Decision tree supports the interpretation of CSF biomarkers in Alzheimer's disease

Abstract

Abstract Introduction We developed and validated a clinically applicable decision tree for the use of cerebrospinal fluid biomarkers in the diagnosis of Alzheimer's disease (AD). Methods Subjects with probable AD (n = 1004) and controls (n = 442) were included. A decision tree was modeled using Classification And Regression Tree analysis in a training cohort (AD n = 221; controls n = 221) and validated in an independent cohort (AD n = 783; controls n = 221). Diagnostic performance was compared to previously defined cutoffs (amyloid β 1‐42 375 pg/ml). Results Two cerebrospinal fluid AD biomarker profiles were revealed: the “classical” AD biomarker profile (amyloid β 1‐42: 647‐803 pg/ml; tau >374 pg/ml) and an “atypical” AD biomarker profile with strongly decreased amyloid β 1‐42 (<647 pg/ml) and normal tau concentrations (<374 pg/ml). Compared to previous cutoffs, the decision tree performed better on diagnostic accuracy (86% [84‐88] vs 80% [78‐83]). Discussion Two cerebrospinal fluid AD biomarker profiles were identified and incorporated in a readily applicable decision tree, which improved diagnostic accuracy.

Keywords

• Alzheimer's disease
• CSF
• Amyloid β 1‐42
• Tau
• Clinical implementation
• Decision tree