Xuan Zhou1, Chao Chen2, Ole Graff3
1 Clinical Pharmacology Modeling and Simulation, GSK, Shanghai, China; 2 Clinical Pharmacology Modeling and Simulation, GSK, London, UK; 3 Discovery Medicine, GSK, Upper Providence, USA
Introduction
At the early stage of drug development, it’s important to raise a question: what proportion of patients can be expected to achieve adequate pharmacological response without unacceptable toxicity? The answer helps us quantify a compound’s pharmacology strength for supporting a progression decision, for assessing the rigor of clinical trials to test the compound and the target, and for differentiating candidate assets. We define the probability of pharmacological success (PoPS) as the probability that a desirable proportion of patients achieve the required level of pharmacological response. This probability is based on our understanding of the target and the compound, such as target dynamics, drug level at site of action and duration of action. We present a GSK funded case of using simulations to evaluate the PoPS of a therapy for treating a brain disease, where target X is present both inside and outside the brain, and two compounds A and B have different brain penetration. The therapeutic approach aims to reduce the target’s activity in the brain, while preserving some of its peripheral activity for safety.
Objectives
To understand 1) what proportion of patients can achieve the required pharmacological response, which is to normalize brain target activity and maintain 10% peripheral activity; 2) what PK or PD parameters have the most significant impact on this proportion; 3) what the trial success rate will be based on certain predefined criteria; 4) what the optimal dose will be for each compound; 5) what the adequate sample size will be in a study to observe the required pharmacological response; and 6) how do the two compounds compare in the above aspects.
Methods
Human pharmacokinetics for both compounds were predicted from non-clinical data. Assuming concentration fluctuation is low, we modelled the steady-state average free concentration as a simple Emax function for both central and peripheral effects without delay. The IC50 value derived in vitro from human cell lines were used, and the maximal inhibition was assumed to be 100%. A moderate 30% between-subject variability was assigned to the pharmacokinetic and pharmacodynamic parameters. Where there was parameter uncertainty, a Bayesian prior of uniform distribution over a credible range was applied. Subject-level data (N=1000) for 1000 trials were simulated taking into account parameter uncertainty (elevation folds of target activity Fe in patients and unbound brain/plasma concentration ratio kp,uu). Success criteria for a clinical trial were set as ≥80% subjects with required activities in the brain and in the periphery, while <5% subjects with undesirable peripheral inhibition. The percentage of the 1000 simulations which met the criteria was calculated as the success rate, or PoPS. The optimal doses for the two compounds were selected from the peak level of the success rate. We also tested the required sample size of a potential trial that can reflect the success rate for the overall population. All simulations were performed by using the Simulx function of the R package mlxR (Lixoft, Orsay, France).
Results
With the assumed parameter uncertainty, typically 83.7% patients given compound A can have sufficient brain response (target activity normalized to the level as healthy subjects’) without undesirable peripheral response (maintain at least 10% as in healthy subjects’), while the percentage for compound B is 78.4%. The most impactful parameter for this proportion and the subsequent overall PoPS is Kp,uu, and Fe has a major effect on the dose selection. The predicted population optimal dose for compound A was 32 mg/day with 72.4% maximal success rate, while the optimal dose for compound B is 10 mg/day but with only 35% maximum success rate. A sample size of at least 35 patients is required to reliably reflect the probably of success for the population. Conceivably, the trial success rate could be enhanced by individualized dose titration using a peripheral pharmacology biomarker.
Conclusion
This work helped us understand the strength of the compounds’ intrinsic pharmacological profiles and the pharmacological underpinning of the trial, as well as the impact of dose, patient population and sample size. It highlights the importance of the “PoPS” concept and illustrates the application of a simulation-based approach for its estimation.
Reference: PAGE 28 (2019) Abstr 8917 [www.page-meeting.org/?abstract=8917]
Poster: Drug/Disease Modelling - CNS