Computerized decision support to select memory clinic patients for amyloid PET: Which patient to test?: Neuroimaging / differential diagnosis. (7th December 2020)

• Record Type: Journal Article
• Authors: Rhodius‐Meester, Hanneke FM; van Maurik, Ingrid S.; Koikkalainen, Juha; Tolonen, Antti; Frederiksen, Kristian Steen; Hasselbalch, Steen Gregers; Soininen, Hilkka; Herukka, Sanna‐Kaisa; Remes, Anne; Barkhof, Frederik; Van Berckel, Bart N.M.; Scheltens, Philip; Lötjönen, Jyrki; van Der Flier, Wiesje
• Is Part Of: Alzheimer's & dementia. Volume 16(2020)Supplement 5
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-12-07
• 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.042687
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: The current diagnostic criteria allow the use of amyloid biomarkers (either CSF or PET) to provide pathophysiological support for the diagnosis of AD. How these criteria should be operationalized by clinicians is unclear. We previously developed a model to select patients in which CSF measurements have added value [1]. Now, we study whether the model can also be used to select patients for amyloid PET. Method: We included 348 subjects (223 controls, 125 Alzheimer's disease dementia) from ADNI2, with available neuropsychology, MRI and Florbetapir PET. Positive (AD like) and negative (normal) amyloid PET were simulated to estimate whether knowledge of amyloid status would impact (confidence in) diagnosis. These simulated markers were added to available demographic, neuropsychology and MRI data in each patient to see whether the diagnostic confidence increased. If the confidence exceeded a predefined cut‐off (probability of correct class (PCC)≥0.80), then the patient was selected for amyloid PET (computerized decision support approach). We compared this approach with proportion of PET scans and patients diagnosed with sufficient confidence (PCC ≥0.80) based on an algorithm without amyloid PET (only neuropsychology and MRI), and amyloid PET for all patients (see Figure 1). Result: The computerized decision support approach recommended amyloid PET in 11%, which yielded a diagnosis with sufficient confidence in 75% of all patients. This approach was most efficient when compared to amyloid PET in none (0% amyloid, 71% diagnosed), or amyloid PET for all (100% amyloid, 71% diagnosed). Results were comparable for CSF instead of amyloid PET. Conclusion: The previous developed computerized decision support approach showed favorable results in selecting patients for amyloid PET testing. In this clear‐cut comparison between controls and AD, amyloid PET adds little over clinical measures – but a similar framework may help to support decision making in more realistic and challenging clinical situations. Expansion of this study is needed with a real life cohort consisting of patients with different types of dementia. In this way, the approach can support clinicians in making a balanced decision in ordering additional biomarker testing. Reference: (1) Rhodius‐Meester HFM et al. Selection of memory clinic patients for CSF, PLoS One. 2020;15(1).