Interval‐specific likelihood ratios for improving differential diagnosis of Alzheimer’s disease using biomarkers in cerebrospinal fluid

Abstract

AbstractBackgroundSingle cutoff values are currently applied to interpret Alzheimer biomarker results. Yet these cutoffs are not uniform among centers, nor highly specific to discriminate AD from other disorders. We examined if the likelihood of AD was higher when biomarker levels deviated more from normal levels. For this purpose, we constructed interval‐specific likelihood ratios (LRs) for CSF biomarkers amyloid Aβ1‐42, total Tau protein (tTau) and hyperphosphorylated Tau protein (pTau181).MethodThis retrospective single center study included 208 patients who 1) attended the Neurology Memory Clinic of the University Hospital Leuven in the period of September 2011 till July 2017, 2) underwent CSF AD biomarkers assessment in the diagnostic workup (Innotest, Fujirebio) and 3) had a stable clinical diagnosis for at least three years after diagnosis. For each biomarker, we calculated LRs for AD for different test result intervals, based on 2 methods, including intervals with boundaries equal to limited multiples of previously published cutoff values (method 1), or to generate intervals well distributed over the entire dynamic range of biomarker levels (method 2). In a graphical representation, post‐test probability of AD was related to pre‐test probability and test result interval for each biomarker.ResultBoth methods for constructing intervals showed that for any of the biomarkers the LR increased significantly once deviating from normal values. Further dividing these aberrant levels into additional intervals (method 2) further increased the LR. Further dividing normal levels into more intervals did further decrease the LR, except for Aβ1‐42. No LRs > 10 or < 0.1, suggesting clinically meaningful differences between pre‐test and post‐test probability, were encountered. LRs between 0.1 – 0.2 or 5 – 10, suggestive of modest clinical differences, were seen for normal Aβ1‐42 levels (800 – 1,000 pg/mL; LR = 0.168), and aberrant tTau levels (≥ 1,000 pg/mL; LR = 5.540) and pTau181 levels (≥ 80 pg/mL; LR between 5.936 and 7.420).ConclusionThe likelihood of AD changed as CSF levels of AD biomarkers deviated more outspoken from normal range, yet related to a modest clinical significance given the partial overlap between AD and non‐AD of biomarker values that deviated from normal values.