Harmonization of tau‐PET in Alzheimer’s disease: comparison of methods to derive CenTauR units for [18F]RO948, [18F]Flortaucipir, and [18F]MK‐6240
Antoine Leuzy, Vincent Dore, Lars Lau Raket, Gregory Klein, Suzanne L. Baker, Maria C. Carrillo, Arnaud Charil, Emily C. Collins, Jessica A. Collins, Samantha Budd Haeberlein, Makoto Higuchi, Eric D. Hostetler, R. Matthew Hutchison, Leonardo Iaccarino, Michael C. Irizarry, William J. Jagust, Keith A. Johnson, Yashmin Karten, Hartmuth C. Kolb, Brian J. Lopresti, Sulantha Mathotaarachchi, Mark A. Mintun, Rik Ossenkoppele, Ioannis Pappas, Tharick A. Pascoal, Michael J. Pontecorvo, Gil D. Rabinovici, Pedro Rosa‐Neto, Ziad S. Saad, Sandra Sanabria Bohorquez, Andrew W. Stephens, Sudhir Sivakumaran, Matteo Tonietto, Ruben Smith, Christopher C Rowe, Victor L Villemagne, Oskar Hansson- Psychiatry and Mental health
- Cellular and Molecular Neuroscience
- Geriatrics and Gerontology
- Neurology (clinical)
- Developmental Neuroscience
- Health Policy
- Epidemiology
Abstract
Background
Tau positron emission tomography (PET) is increasingly used in the clinical evaluation of patients and as an outcome measure in Alzheimer’s disease (AD) clinical trials. Due to differences in tracer properties, instrumentation, and methods of analysis, however, tau‐PET outcome data cannot currently be meaningfully compared or combined. Here, we tested i) the feasibility of adapting the Centiloid method—an approach originally developed to standardize amyloid PET—to harmonize tau‐PET quantification (CenTauRs); ii) the performance of a non‐linear mixed model‐based approach (Joint Propagation Model) that does not require the use of a reference tracer.
Method
Head‐to‐head tau‐PET data (Table 1) was included from two cohorts ([18F]RO948 vs [18F]flortaucipir, n = 37 [BioFINDER‐2]; [18F]flortaucipir [Avid A05] vs [18F]MK‐6240, n = 15, University of Pittsburgh) in which each participant was scanned with two tau‐PET tracers. Standardized uptake value ratio (SUVR) values were calculated using the inferior cerebellar cortex as the reference region. Anchor point data (Table 1) was derived for each tracer using the following criteria: CenTauR‐0, cognitively unimpaired (CU), amyloid PET negative (<10 Centiloids); CenTauR‐100, amyloid PET positive (>50 Centiloids), typical (temporoparietal) AD pattern on tau‐PET visual read, age<65 andMMSE>20. Regions‐of‐interest (ROIs) included a universal tau‐PET ROI—based on the intersection of tracer specific ([18F]flortaucipir, [18F]MK‐6240, [18F]PI‐2620, [18F]PM‐PBB3, [18F]GTP1 and [18F]RO948) masks that had been derived by subtracting average of amyloid‐negative CU images from the average AD image—as well as four subregions delineated within this ROI (medial temporal, meta‐temporal, temporoparietal and frontal) (Figure 1A). An overview of the adapted Centiloid‐like approach and the joint propagation model is shown in Figure 1B.
Result
High R2 values were observed between tracers across all ROIs: [18F]RO948 vs [18F]flortaucipir, average 0.965 [range 0.923 (frontal) to 0.986 (universal)]; [18F]MK‐6240 vs [18F]flortaucipir, average 0.985 [range, 0.923 (medial temporal) to 0.991 (frontal)]. The Centiloid‐like and joint propagation model approaches provided near identical CenTauR values (Figure 2).
Conclusion
Preliminary findings support the development of standardized scale for tau‐PET using both a Centiloid‐like or joint propagation model approach. Additional data and consideration of the advantages and disadvantages of each will be needed to recommend one approach over the other.