Quantitative amyloid‐PET measures in a memory clinic population

Abstract

AbstractBackgroundIn clinical practice, visual assessment of amyloid‐PET images results in a binary classification (negative or positive) of amyloid status. While this approach has shown high clinical utility, the expected approval of disease‐modifying treatments warrants the need for objective amyloid burden quantification, to further support a high‐confidence diagnosis of Alzheimer’s disease (AD) and treatment decisions. We investigated two widely available quantification methods, i.e. Centiloid (CL) and z‐scores to optimize clinical use of amyloid‐PET.MethodsPatients with subjective cognitive decline plus (SCD+; N=220), mild cognitive impairment (MCI; N=293) or dementia (N=216) enrolled in the AMYPAD‐DPMS that have undergone amyloid‐PET imaging and passed quality control were included in the study. Acquired static (i.e. 90‐110 min. post‐injection) [18F]flutemetamol (N=380) or [18F]florbetaben (N=349) PET images were processed using GE Healthcare’s AMYPYPE PET‐only pipeline, providing global CL, z‐scores, and z‐scores for 7 cortical ROIs. Local readers provided visual read (VR) status and regional assessments.ResultsDemographics are shown in Table 1. In total, 364 (49.9%) patients were assessed as amyloid VR‐positive and showed significantly higher CL (U=3954, p<0.001) and global z‐score (U=4197, p<0.001) values across clinical stages compared to VR‐negative patients (Figure 1A). The proportion of VR‐positive patients increased with clinical stage (SCD+: 30.0%, MCI: 48.8%, dementia: 71.8%) and global amyloid burden (CL: H=24.97, p<0.001; z‐score: H=24.43, p<0.001) was also associated with disease severity (Figure 1B). VR‐positive patients with a primary etiological diagnosis of AD showed higher overall higher amyloid burden (CL: H=6.93, p=0.008; z‐score: H=5.54, p=0.019) and specifically in the pre‐frontal cortex (H=8.18, p=0.004) compared to patients with a non‐AD primary etiological diagnosis, though percentage of frontal VR‐positivity was highly similar between groups (AD: 97.2% vs. non‐AD: 96.6%). Global CL and z‐scores were highly correlated (ρ=.987, p<0.001), by a factor of 10 (Figure 1C).ConclusionOur results demonstrate high agreement between CL and z‐scoring quantitative measures in a clinical population. While initial results do not indicate differences in patterns of regional positivity between AD and non‐AD subjects, differences in global amyloid burden seem to be driven by known AD‐related early accumulating regions. Additional analyses comparing other etiological groups will be presented at the conference.