Quantitative PET imaging of radioligands with slow kinetics in human brain

Abstract

Dear Sir, We read with interest the commentary by Hirvonen et al. [1] on the quantification of radioligands with slow kinetics. Their critique of our work on the quantification of the type 1 cannabinoid receptor (CB1R) ligand [F] MK-9470 in the human brain [2] was of particular interest and triggered our response to support the methods that were used in the analysis of the behavior of this novel radioligand. Here we specifically address the vulnerabilities and pitfalls highlighted by Hirvonen et al. and argue that our assumptions can be reasonably justified even in the absence of a “gold standard” for radioligands with slow tissue kinetics. The general approach for analyzing neuroreceptor PET studies is based on compartmental modeling approaches formulated by Mintun et al. [3]. Even though the mathematical formulation of this model relies on several assumptions (e.g., rate constants are not changing during the PET study and the tracer concentration in the compartments is homogeneous), brain PET imaging analyzed with this approach has been a powerful tool that has enabled the study of neuroreceptor systems in normal and pathological conditions as well as supporting drug development. The compartmental model formulation is such that the parameters (K1; k2; k3 1⁄4 fNDkon Bavail CS ð Þ and k4=koff, where CS is the tracer concentration bound to the receptors and fND is the tracer free fraction in tissue [4]) could be related to the biochemical and physiological characteristics of the tracer and the receptor system. The estimation of all five parameters (rate constants and Bavail) is however only possible when using multiple injection protocols [5–8]. Another approach is to perform a series of studies at different specific activities [9–11] and use an in vivo Scatchard analysis to estimate Bavail and Kd. It should be noted that very often, when performing single bolus or bolus plus constant infusion (B/I) tracer administration experiments, the individual model parameters cannot be identified (even when the number of parameters is reduced by using high specific activity tracer, so that kon Bavail CS ð Þ ! HighSA konBavail). The reason that it is not possible to find a unique solution to the individual rate constants is because of the large correlation between parameters and the large coefficients of variation of the estimates. In general, the macro parameters (volume of distribution, binding potential, irreversible uptake rate constant) obtained from the combination of individual rate constants are more stable. As a consequence caution is required when interpreting any estimates of individual rate constants particularly when derived from single tracer injection experiments. Hirvonen et al. [1] asserted that our assumption that k4 is uniform across brain regions is invalid and that k4 differs across brain regions in correlation with receptor density. In support of their argument they discussed [I]iomazenil, a SPECT ligand for benzodiazepine receptors that are widely distributed at high density throughout the brain. In nonhuman primates for [I]iomazenil, the in vivo kon and koff (estimated using a combination of parameters from in vivo Scatchard analysis [11] and from single bolus tracer studies [12]) were much slower than the in vitro measurements, whereas the ratio of the in vivo rates kon/koff was similar to the in vitro measured Kd values. S. M. Sanabria-Bohorquez (*) Imaging, Merck Research Laboratories, Sumneytown Pike WP44D-2, West Point, PA 19486, USA e-mail: sandra_sanabria@merck.com