Updated prevalence estimates of amyloid positivity from cognitively normal to clinical Alzheimer's disease dementia: The Amyloid Biomarker Study.

Background: We have updated our previous amyloid positivity prevalence estimates (Jansen, 2015; Ossenkoppele, 2015; Table 1) with data from 19,097 individuals from 85 cohorts. We tested if prevalence estimates differed for positron emission tomography (PET) or cerebrospinal fluid (CSF) amyloid. We calculated data‐driven CSF cut‐points as previous studies suggested a drift in Innotest‐assay performance. Method: We included 9,908 participants with normal cognition (CN); 1,542 with subjective cognitive decline (SCD); 5,405 with mild cognitive impairment (MCI); and 2,260 with clinical Alzheimer's disease (AD) dementia. 10,139 participants had an amyloid‐PET measure and 8,958 participants had an amyloid‐CSF measure. Amyloid‐ß aggregation was originally dichotomized as normal or abnormal according to study‐specific cut‐points, and we defined data‐driven CSF cut‐points using Gaussian mixture modelling. Generalized‐estimating‐equations adjusting for within‐study clustering of individuals were used to estimate amyloid positivity prevalence according to age, cognitive status, and biomarker modality and compared original with updated cut‐points. Result: Amyloid positivity prevalence was higher with older age and advancing disease severity. Using original cut‐points as provided by the cohort, the prevalence of amyloid positivity at the median age of 70 was similar for PET and CSF estimates for CN (24%), SCD (26%), and MCI (51%), and was higher in PET (87%) than CSF (79%) for clinical AD dementia. Using the data‐driven CSF cut‐points, the prevalence of amyloid positivity was higher for CSF‐based estimates than PET‐based estimates in CN (CSF 33%, PET 24%), SCD (CSF 36%, PET 27%), and MCI (CSF 60%, PET 49%), but comparable in CSF and PET in clinically diagnosed AD dementia (CSF 83%, PET 87) (Table 1, Figure 1). Conclusion: We provide updated amyloid positivity estimates from the Amyloid Biomarker Study. CSF‐based estimates using a data‐driven approach resulted in higher estimates (up to 11% higher) in people without clinical AD dementia than PET‐based estimates. Whether CSF‐based estimates are more sensitive than PET‐based estimates for amyloid pathology among people without dementia needs to be explored in cohorts that use both modalities. These updated estimates may be useful to understand potential eligible patient population sizes for anti‐amyloid therapies and to inform recruitment strategies for clinical trials investigating anti‐amyloid therapies. [ABSTRACT FROM AUTHOR]