Premium
Application of two‐graph receiver operating characteristics for defining intermediate amyloid‐beta CSF cutoffs
Author(s) -
Brum Wagner Scheeren,
De Bastiani Marco Antônio,
Souza João Pedro Ferrari,
Bieger Andrei,
Therriault Joseph,
Benedet Andréa Lessa,
Pascoal Tharick A.,
Povala Guilherme,
Ferreira Pamela C.L.,
Souza Diogo O.,
RosaNeto Pedro,
Zimmer Eduardo R.
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.045873
Subject(s) - biomarker , cutoff , receiver operating characteristic , medicine , cerebrospinal fluid , oncology , biology , physics , biochemistry , quantum mechanics
Background The NIA‐AA 2018 Research Framework envisioned Alzheimer's disease (AD) as a biological construct, dichotomizing individuals into normal and abnormal for each biomarker category. Furthermore, it suggests that developing intermediate‐range cutoffs would aid in advancing AD research. However, few statistical methodologies have been proposed to define what would constitute an intermediate range for an AD biomarker. Here, we apply a three‐range method for defining cutoff in amyloid‐β (Aβ) cerebrospinal fluid (CSF) measurements to assess intermediate range's predictive value over clinical progression. Considering it is expected that a biomarker loses its predictive value in the perithreshold zone, we hypothesized that Aβ‐intermediate (Ab Int ) status of cognitively unimpaired (CU) would not be able to predict progression to symptomatic stages of AD. Method For defining intermediate‐range cutoffs with two‐graph receiver operating characteristics (TG‐ROC) method, 807 individuals with baseline [ 18 F]AV45 Aβ‐PET and CSF Aβ 1‐42 Elecsys ® biomarkers were selected from the ADNI. Mean standardized uptake value ratio cutoff of 1.11 for [ 18 F]AV45 was employed to stratify individuals into true‐positive (n=433) and true‐negative (n=374). Based on TG‐ROC cutoffs, 336 CU individuals with baseline CSF Aβ 1‐42 Elecsys ® were divided into Aβ‐, Aβ int and Aβ+ and followed‐up for 6 years. Predictive values of baseline biomarker status for clinical progression were assessed using Cox proportional hazards models adjusted for age, gender, APOE and baseline MMSE. Result TG‐ROC analysis yielded cutoffs for Aβ‐ (>1131 pg/mL; n=199), Aβ Int (820.1