Determining the amyloid PET and CSF inclusion criteria for the SKYLINE secondary prevention study with gantenerumab

Abstract

AbstractBackgroundSKYLINE is a secondary prevention study in Alzheimer's disease (AD) enrolling amyloid‐positive, cognitively normal (CN) individuals at risk for clinical AD between 60–80 years and treating them with gantenerumab or placebo for 4 years. There is no gold standard for determining amyloid positivity for this purpose. Thresholds for amyloid positivity used in trials in symptomatic AD may be suboptimal in a secondary prevention setting, where amyloid positivity prevalence is lower. We compared various amyloid‐positive definitions in a CN elderly population based on how well they enrich for clinical progression over 4 years and operational feasibility.MethodFour cohorts (ADNI, BioFINDER‐1, AIBL, and HABS) were used to explore the relationship between amyloid PET visual read, centiloid values and CSF pTau/Abeta42 values in a CN population. For amyloid PET, we compared visual read with centiloid thresholds by evaluating the amyloid positivity prevalence, magnitude and variability of endpoint progression, and positive and negative predictive value (PPV and NPV). Operational advantages of CSF motivated a comparison of elevated CSF pTau/Abeta42 thresholds with amyloid PET. We interrogated which CSF thresholds were robust to preanalytical differences, had a high PPV and NPV compared with amyloid PET, and resulted in an acceptable degree of cognitive decline and variability.ResultHigher centiloid thresholds (≥40 CL) showed good agreement with visual read, but visual read resulted in a lower coefficient of variation (CV) for change from baseline in PACC5, the primary endpoint in SKYLINE (CV=118–210 vs. CV=150–336). An elevated CSF pTau/Abeta42 threshold resulted in a similar amyloid positivity prevalence to amyloid PET (14–17% vs. 10–20%), had acceptable NPV (95–97%), PPV (69–87%), sensitivity (71–82%), and specificity (95–98%) for visual read, and resulted in a similar CV for change from baseline in PACC5.ConclusionThere is no gold standard for determining the optimal amyloid positivity threshold in a secondary prevention trial. Determining patients to be amyloid‐positive if they are amyloid PET visual read‐positive or have elevated CSF pTau/Abeta42 offers not only operational advantages but also likely results in a trial population that has an acceptable level of endpoint progression and variability over the trial duration.