Evaluation of novel data‐driven metrics of amyloid β deposition for longitudinal PET studies. (20th December 2022)

• Record Type: Journal Article
• Authors: Bollack, Ariane; Cash, David M; Markiewicz, Pawel J; Wink, Alle Meije; Prosser, Lloyd; Coath, William; Pemberton, Hugh; Farrar, Gill; Barkhof, Frederik
• Is Part Of: Alzheimer's & dementia. Volume 18(2022)Supplement 5
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-12-20
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.064612
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Positron emission tomography (PET) provides in vivo quantification of amyloid‐β (Aβ) pathology. The most common method for assessing PET is standardized uptake value ratio (SUVr), which can be affected by physiological and technical factors. Novel, data‐driven metrics have been developed to address these sources of variability. We evaluate cross‐sectional and longitudinal performance of four data driven metrics. Method: Three cohorts were used for evaluation: Insight 46, a neuroimaging sub‐study of the 1946 British birth cohort, Australian Imaging, Biomarker & Lifestyle Flagship Study of Ageing, and a test‐retest flutemetamol dataset (Table 1). Data from participants were included if good quality volumetric MRI and amyloid PET, using florbetapir (Insight 46) or flutemetamol (AIBL and test‐retest data), was available. No partial volume correction was applied. Four data driven metrics were extracted: the Centiloid derived from non‐negative matrix factorisation (CLNMF), the Aβ PET pathology accumulation index (Aβ‐index), amyloid load (Aβ‐load), and amyloid pattern similarity score (AMPSS). These data‐driven metrics were compared to a global composite SUVr, with either a pons (flutametamol) or cerebellar grey matter (florbetapir) reference region, and a Centiloid (CL) score. They were evaluated by measures of repeatability in test‐retest data, associations with non‐displaceable binding potential (BPND, computed from Logan graphical analysis), sample size estimates to detect a 20% slowing in Aβ accumulation with 95% significance and 80% power, and accuracy in predicting the second follow‐up visit. Result: All metrics show good to excellent reliability. The repeatability is highly influenced by the amyloid burden, the range, and the offset of each metric. (Table 1) All metrics are strongly correlated to the BPND ( R 2 : 0.8 to 0.94), with SUVr and CL explaining more variance in BPND than the data‐driven metrics. (Figure) Sample size estimates were lowest in CL and CLNMF compared to the SUVr. (Table 2) The Aβ‐index had the best predictive power over ∼3 years while the other metrics were comparable. (Table 3) Conclusion: Novel data driven metrics provide comparable performance in terms of accuracy and performance to more established quantification methods of Aβ PET tracer uptake, but with additional benefits, such as being MRI‐free, reference region independent, or more robust to change in tracer.