DEVELOPMENT OF A NOVEL MULTI-BIOMARKER ASSESSMENT SCORE FOR ALZHEIMER’S DISEASE

Abstract

There is an unmet need for laboratory tools that detect Alzheimer's disease (AD) in its early stages. Biomarkers in cerebrospinal fluid (CSF) have shown great potential in fulfilling this need. The ratio of the peptides β-amyloid (Aβ) 40 and 42 in CSF predicts disease progression. The levels of apolipoprotein E (ApoE) in CSF, and the presence of 1 or more ε4 alleles (APOE4), indicate disease likelihood. Here, we describe a novel algorithm that combines these biomarkers. We have used the algorithm to develop an assessment score that categorizes an individual's likelihood of having AD as higher, lower, or intermediate. Aβ42/ Aβ40 ratio, APOE4allele count, and total ApoE (μg/mL) were determined using mass spectrometry in a training set of 151 specimens from clinically-diagnosed AD and non-AD patients (supplied by UC San Diego ADRC Neuropathology Core and Brain Bank). A diagnosis of AD was based on imaging and cognition tests. The results from the training set were used to develop a logistic regression model. The linear predictor from the regression model was used to create an AD likelihood score. The score was partitioned into 3 groups: higher, intermediate, and lower likelihood. The higher likelihood threshold was selected to allow for a 10% false positive rate, while the lower likelihood threshold was selected to allow for a 10% false negative rate. We validated the model in an initially-blinded separate test cohort of 159 patients independently diagnosed as AD or non-AD using combined genomic and immunochemical methodology. Of the 159 test subjects, 79 were previously diagnosed as AD and 80 were diagnosed as non-AD. Categorization based on logistic regression analysis results is presented in Table 1. Of those in the higher likelihood category, 85.7% (36/42) had at least one APOE4 allele: 47.6% ε2/ε4 or ε3/ε4 (20/42), and 38.1% ε4/ε4 (16/42). The remaining higher likelihood samples, 14.3% (6/42), were APOE4-negative. We have devised a logistic regression model using biomarkers in CSF that has potential to stratify the likelihood of Alzheimer's disease.