Translational modelling and estimation of probability of molecule success on balance of pharmacology benefit and safety risk
Silvia Maria Lavezzi (1), Chao Chen (2), John Upson (2), John Toomey (2), Kevin French (2), Liangfu Chen (2), Alistair Lindsay (2)
(1) Parexel International; (2) GlaxoSmithKline R&D
Objectives: Recent drug industry statistics show a high failure rate of clinical proof of concept, suggesting an inadequate benefit-risk ratio at this critical stage for many drug candidates [1,2]. Therefore, an early understanding of a molecule’s therapeutic potential is important for a sound decision on whether to progress it to the clinic or to redirect the investment. In the absence of direct evidence of clinical efficacy and safety, the estimation of probability of pharmacological success (PoPS) has been proposed as a framework for early prediction of the benefit-risk ratio [3,4]. Based on translated non-clinical data, PoPS quantifies the overall pharmacological strength and safety risk of a molecule for its therapeutic goal. Here, we present an illustrative example to outline the PoPS approach based on translational pharmacology modelling.
Methods: The PoPS for a molecule approaching the first human trial was defined as the probability of having appropriate proportion of the population achieving adequate pharmacology and not exceeding a safety limit.
The pharmacology endpoint was a blood biomarker, reflecting target engagement (proximal pharmacology). The concentration-response data from an ex vivo assay were analysed using a transduction model [5] to estimate the production and decay rates of the biomarker, as well as the pharmacokinetics-pharmacodynamics (PK-PD) of the drug effect for enhancing the biomarker production. The ex vivo biomarker PK-PD was translated into a dose-response relationship in human, by combining it with the predicted in vivo human PK, derived via allometry from three preclinical species.
The in vivo PK-PD relationship for a surrogate efficacy endpoint (distal pharmacology) was modelled from a preclinical efficacy model, to identify the efficacious range of the biomarker response.
The safety endpoint was the molecule’s total plasma exposure at the doses without observed adverse effects in three preclinical species.
Results: An Emax model successfully characterised the ex vivo drug effect on biomarker production within the transduction model. The PK in human was predicted as a 2-compartment linear model with first-order absorption. The in vivo efficacy was described by an Emax function driven by drug concentration, and the predicted efficacious exposure range corresponded to ~2-10 fold change in biomarker.
The PoPS was specified as the probability that ≥90% of patients achieve adequate pharmacology, and ≤5% exceed the safety limit. The biomarker response and the plasma exposure were simulated over a wide dose range for 1000 trials, each with 1000 patients per dose level, accounting for PK-PD parameter and translational uncertainties and for between-subject variability for key PK-PD parameters. Each trial was assessed for success according to the pharmacology and safety criteria, simulated based on respective priors. Adequate pharmacology was defined based on the efficacious range of the biomarker response (2-10 fold change), using a beta distribution prior. The safety limit was the exposure at the no-observed-adverse-effect doses, with prior weighted among the three species (60:30:10) based on the knowledge about species relevance at the time. The PoPS (the proportion of trials achieving success) was estimated as 62%.
Subsequent in vivo and in vitro toxicology assessments combined with consultation with experts concluded that dose-limiting safety findings in two out of three preclinical species were unlikely to translate to humans at the proposed clinical dose range. Based on this, using the safety limit from the third species only, the PoPS was re-estimated in various scenarios of varying pharmacology requirement and variability. In all scenarios, PoPS was ≥90%.
Conclusions: The PoPS has been proposed as a tool for evidence-based decision making, by considering both the molecule’s pharmacological properties and, importantly, the intervention needs for the clinical objective [4]. As such, it conceivably leads to more meaningful and credible therapeutic dose prediction. This work illustrates the general principles for situational application of the framework. Through effective integration of multi-source data and transparent description of key assumptions, the PoPS captures multiple uncertainties in a single probability term. This methodology has the potential to enhance the confidence and clarity of investment decisions for drug candidates.
References: [1] DiMasia JA, Grabowski HG, Hansen RW. Innovation in the pharmaceutical industry: new estimates of R&D costs. J Health Econ 2016; 47:20–33.
[2] Hay M, Thomas DW, Craighead JL, Economides C, Rosenthal J. Clinical development success rates for investigational drugs. Nat Biotechnol 2014; 32:40–51.
[3] Chen C. Opportunities and pitfalls in clinical proof-of-concept: principles and examples. Drug Discov Today 2018; 23:776-787.
[4] Zhou X, Graff O, Chen C. Quantifying the probability of pharmacological success to inform compound progression decisions. PLoS One 2020; 15:e0240234.
[5] Mager DE, Jusko W J. Pharmacodynamic modeling of time-dependent transduction systems. Clinical Pharmacology & Therapeutics 2001; 70:210-216.
