC curves for the same models are shown in the bottom row. The colours in the forest plots in the top row for each model match the corresponding colours of the ROC curves in the bottom row. The numbers in the bottom right of each ROC plot represent the C statistic estimate from the corresponding forest plot above it. Separate models were fit for the amyloid PET magnitude stage outcome, the tau PET topographic stage outcome and the tau PET magnitude stage outcome. The columns represent different contrasts between the PET stages. Models were compared to answer the following: (i) how well does each plasma analyte discriminate among amyloid PET and tau PET stages when added individually to a base model consisting of risk factors? and (ii) is discrimination between PET stages improved over the base model or models with individual analytes plus base when all plasma analytes were included in the model simultaneously? Footnotes below the top row show which comparisons were significantly different with P < 0.05. The risk factor model included age, sex, APOE genotype, a three-level study and clinical diagnosis variable (MCSA CU, MCSA CI, ADRC CI) and an interaction with age and the study/clinical diagnosis variable.

Figure 6

Discrimination between a combined amyloid and tau PET stage as a proxy of intermediate/high versus none/low Alzheimer's disease neuropathologic change among CU, CI and all participants for models including plasma analyte(s) as predictors and a base model with risk variables. Concordance (95% confidence interval) estimates from logistic regression models fit among CU (left), CI (middle) and all participants (right) are shown in the top row and ROC curves for the same models are shown in the bottom row. The numbers in the bottom right of each ROC plot represent the C statistic estimate from the corresponding forest plot above it. Models were compared to answer the following: (i) how well does each plasma analyte discriminate among the combined amyloid and tau PET stage when added individually to a base model consisting of risk factors? and (ii) is discrimination between PET stages improved over the base model or models with individual analytes plus base when all plasma analytes were included in a model simultaneously? Footnotes below the top row show which comparisons were significantly different with P < 0.05. The base model for the CU and CI models included age, sex and APOE genotype. The base model for the models fit among all participants included age, sex, APOE genotype, a three-level study and clinical diagnosis variable (MCSA CU, MCSA CI, ADRC CI) and an interaction with age and the study/clinical diagnosis variable.

Figure 7

Discrimination between lower and higher PET stage pairs among the subset of CU participants with Lilly plasma p-tau217 for several models including plasma analyte(s) as predictors as well as a base model with risk variables. Concordance (95% confidence interval) estimates from multinomial regression models fit among the subset of CU participants with Lilly p-tau217 data are shown on the top row (forest plots) and ROC curves for the same models are shown in the bottom row. The colours in the forest plots in the top row for each model match the corresponding colours of the ROC curves in the bottom row. The numbers in the bottom right of each ROC plot represent the C statistic estimate from the corresponding forest plot above it. Separate models were fit for the amyloid PET magnitude stage outcome, the tau PET topographic stage outcome and the tau PET magnitude stage outcome. The columns represent different contrasts between the PET stages. T^INT and T^HIGH were combined and Braak 1–2, 3–4 and 5–6 were combined due to small numbers in these groups among CU in the Lilly subset. Models were compared to answer the following: (i) how well does Quanterix p-tau181, Lilly p-tau217, and the combination of all four Quanterix biomarkers discriminate among amyloid PET and tau PET stages when added to a base model consisting of risk factors? (ii) is discrimination between PET stages improved when using Lilly p-tau217 added to the base model compared to Quanterix p-tau181 added to the base model? and (iii) is discrimination between PET stages improved when using Lilly p-tau217 added to the base model compared to a model with all Quanterix plasma analytes plus the base model? Footnotes below the top row show which comparisons were significantly different with P < 0.05.