Population PK/PD modelling of platelet dynamics for dose selection in patients with haematological malignancies
Tomás Sou (1), Christophe Meille (1), Sebastien Lorenzo (1), Romain Sechaud (1)
(1) Novartis, Basel, Switzerland
Introduction: Overexpression of murine double minute 2 (MDM2), a key negative regulator of the tumour suppressor protein p53, has been reported in a variety of cancers [1]. Siremadlin, a MDM2 inhibitor, is being investigated as a new treatment for acute myeloid leukaemia (AML). The effect of siremadlin in patients with solid tumours has been previously reported, with delayed thrombocytopaenia being the primary dose-limiting toxicity [2]. However, the effect of siremadlin in patients with haematological malignancies have not been fully evaluated. Specifically, the nature of the underlying disease may have different clinical manifestation of thrombocytopaenia than in solid tumour patients. A population PK/PD model characterising the relationship between the plasma pharmacokinetics (PK) of siremadlin and platelet counts in haematological patients can therefore be beneficial to support dose optimisation.
Objectives: This work aims to develop a population PK/PD model characterising the relationship between siremadlin plasma PK and platelet levels to support dose selection.
Methods: Plasma drug concentrations and platelet data were obtained from a phase I study on patients with p53 wild-type solid tumours and haematological malignancies following different dosing regimens. Plasma drug concentrations and platelet data were analysed and modelled using a population approach in Monolix 2021R2. Previously, a population PK model was developed using the plasma PK data in the trial. In the current analysis, the individual PK predictions from the population PK model were used to drive drug effect on platelets. The platelet model was a cell maturation model adapted from Friberg et al (2002) [3] and different drug effect functions were evaluated. Model selection was guided by mechanistic understanding of drug action, scientific plausibility of parameter estimates, parameter precision, the Akaike information criterion (AIC), and goodness-of-fit plots. With the selected model, a Shiny application [4,5] was developed to allow interactive exploration of different dosing scenarios. Selected dosing regimens from 10 to 40 mg QD for 5 days with different periods of drug holiday and number of cycles were evaluated. In the population simulations, platelet profiles from 1000 virtual subjects were generated to assess the risk of thrombocytopaenia resulting from the different dosing regimens.
Results: The plasma concentration-time profiles of siremadlin were well-described by a one-compartment disposition model with linear clearance (CL/F) and delayed absorption as described by a transit compartment model. Body weight was included as a covariate on volume of distribution (V/F) and the correlation between CL/F and V/F was considered. The population estimates (RSE%) of the PK model were: ktr = 6.28 h-1 (7.90%); MTT = 0.812 h (4.06%); ka = 3.89 h-1 (17.8%); V/F = 116 L (2.57%); CL/F = 5.9 L/h (4.20%); beta_V_tBWKG = 0.888 (9.23%); corr_V_Cl = 0.650 (7.85%). In the current analysis, drug effect on platelets, driven by drug concentrations in the central compartment, was described by a drug effect function potentiating cell apoptosis in the proliferating precursor compartment representing bone barrow cells, leading to a reduced number of matured cells available for development into circulating platelets. Baseline platelet count (PLTZ) was noticeably lower than the ones in solid tumour patients (PLTZ: 28.0 vs 241 G/L. The population estimates (RSE%) of the current PD model were: PLTZ = 28.0 (12.0%); MMT = 177 (13.3%); gamma = 0.145 (20.8%) and SLP = 0.000808 (15.2%). The model was able to simulate the delayed thrombocytopaenia resulting from different doses of siremadlin. The simulations showed that following a single cycle of treatment, platelet count decreased to the lowest level after approximately 15 days for a typical subject before a gradual recovery. In addition to the dose level, baseline platelet count was an influential factor impacting on the proportion of patients with severe thrombocytopaenia.
Conclusion: This work shows an application of a population PKPD model with a safety endpoint in view of supporting the selection of dosing regimens, considering platelet dynamics and thrombocytopenia as a clinical constraint. Model development is ongoing for the evaluation of disease effect in haematological malignancies.
References:
[1] Guerreiro N et al. 2021. “Translational modeling of anticancer efficacy to predict clinical outcomes in a first-in-human phase 1 study of MDM2 inhibitor HDM201.” AAPS J. 23(2):28
[2] Meille C et al. PAGE 27 (2018) Abstr 8612 [www.page-meeting.org/?abstract=8612]
[3] Friberg L et al. 2002. “Model of chemotherapy-induced myelosuppression with parameter consistency across drugs.” J. Clin. Oncol. 20:4713–4721.
[4] https://CRAN.R-project.org/package=shiny
[5] https://CRAN.R-project.org/package=mrgsolve
