Evaluation of Darigabat QTc Prolongation in Healthy Volunteers: a Concentration-QTcF Analysis
Marije E. Otto [1,2], Koshar Safai Pour [1,3], Joop van Gerven [1,3], Gabriel Jacobs [1,3], Michiel van Esdonk [1], Gina Pastino [4], Jagan Parepally [4], Sridhar Duvvuri [4]
[1] Centre for Human Drug Research (CHDR), Leiden, The Netherlands [2] Leiden Academic Centre for Drug Research (LACDR), Leiden University, Leiden, The Netherlands [3] Department of Psychiatry, Leiden University Medical Centre (LUMC), Leiden, The Netherlands [4] Cerevel Therapeutics, LLC, Cambridge, MA, USA
Objectives: Darigabat (CVL-865; PF-06372865) is a novel, brain-penetrable positive allosteric modulator (PAM) of α2/3/5 subunit-containing gamma-amino-butyric-acid type A (GABAA) receptors, with minimal functional activity at α1 subunit-containing receptors, and is currently developed as potential treatment for focal onset epilepsy and anxiety-related disorders. Administered as single (0.04-100mg) and multiple (2.5-42.5mg BID) oral doses, darigabat demonstrated a favourable safety and pharmacodynamic profile compared to non-selective GABAA PAMs such as benzodiazepines[1-3]. No clinically significant dose-dependent prolongation of QTcF or change in heart rate were observed with up to 100mg single doses of darigabat. To further evaluate the cardiodynamic effects of darigabat, a concentration-QTcF (conc-QTcF) analysis was performed using the single dose clinical trial data in healthy volunteers.
Methods: Data from a randomized, double-blind, placebo-controlled, cross-over study in healthy volunteers were used for conc-QTcF analysis[2]. Subjects participated in 5 treatment periods (4 active, 1 placebo) with a minimal 7 days washout period. Cohorts 1-3 included 10 subjects each and received single oral doses of 0.04-100mg of darigabat; Cohort 4 included 15 subjects and received two single oral doses of darigabat (15mg or 65mg), 2mg lorazepam, or 65mg darigabat combined with 2mg lorazepam. Pooled ECG and PK data from the 6.0-100mg (solely) darigabat dose levels was included for the analysis, along with placebo data of included subjects. ECG measurements were performed in triplicate and individual mean derived QTcF values were matched to the time of PK sampling (baseline [-2h pre-dose], 1h, 2h, 4h, 8h,12h, 24h and 48h post-dose). The cross-over design allowed for placebo correction on an individual level. In total, 639 matched placebo- and baseline-corrected (ΔΔ) QTcF observations from 43 subjects were available for analysis.
The pre-specified model for conc-QTcF analysis, as described by Garnett et al. (2018)[4], was applied. Model assumptions on heart rate effects, hysteresis and linearity were explored. Both linear and saturable effect (Emax) relationships, with different variance structures, were investigated. The developed model was evaluated with goodness-of-fit figures and a confidence interval visual predictive check. The mean and two-sided 90% confidence interval (CI) of the simulated ΔΔQTcF at the (geometric) mean maximal concentration (Cmax) for each cohort were calculated.
NONMEM(7.5) was used for nonlinear mixed-effects modelling and R(4.0.3) for data management and visualization[5,6].
Results: Data exploration showed that model assumptions regarding no effect of darigabat on heart rate and no presence of hysteresis were met. The baseline correction in the pre-specified linear model describing the conc-ΔΔQTcF relationship of darigabat introduced a bias in predicted versus observed concentrations and was therefore removed. Still, use of a linear relationship resulted in overprediction of observations in the higher concentration range and thus development continued with a non-linear Emax model. Inclusion of inter-individual and inter-occasion variability on the intercept significantly improved the model fit.
Darigabat showed an effect on the QTcF interval, with an estimated Emax parameter of 7.43ms (95%CI: 0.539-14.3, p=0.017) and EC50 of 159ng/mL (95%CI: 21.1-1199.9, p<0.0001). The predicted ΔΔQTcF at the geometric mean Cmax (559.3ng/mL) for the highest dose level of 100 mg was 4.33ms with an upper limit of the 90%CI of 7.54ms, which is below the 10ms threshold necessary for regulatory significance. The darigabat concentration at which the mean ΔΔQTcF crossed the 5ms level and the upper limit of the 90%CI crossed the 10ms threshold was estimated at 946ng/mL and 2062ng/mL, respectively, providing a 3.7-fold safety margin over the exposure achieved at the therapeutic dose (25 mg BID, Cmax=253 ng/mL)[1].
A significant linear relationship between observed QTcF and RR interval in the active treatment group was found. Based on the overlap between high RR values and Tmax, this might result in positively biased ΔΔQTcF values for observations in the higher concentration range. Therefore, analysis outcomes may have an increased false positive risk of QTcF prolongation.
Conclusions: The darigabat conc-QTcF effect relationship in healthy volunteers was optimally described by a non-linear Emax model. Simulations with this model preclude significant QTc interval prolongation at clinically relevant darigabat plasma concentrations in future clinical studies.
References:
[1] Gurrell, R., Whitlock, M., Wei, H., Shen, Z. & Ogden, A. Safety, Tolerability, and Pharmacokinetics of Multiple Repeated Oral Doses of the α2/3/5-Subtype Selective GABAA-Positive Allosteric Modulator PF-06372865 in Healthy Volunteers. Clin. Pharmacol. Drug Dev. 10, 756–764 (2021).
[2] Nickolls, S. A. et al. Pharmacology in translation: the preclinical and early clinical profile of the novel α2/3 functionally selective GABAA receptor positive allosteric modulator PF-06372865. Br. J. Pharmacol. 175, 708–725 (2018).
[3] Cerevel Therapeutics. [Press Release Details] Cerevel Therapeutics Announces Positive Topline Results for Darigabat in Phase 1 Clinical Trial in Acute Anxiety. https://investors.cerevel.com/news-releases/news-release-details/cerevel-therapeutics-announces-positive-topline-results (2022).
[4] Garnett, C. et al. Scientific white paper on concentration-QTc modeling. J. Pharmacokinet. Pharmacodyn. 45, 383–397 (2018).
[5] Beal, S. L., Sheiner, L. B. & Boeckman, A. J. NONMEM 7.5.0 User Guides. (1989-2020). ICON Dev Solut Hanover, MD.
[6] R Core Team. R: A language and environment for statistical computing. R Found. Stat. Comput. Vienna, Austria (2020).