Abstract
INTRODUCTION: The ATN framework provides an in vivo diagnosis of Alzheimer's disease (AD) using cerebrospinal fluid (CSF) biomarkers of pathologic amyloid plaques (A), tangles (T), and neurodegeneration (N). ATN is rarely evaluated in pathologically confirmed patients and its poor sensitivity to suspected non-Alzheimer's pathophysiologies (SNAP), including frontotemporal lobar degeneration (FTLD), leads to misdiagnoses. We compared accuracy of ATN (ATN(TAU) ) using CSF total tau (t-tau) to a modified strategy (ATN(NfL) ) using CSF neurofilament light chain (NfL) in an autopsy cohort. METHODS: ATN(TAU) and ATN(NfL) were trained in an independent sample and validated in autopsy-confirmed AD (n = 67) and FTLD (n = 27). RESULTS: ATN(NfL) more accurately identified FTLD as SNAP (sensitivity = 0.93, specificity = 0.94) than ATN(TAU) (sensitivity = 0.44, specificity = 0.97), even in cases with co-occurring AD and FTLD. ATN(NfL) misclassified fewer AD and FTLD as "Normal" (2%) than ATN(TAU) (14%). DISCUSSION: ATN(NfL) is a promising diagnostic strategy that may accurately identify both AD and FTLD, even when pathologies co-occur.