Anne Keunecke 1, Lars van der Veen 2, Paramjit Kaur 2, Peter Vis 1
1 LAP&P Consultants BV (Leiden, The Netherlands), 2 iOnctura SA (Genève, Switzerland)
Introduction and Objectives: The highly selective, oral allosteric modulator of PI3Kδ, roginolisib (IOA-244), blocks the activity of PI3Kδ-dependent signalling, in both tumor cells and Tregs. The activation of basophils (CD63+) is PI3Kδ-dependent, which makes CD63+ basophils a potential surrogate to characterise the inhibitory effect of roginolisib on PI3Kδ in tumor cells.
Here, we discuss the development of a population PK (popPK) for roginolisib in healthy volunteers and the influence of food, sex and alpha1-acid glycoprotein (AAG) binding on clearance exposure in healthy volunteers and its implication when predicting patient studies. Additionally, we describe the importance of appropriate data transformation when developing a sequential PKPD model to describe the drug effect on CD63+ basophils (%) as a potential surrogate biomarker of PI3Kδ inhibition.
Methods: Plasma concentration–time data from healthy volunteers were analysed using nonlinear mixed-effects modelling in NONMEM. A two-compartment model with linear elimination was developed to describe roginolisib pharmacokinetics (PK). Absorption was characterised using three transit compartments, with mean transit time (MTT) dependent on meal status. Alpha-1-acid glycoprotein (AAG) concentrations were incorporated as a time-varying covariate to account for protein binding effects on exposure.
Individual post hoc PK parameters from the final population PK model were used to drive a sequential PK/PD analysis. A direct-response model was developed to describe the effect of roginolisib on CD63⁺ basophils (%). To ensure model predictions remained within the physiological range (0–100%), the CD63⁺ basophil data were transformed to the logit scale prior to modelling.
Results: Roginolisib PK in healthy volunteers was adequately described by a two-compartment model with linear elimination, incorporating AAG concentrations as a time-varying covariate. Higher AAG concentrations were associated with a lower free fraction of roginolisib and, consequently, reduced clearance. Absorption was characterised by three transit compartments, with a longer mean transit time in the fed compared with the fasted state. A slightly lower relative bioavailability was estimated at the 80-mg dose compared with 40 mg, but only under fasted conditions. Inter-individual variability was estimated on clearance, relative bioavailability, and MTT.
Covariates including age, AAG, sex, and body weight were evaluated. Sex, AAG, and body weight were correlated; women appeared to have higher clearance than men.
To investigate exposure differences between volunteers and patients where AAG was not determined, the fold change in median AAG concentration was determined through empirical Bayes estimation. The AAG concentrations determined this way in patients ranged from 0.5 to 3.5 g/L and is consistent with reported values in oncology populations[1]. This finding supports the hypothesis that the increased exposure in patients can be explained by potentially higher AAG concentrations.
CD63⁺ basophil data were described using a direct-response model on the logit scale, with a linear drug effect. The model provided a satisfactory fit. Inter-individual variability was estimated on both baseline CD63⁺ basophils and the drug effect slope, with parameters estimated with good precision. The typical baseline CD63⁺ level was 77% and the slope of the drug-effect was 0.67 mL/μg on logit scale.
Replacing total roginolisib concentrations with model-derived free concentrations did not significantly improve the fit. No trends were observed between individual effects and dose, food status, sex, body weight, age, or AAG concentration.
Conclusions: Roginolisib PK in healthy volunteers were well described by a two-compartment model incorporating time-varying AAG concentrations, highlighting the importance of drug binding to this protein in explaining variability in total exposure. Higher AAG levels were associated with reduced free fraction and lower apparent clearance, providing a plausible explanation for the higher total exposure observed in patient populations.
The sequential PK/PD analysis showed that inhibition of CD63⁺ activation could be adequately described by a direct-response model on the logit scale, with a linear drug effect and good parameter precision. Use of free drug concentrations did not significantly improve the PD model, suggesting that total concentrations were sufficient to characterise the observed biomarker response in this setting.
Overall, the developed PK/PD framework provides understanding of the impact of AAG-mediated protein binding on roginolisib exposure and supports the use of CD63⁺ basophil data as a surrogate marker of PI3Kδ target occupancy in early clinical development.
References:
[1] Kenmotsu H et al, Br J Clin Pharmacol (2017) 83 2416–2425
Reference: PAGE 34 (2026) Abstr 12004 [www.page-meeting.org/?abstract=12004]
Poster: Drug/Disease Modelling - Oncology