Presentation Abstract

Session: Imaging Posters
Saturday, Jul 10, 2010, 11:15 AM - 1:15 PM
Presentation: IC-P-119 - A model-based approach to improve intra-individual comparison of clustered amyloid-ß brain PET imaging data: Accounting for overlapping imaging time windows
Pres. Time: Saturday, Jul 10, 2010, 11:15 AM -12:15 PM
Location: 313-C
Category: +New imaging methods
Author(s): Florian Hiemeyer1, John P. Seibyl2, Michael Kunz1, Olivier Barret2, George Zubal2, Henryk Barthel3, Cornelia Reininger1, Osama Sabri3.
1Bayer Healthcare AG, Berlin, Germany, 2Instit Neurodeg Disorders, New Haven, CT, USA, 3University of Leipzig, Leipzig, Germany.
Abstract: Background:
Data used for the described models was generated from a non-randomized, global Phase 2 trial evaluating the efficacy of the ß-amyloid targeted PET tracer florbetaben. A total of 150 subjects (69 Healthy volunteers (HV) and 81 probable AD patients) were included in the study. Both visual and quantitative measurements were used to evaluate the efficacy of florbetaben in detecting ß-amyloid in the brain. The quantitative results consisted of decay corrected, standard uptake value (SUV) stemming from various cerebral volumes of interest (VOIs), for 3 different, but overlapping imaging time windows. All SUV were assessed using two different strategies of assigning the boundaries to the VOI. The purpose of the present analysis was to evaluate, whether the second VOI boundary strategy resulted in a higher mean separation in the SUV between AD and HV across different cortical regions.
The proposed analysis is a mixed effects ANOVA model that explains the clustered and repeated SUV measurements by study design (i.e. patient group, imaging window, VOI boundary method) as well as brain VOIs. As the 3 imaging windows overlapped (90 to 110 min., 100 to 120 min. and 110 to 130 min. p.i.), the induced correlation needed to be accounted for. This was done by specifying a sophisticated residual covariance matrix. The results of the ANOVA model were compared with the results of a simplified model which included only data from the 2 non-overlapping imaging windows.
The analysis model proposed above resulted in approximately 66% of the standard error of the estimates compared to the simplified model. There was no substantial change in the magnitude of the estimates. Due to the smaller standard error, precise estimates for the comparison of the two VOI boundary strategies for cortical VOI could be established.
The additional data included in the proposed analysis model is highly correlated with the data from the non-overlapping imaging windows. Nevertheless, we could show, that the proposed analysis model results in an evident decrease in the length of the confidence interval of about 33%, compared to the simplified model.
Disclosures:   F. Hiemeyer, Bayer Healthcare AG; J.P. Seibyl, Instit Neurodeg Disorders; Bayer Healthcare AG; Molecular NeuroImaging LLC; M. Kunz, Bayer Healthcare AG; O. Barret, Molecular NeuroImaging LLC; Instit Neurodeg Disorders, New Haven; Bayer Healthcare AG; G. Zubal, Molecular Neuroimaging LLC; Instit Neurodeg Disorders; Bayer Healthcare AG; H. Barthel, University of Leipzig; Bayer Healthcare AG; C. Reininger, Bayer Healthcare AG; O. Sabri, University of Leipzig; Bayer Healthcare AG.
EMBARGO: As a reminder: All abstracts accepted for presentation at ICAD are embargoed until the date and time of presentation at the conference, unless the Alzheimer's Association provides written notice of change of date and/or time in advance. Embargo is the prohibition for copyright reasons from releasing any content from a submitted or accepted abstract to the public until after it has been presented at ICAD. Prior to presentation at ICAD, authors may not release information to the news media or the public, nor publish the results in any manner, and must ensure that colleagues and other organizations or institutions are aware of this policy. Violation of the embargo may include retraction of the accepted abstract and/or loss of privileges of presenting research at ICAD in the future.

Technical Support
Phone: 217-398-1792