Introduction
Reference tissue (RT) and simplified reference tissue (SRT) models used in conjunction with [carbonyl-11C]WAY-100635 data underestimate the binding potential (BP) when compared to values obtained using kinetic modeling with arterial input function. This underestimation is due to violation of the assumptions underlying RT and SRT. The purpose of this study was to quantify by simulation the extent of BP bias in presence of one or more RT and SRT assumption violations.
Materials & Methods
Healthy brain time activities were simulated using a representative plasma input function for [carbonyl-11C]WAY-100635 and a two-tissue model (2TM) to obtain typical lateral orbital frontal cortex (Lof) and hippocampus (Hip) ROI, or one-tissue model (1TM) to obtain typical cerebellum (Cer - reference) ROIs. Since RT and SRT assume a negligible ROI blood volume (Vb=0) and equal distribution volumes of not specifically bound tracer between reference and ROI (Vf+ns(ROI)=Vf+ns(Cer)), to test the effect of these assumptions, Lof, Cer, and Hip activities were generated assuming Vb=0%, 5% or 10% and Vf+ns(ROI)=Vf+ns(Cer) or Vf+ns(ROI)Vf+ns(Cer): ±20%, 35%, & 45%. We also tested the assumption that the reference region is modeled with a 1TM, therefore a typical Cer activity was also generated by using a 2TM.
Results
The mean results are: if Cer is modeled with a 1TM model, Vf+ns(ROI)=Vf+ns(Cer), and Vb=0% parameters R1 and BP are not biased, while the other parameters (k2, k3, k4 for the RT and k2 for the SRT) are either under- or over-estimated. As soon as one of RT and SRT assumptions fails, BP is biased. For example, with Cer described by 1TM, Vf+ns(ROI)=Vf+ns(Cer) and Vb=5%, BP bias is ∼25–30% and BPRTBPSRT. If Cer is modeled by 2TM, Vf+ns(ROI)=Vf+ns(Cer) and Vb=5%, BP underestimation reaches 40% and BPRT=BPSRT. If Vb=0%, and Cer is modeled by 1TM, but Vf+ns(ROI)<Vf+ns(Cer), BP underestimation is of 25% when the difference between Vf+ns is 20%; underestimation increases when this difference is larger. If Vb=0%, and Vf+ns are equal but Cer is modeled by 2TM, there is a 20% BP underestimation. When all the three assumptions are violated, eg Vb=5%, Vf+ns(ROI)<0.35Vf+ns(Cer) and Cer described by 2TM, BP is underestimated of about 60%. Of note is that simulations always give BPRT=BPSRT whenever Cer is not modeled as postulated by RT and SRT model, i.e. by 1TM.
Conclusion
Simulations indicate that a correct estimation of BP with RT and SRT is possible if all the assumptions underlying these models are verified. As soon as one of them fails, BP is underestimated (at least 25%) and the other RT and SRT parameters are also biased. In addition, only if Cer is modeled with a 1TM then BPRTBPSRT. Since our previous RT and SRT analysis on real [carbonyl-11C]WAY-100635 data 1 showed BPRT=BPSRT, our findings support that Cer has a more than 1TM structure and that BP is heavily underestimated (∼50% by assuming a Vb=5%).