Optimisation of Screening Experiments for the Assessment of Analgesic Effects
Taneja A(1),Nyberg J (2),Danhof M(1),Della Pasqua O (1)
(1)Pain Project Members of the TI Pharma mechanism-based PKPD modelling platform;(2)Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden
Objectives:Experimental protocols in preclinical drug screening are often based on empirical criteria such as best guess approaches. Practical constraints prevent the use of a model-based approach, which can have major impact on the ranking of compounds at this early stage of development. In the current investigation, we apply robust ED-optimality in a prospective setting for a new chemical entity (NCE) wherein prior information from a paradigm compound is used to optimise the study design, and thereby accurately estimate the parameter of interest (EC50).
Methods:We demonstrate the advantages of the concept based on two compounds commonly used in neuropathic pain treatment, namely gabapentin (reference) & pregabalin (NCE). We apply a logistic regression model, under the assumption that EC50 is the only unknown parameter. Initial estimates for the NCE were based on an in vitro potency ratio of 2:1. The design variables for optimisation include the PK sampling times and doses. Uncertainty of 50% is assumed for the between-subject variability as well as in the parameter EC50. The design was validated using a simulation/estimation study (SSE) with n=500. The precision and bias of parameter estimates for standard and optimised protocol designs were then compared. Simulated PKPD profiles of each design were also compared. POPED 2.10/ MATLAB 7.9 were used for the optimisation procedures, whilst NONMEM 6 was used for simulation purposes.
Results:The ED-optimal doses for the paradigm (5-150mg/kg) were lower than the original doses (30-300mg/kg).Optimal sampling times were around the expected EC50.Relative standard errors for the optimal design were 32.33(14.7%) vs. 79.3(123.30)% for the original protocol. For the NCE, the EC50 estimates were close to the hypothesised true estimate, 108ng/ml, while modelling of experimental data yielded estimates of 3200 (110-7000) ng/ml. The RSE for this parameter was also considerably lower after optimisation.
Conclusions: We show that a model-based approach can be used to optimise the screening of NCEs. Empirical protocols for the assessment of drug potency in the presence wide parameter uncertainty leads to biased estimates and inaccurate ranking of candidate molecules. The use of ED-optimality concepts support the design of more informative experimental protocols, since it can warrant sampling times and dose selection that support the estimation of EC50, taking into account its uncertainty. .
References:
[1] Atkinson AC, Donev AN, Tobias AD, Optimum Experimental Designs with SAS. Oxford University Press 2006.
