[11C]flumazenil kinetics in the rat brain: model preference and the impact of non-specific and non-selective binding in reference region modeling

Aim: The analysis of [¹¹C]flumazenil pre-clinical studies currently follows the quantitative models validated in the clinical setting. Since tracer kinetics can differ between species, [¹¹C]flumazenil kinetic modeling was evaluated for the rat brain.
Material and Methods: 60min [¹¹C]flumazenil brain PET scans with arterial sampling were performed in two groups of male Wistar rats (n=10, tracer dose only; n=2, pre-saturated with 330nM of flumazenil). Time-activity curves (TACs) were generated for frontal cortex, hippocampus, cerebellum, medulla and pons (reference). Next, noiseless TACs (n=10) were simulated using an average input function and representative rate constants from animal data. A three-tissue compartment model including a compartment for non-selective binding was simulated with different levels of specific-binding (k₃ ranging from 0.2 to 2.6min^-1). For animal and simulated data, distribution volume (V_T) and distribution volume ratios (DVR) were calculated using one and two-tissue compartment models (1TCM and 2TCM) and spectral analysis (SA). Binding potential (BP_ND) was determined from full and simplified reference tissue models with one or two-tissue compartments for the reference (FRTM, SRTM and SRTM-2Ref). Parameter agreement was assessed by Spearman’s correlation and Bland-Altman plots. Akaike information criterion (AIC) was used to determine model preference.
Results: 1TCM and 2TCM V_T of regions with high specific-binding (frontal cortex, hippocampus and cerebellum) showed similar AIC (143±7 and 148±10 respectively), strong correlation (r_s=0.99) and good agreement (0.1% difference). In contrast, low specific-binding regions (pons and medulla) showed worse correlation (r_s=0.77) and agreement (17.6% difference), while AIC were lower for 2TCM (134±10) than for 1TCM (161±8). The pre-saturated group displayed similar results to those of low specific-binding regions. High levels of non-specific and/or non-selective binding (2TCM V_T=2.5±0.4) was observed in pons and affected BP_ND estimation by all tested reference tissue models. Simulations showed a similar pattern: 2TCM V_T demonstrated better agreement (<8.0% bias) than 1TCM for all levels of specific-binding, while 1TCM V_T demonstrated smaller bias with increasing specific-binding (from -55.1% to 3.1%). SA generated accurate V_T (<0.9% bias) for all specific-binding levels. 1TCM DVRs resulted in the largest BP_ND overestimation (up to 58% bias), while SRTM-2Ref showed an overall smaller bias (<1.5%).
Conclusion: [¹¹C]flumazenil kinetics in rats was substantially different from that in humans, especially in low specific-binding regions. In those, the 2TCM is preferred and the standard 1TCM and SRTM can lead to major errors in parameter estimation. Instead, models which can better account for non-specific and/or non-selective binding should be used, such as 2TCM or SA.