Longitudinal [18F]flortaucipir PET: comparison of quantitative and semi-quantitative parameters

Abstract

1597 Introduction: In vivo [18F]flortaucipir PET scans can be used to assess changes in tau pathology over time. Most studies use semi-quantitative measures derived from static images for this purpose due to practical applicability and computational simplicity. However, dynamic imaging allows for more accurate quantification of (changes in) specific binding. Quantification is especially important in longitudinal PET imaging in Alzheimer’s disease (AD) because disease progression or drug intervention may lead to flow changes over time, which have to be taken into account when measuring changes in specific binding (van Berckel et al. 2013). We aimed to compare semi-quantitative (SUVr) and quantitative (BPND) parameters for [18F]flortaucipir PET in a longitudinal study. Methods: We includedsubjects with subjective cognitive decline (SCD; n=38) and AD patients (n=24). All subjects underwent baseline (BL) and 2-year follow-up (FU) dynamic [18F]flortaucipir PET. BPND and R1 (relative cerebral blood flow) were calculated using receptor parametric mapping (RPM) with cerebellar gray matter as reference region. In addition, SUVr was derived using 80-100 min p.i.. Regional SUVr values were compared against respective DVR (BPND+1) values in all analyses. (Trans)entorhinal, limbic and neocortical regions-of-interest (ROIs) were assessed. For each parameter % change was calculated based on the following formula: 100*(FU-BL)/BL. Paired T-tests in R software were used to evaluate the significant differences for all the comparisons in this study. Additionally, simulations were performed to model the effects of flow changes on BPND and SUVr. For this end, we classified three categories: high, medium or low binding and for each category, R1 values were varied to mimic 5 to 25 % increase or decrease of flow. Thereafter, %bias for SUVr with respect to the underlying binding were calculated and the association of this %bias with respect to the change in flow was also evaluated. Results: In SCD subjects, there was a significant difference between %change in DVR (2.56%) and SUVr (1.85%) in the (trans)entorhinal ROI (pl p&lt0.05), but not the (trans)entorhinal (DVR 3.48% vs. SUVr 3.25%) or neocortical (DVR 7.25% vs. SUVr 8.21%) ROIs. R1 increased on average with 0.7% in SCD, and decreased with 1.6% in AD. Simulations showed varying %bias in SUVr depending on the underlying binding. With increasing flow a decreased bias in low binding condition was observed, whereas in high binding condition an increased bias was seen. DVR remained relatively stable under different flow conditions, irrespective of the underlying binding. Conclusion: There were minor differences between BPND and SUVr in these cohorts, that were disease stage dependent. Overall, compared to BPND, SUVr provided a relatively good estimate of specific binding for [18F]flortaucipir over a two-year follow-up. However, simulations showed that [18F]flortaucipir SUVr is subject to variable bias in case of larger flow changes. Moreover, the effect of flow changes on SUVr seemed to be depending on the underlying tau load. As such, BPND should be preferred over SUVr, especially in the context of pharmacotherapeutic interventions that could induce significant changes in blood flow.