CENTILOID CUT-OFF VALUES FOR OPTIMAL AGREEMENT BETWEEN AMYLOID PET AND CSF CORE AD BIOMARKERS

Abstract

The Centiloid scale was developed to standardize amyloid (Aβ) PET quantification across centers and tracers. This scale allows better comparability in Aβ PET continuous measurements for multicenter studies and clinical trials. In this study we aimed to derive Aβ positivity Centiloid cut-offs against cerebrospinal fluid (CSF) core Alzheimer's disease (AD) biomarkers. A total of 516 participants of the ALFA+ Study (N=205) and ADNI (N=311) covering all the AD spectrum (Figure 1) underwent amyloid PET imaging and determination of core AD CSF biomarkers with the Elecsys platform. Tracer uptake was quantified in Centiloid units (CL). Centiloid cut-offs were sought that maximize the agreement between PET and dichotomous determinations based on the CSF levels of Aβ42, tTau, pTau and their ratios, using pre-established threshold values. Optimal cut-offs were selected as those that maximized either Youden's Index (YI) or the overall percentage agreement (OPA). Positive and negative percentage agreements (PPA and NPA) of these cut-offs were also calculated. Receiver operating characteristic (ROC) analysis was performed for comparison of all CSF biomarkers. As additional analysis, we also derived Centiloid cut-offs using more liberal CSF thresholds (including “grey zone”) and on both cohorts separately. A cut-off of 12 CL maximized the agreement with CSF Aβ42 participants' classification. Cut-offs derived against tau CSF biomarkers felt between 25–35 CL (Table 1). Tau ratios showed better agreement than the remaining CSF biomarkers (Figure 2). Centiloid cut-offs robustness was confirmed even when established against variations of CSF cut-offs (Table 2), but showed some variations when derived on both cohorts independently.