Amyloid-targeting therapies (ATTs) have been approved in the US for a few years, which means that patients receiving these treatments now exist outside of clinical trials, in the "real world," and are being included in observational studies of patients with mild cognitive impairment or mild dementia.
This is a great opportunity to study the effects of these drugs, but it also comes with the major challenge of correctly identifying treated patients. This is particularly difficult in multisite studies, where the clinicians who treat patients and manage their ATTs are not necessarily involved in the observational studies.
This challenge came up in the Longitudinal Early-Onset Alzheimer's Disease Study (LEADS). Our group at UCSF serves as the PET Core for the entire study, which means that we centrally and continuously perform quality control, processing, and quantification of PET scans acquired across all 18 sites. In doing so, we observed a few patients who appeared to show a dramatic decrease in amyloid PET signal over time, despite not being labeled as receiving ATTs. Communication with the clinical sites usually clarified that these patients were indeed being treated (with a couple of interesting exceptions described in the paper).
As a result, we decided to establish a standardized process that would raise a red flag whenever a patient showed an "unusual" decline in Centiloids. However, this required defining what constitutes a "normal" versus "unusual" decline in Centiloids, based on test-retest variability and measurement error.
This is exactly what we addressed in this paper, now published in Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring (DADM) - link here.
