Establishing an Autopsy-Based Cerebrospinal Fluid Biomarker Signature in Alzheimer Disease Patients.

Abstract

Featured Article: Shaw LM, Vanderstichele H, Knapik-Czajka M, Clark CM, Aisen PS, Petersen RC, et al. Cerebrospinal fluid biomarker signature in Alzheimer's Disease Neuroimaging Initiative subjects. Ann Neurol 2009;65:403-13.† At the time this work was conducted (2007 to 2008), the research use of cerebrospinal fluid (CSF) biomarker proteins linked to Alzheimer disease (AD) hallmarks—β-amyloid (Aβ) plaque load and tau pathology (Aβ1-42, t-tau, and p-tau181) —was becoming established practice amongst a number of experienced laboratories as reliable and accurate indices of brain pathology in living subjects (1–3). These studies mainly used either a widely employed ELISA methodology (4) or the newer (at the time) microbead-based biomarkers multi analyte profiling (xMAP) immunoassay format, developed by Innogenetics (now Fujirebio) (5). The data generated from more than 25 studies that used these techniques and CSF provided strong support for a more objective measurement of amyloid and tau pathology in living individuals, compared to clinical diagnosis of AD dementia (6). The clinical diagnosis of AD is only up to about 80% accurate and our 2009 study in Annals of Neurology and other studies provided the foundation for moving towards a biological basis (protein biomarker profile) for more accurate clinical diagnosis of the presence of ongoing AD brain pathology. Two populations were used in our study: (a) participants in the multicenter study, the Alzheimer's Disease Neuroimaging Initiative (ADNI; 100 patients with early AD, 196 patients with mild cognitive impairment, and 115 cognitively normal age-matched controls, all clinically diagnosed), provided CSF samples collected at 56 centers at the participants’ baseline visit, and (b) an independent population, followed at the University of Pennsylvania Alzheimer's Disease Clinical Core, of 56 autopsy-confirmed AD participants and 52 age-matched living cognitively normal controls. Besides inclusion of an ADNI-independent population for assessment of cutpoint values for each of the 3 AD biomarkers as well as for the ratios t-tau/Aβ1-42 and p-tau181/Aβ1-42, another key feature in this study was the close collaboration between an in vitro diagnostic company and a university laboratory with the aim of generating best-in-class assays for the benefit of patients, and the inclusion of autopsy diagnosis for the patients with AD, the accepted gold standard for diagnosis of this disease. This enabled (a) determination of the accuracy of prediction of AD pathology assessed by receiver operating characteristic curve (AUC) analyses, (b) determination of cutpoints for the individual proteins and their ratios, and (c) establishment of a logistic regression model that combined Aβ1-42, t-tau, and APOε4 genotype. AUC values ranged from 0.913 for CSF Aβ1-42 concentration alone up to 0.942 for the combination of Aβ1-42, t-tau, and APOε4 genotype. Importantly, the same lot of reagents was used for the analyses in both populations, thus removing lot-to-lot variability from measurement errors. At the time this work was done, it was a considerable source of variability using this “for research-use-only (RUO)” immunoassay. It was an important additional feature of this study.