Andrés Olivares-Morales, Pilar Garcés, Joerg Hipp, Jeannine Petrig Schaffland, Giuseppe Cecere, Damien Docquir, Martin Kapps, Lorraine Murtagh and Maria-Clemencia Hernandez
Roche Pharma Research and Early Development (pRED), Roche Innovation Center Basel.
Objectives: RO1 is a small molecule, selective GABAA α5 receptor positive allosteric modulator under development for the treatment Neuro-psychiatric disorders. Quantitative electroencephalography (qEEG) has previously shown utility as a biomarker for related compounds (Benzodiazepines). Here we investigate the pharmacokinetics (PK) and pharmacodynamics (PD) relationship of the qEEG modulation by RO1 and diazepam (DZP) in Wistar Rats using a previously published mechanism-based PK/PD model [1] in the context of a Bayesian framework.
Methods: Rats (n=12 per cohort) were surgically implanted with EEG electrodes 2 weeks prior to the compound administration. RO1 (10 mg/kg), DZP (2 mg/kg) and vehicle were administered intravenously as five-minute infusions. Blood samples were obtained from all the animals and plasma samples were analyzed for RO1 and DZP with a qualified analytical method using LC-MS/MS. In addition, EEG was recorded after drug administration over a period of 17 h in the freely moving rats during dark cycles. The PK and PD data (fronto-central wake EEG beta-band amplitude, at 20 – 30 Hz) for each treatment were analyzed in NONMEM 7.3 using a Bayesian approach. Compartmental models were fit to the PK data, whereas the PD data was fitted to a mechanism-based model previously described for other GABAA modulators by Visser et al [1]. For the Bayesian analysis, informative priors were derived from exploratory analysis of the PK (RO1 and DZP) and PD (vehicle) data. Additional priors were derived from the work of Visser et al. [1] for the drug-independent elements of the PK/PD model (i.e., parameters A, b and d). Markov chain Monte Carlo (MCMC) analysis was conducted by initiating two chains for each compound (1×106 iterations each). The first part of each chain was discarded as a burn-in phase while the second part was kept and split into two. To assess the model convergence, multi-chain diagnostics (total 4 chains) were used as described by Gelman and Rubin [2]. Visual inspection of the MCMC chains and calculation of the potential scale reduction factor (Rhat) and effective sample size (Neff) for each parameter was implemented in R version 3.4.4 (https://cran.r-project.org/)using the latest versions of the “coda” and “MCMCpack” packages.
Results: In total 11 and 9 animals had evaluable PK samples for RO1 and DZP, respectively, whereas clean EEG recordings were obtained from 9, 7 and 10 rats for RO1, DZP and vehicle, respectively. The PK of RO1 and DZP were best described by a 3-compartment model, whereas the baseline qEEG data of the vehicle was best described by a cosine function. Using the Bayesian approach and prior information, all the PK and PD parameters models converged successfully (Rhat <1.1) and the posterior distributions, means and credible intervals (CI) were within acceptable limits. In addition, the mechanism-based PK/PD model provided reliable estimates for both RO1 and DZP’s potency and efficacy (Kpd and Epd, respectively). In addition, the estimates for DZP (Kpd =22 [CI: 18- 27] ng/mL and Epd = 0.49 [CI: 0.44 – 0.54]) were in line with the previously reported values for DZP in rats [1].
Conclusions: A PK/PD relationship for qEEG modulation by RO1 and DZP was established in rats using a Bayesian approach, allowing the fit of complex models to limited data with the use of prior information. This approach facilitated a full PK/PD modelling without the need for a two stage PK/PD assessments, where parameters are generally fixed. Finally, this study contributes towards the establishment of qEEG as a translational biomarker for use in the potential clinical development of RO1.
References:
[1] Visser et al. JPET. 2003 (doi: 10.1124/jpet.102.042341)
[2] Gelman, A. and D. B. Rubin (1996). Statistical Methods in Medical Research 5(4): 339-355.
Reference: PAGE 28 (2019) Abstr 9057 [www.page-meeting.org/?abstract=9057]
Poster: Drug/Disease Modelling - CNS