Amyloid-PET pipeline choice influences classification of preclinical Alzheimer's disease

<jats:p>BACKGROUND: Quantitative amyloid-beta (Aβ)-PET is increasingly used in AD prevention trials. Although the Centiloid (CL) framework provides a common scale, variability persists across processing pipelines, including differences in template/native space, partial volume correction (PVC), and reference region. These choices may influence cut-points, and in turn positivity rates, as well as longitudinal accumulation rates. We examined cut-point estimates and inter-pipeline discordance in a community cohort where many are expected to have early Aβ deposition. METHODS: We analysed [18F]florbetapir PET/MR data from predominantly cognitively unimpaired (~95%) individuals aged ~71 years at baseline (n=433) and at follow-up (n=328; ~2.4-year interval) in Insight 46 (1946 British birth cohort). Centiloids were derived using the standard pipeline and ten in-house pipelines employing alternative reference regions and PVC in native space. Gaussian mixture modelling estimated cut-points with bootstrapped uncertainty. We assessed Aβ-discordance across pipelines as a function of standard CLs and examined follow-up CSF Aβ42/Aβ40 (n=120) and PET in individuals with discordant baseline classifications. RESULTS: Baseline cut-points were 10–23 CL across pipelines, classifying 16–25% as Aβ-positive. Reliable accumulation cut-points were 3.5–6 CL/year, identifying 16–22% as accumulators. Uncertainty varied across pipelines. At baseline, 18% were discordant across PET measures, predominantly between 11–35 standard CLs. The discordant group showed higher Aβ-PET accumulation and lower CSF Aβ42/Aβ40 than concordant negatives. CONCLUSIONS: Disagreement between Aβ-PET methods was highest between 11–35 standard Centiloids and was frequently associated with accumulating Aβ. These findings highlight the importance of considering cut-point uncertainty and methodological influences when interpreting early-stage amyloidosis.</jats:p>