An unbiased, staged, multicentre, validation strategy for Alzheimer's disease CSF tau levels
ABSTRACT Newly proposed diagnostic criteria for Alzheimer's disease include cerebrospinal fluid (CSF) tau levels as one core supportive criterion. The published high sensitivity and specificity figures for CSF tau levels in Alzheimer's disease are offset by the large range of proposed cutoff values (9.6 pg/mL to 1140 pg/mL). This study aimed to provide guidance on how to establish, validate and audit CSF tau cutoff values using an unbiased, two-stage multicentre strategy. Both receiver operator characteristics (ROC) optimised and population-based cutoff values were calculated on a pilot dataset (n=99), validated in a large dataset (n=560) and then compared to the literature. The data suggest using an ROC optimised cutoff level of 323+/-51.7 pg/mL allowing for the published inter-laboratory coefficient of variation of 16%. This cutoff level was confirmed in a prospective audit (n=100). As demand for CSF tau levels will increase globally, the accuracy of local CSF hTau cutoff levels can be compared against this benchmark.
[show abstract] [hide abstract]
ABSTRACT: The objective of this work was to improve the clinical diagnosis of Alzheimer's disease (AD) by proposing a simple decision tree based on three major biomarkers of AD found in the cerebrospinal fluid (CSF): amyloid peptide Aβ1-42, total Tau (t-Tau) and Tau phosphorylated at Thr181 (p-Tau). Two consecutive cohorts comprising 548 patients in total were recruited by the Memory and Neurology Clinics at Lille University Hospital (France). These included 293 patients with AD, 171 patients with other dementias and 84 healthy controls. All patients underwent lumbar puncture for the assessment of CSF concentrations of Aβ1-42, t-Tau and p-Tau. International criteria for dementias were used for diagnosis by investigators blind to CSF test results. To identify the combination of biomarkers that best predicted the 3 diagnoses, we used the CHAID decision tree method with the first cohort. Our analysis yielded a two-step decision tree, with a first stratification step based on the Aβ1-42/p-Tau ratio of the CSF, and a second step based on CSF p-Tau concentrations. The second cohort was then used to determine the power (0.618), sensitivity (82%) and specificity (81%) of this tree in AD diagnosis. These were found to be at least as high as those of other known algorithms based on the three CSF biomarkers, Aβ1-42, t-Tau and p-Tau. For the first time, diagnostic rules for AD based on CSF variables were compared in a single study. Our findings indicate that the measurement of Aβ1-42 and p-Tau levels in the CSF is sufficient to diagnose AD.Current Alzheimer research 09/2012; · 4.97 Impact Factor