Parameterisation of biomarker response in the assessment of long term safety
Tarjinder Sahota (1), Ian Sanderson (1), Meindert Danhof (1), Oscar Della-Pasqua (1,2)
(1) Division of Pharmacology, LACDR, Leiden University, Leiden, The Netherlands; (2) Clinical Pharmacology Modelling & Simulation, GlaxoSmithKline, Stockley Park, UK
Objectives: Despite major regulatory concern regarding long term safety, traditional methods are still applied to the evaluation of the so-called safe drug exposure, often expressed as AUC and CMAX. Given current understanding of PK, PD and disease processes, these parameters are unlikely to be predictive of long term toxicity.
We use naproxen as a paradigm compound to assess the ability of current preclinical experimental protocols to provide necessary information for predicting drug-induced effects in humans using nonlinear mixed effects modelling (NLME). We determine the feasibility of parameterising drug exposure in terms of biomarkers (prostaglandin E2 (PGE2) and thromboxane B2 (TXB2)) for prediction of target related toxicity in rodents (ulcer formation). We then show that extrapolations based on these biomarkers can be used to predict treatment effects in humans.
Methods: A general toxicity study protocol was conducted in rodents according to standard practice with the exception of biomarker samples being collected at PK sampling times. PKPD modelling was performed in NONMEM VII. Model validation included VPCs, NPCs, and NPDE analysis. Translation to drug effects in humans took account of expected differences in the therapeutic exposure in patients. The accuracy and precision of predictions were assessed using literature data in humans. Measures of exposure (dose strength, AUCτ, CMAXτ, cumulative AUC, area under % inhibition curve (AUICτ), maximum inhibition IMAXτ, cumulative AUIC) were tested for their ability to predict gastric ulceration in rodents as measured by % surface area of the stomach. Measures were compared for predictive performance with ulceration by modelling the exposure-risk relationship in NONMEM and comparing model diagnostics.
Results: The PKPD model showed good model performance and concordance with literature data in humans. The exposure-risk model with the best goodness-of-fit for gastric ulceration was cumulative TXB2 inhibition.
Conclusions: NLME was used to characterise PK and PKPD relationships for relevant biomarkers without increasing the experimental burden of toxicity study protocols. Parameterisation of drug effects in terms of cumulative TXB2 inhibition provided good prediction of the adverse events in rodents and predicted drug effect in humans. The use of systemic exposure as surrogate for target engagement in toxicology experiments must be revisited, as it may lead to inaccurate decisions regarding drug safety.