Context: We have developed the Splus and R functions, PFIM 1.2 and PFIMOPT 1.0, for population design evaluation and optimisation, respectively [1]. Their relevance has been demonstrated in several studies [2-3]. However, those functions allow only structural models defined with an analytical form, and can not deal with ordinary differential equations (ODE). Moreover, the user has to write the model, which can be complex, especially for multiple doses. Last, the Simplex algorithm used in PFIMOPT may converge towards local optima and optimises the sampling times in continuous intervals, leading sometimes to unfeasible sampling times in practice.

Objectives: 1) To extend PFIM and PFIMOPT for models defined by ODE systems and to propose a library of usual pharmacokinetics models. 2) To implement a new algorithm in PFIMOPT ensuring the convergence and integrating the clinical constraints.

Methods: The lsoda function of the odesolve package of R to solve stiff and non-stiff systems of ODE has been integrated in PFIM and PFIMOPT. The numerical derivatives needed for the computation of the Fisher information function in case of ODE are computed using the fdHess function of the nlme library. For usual PK models, a library has been developed in R. This library is easily usable in PFIM and PFIMOPT. It handles four types of drug administration (PO, infusion, bolus, infusion with load dose) with a one or a two compartment model after either one dose, repeated doses or at steady state. Last, we have implemented the Fedorov-Wynn algorithm in PFIMOPT, using a C code linked with R via a dynamic link library. This algorithm converges towards the D-optimal design and its usefulness has already been demonstrated [4]. It optimises both the group structure of a design (number of groups, number of subjects per group, number of samples per subject), and the times defined in a given finite set of times. This set can be specified by the user as a set of clinically feasible times.

These new functions, PFIM 2.0 and PFIMOPT 2.0 will be demonstrated on several examples. A windows interface of those functions will also be demonstrated.

Acknowledgements. Part of this work has been funded by Roche during a post-doctoral position of S.Retout.

References.
[1] http://www.bichat.inserm.fr/equipes/Emi0357/download.html
[2] Retout, Mentré, Bruno. Stat Med,. 2002, 21: 2623-39.
[3] Retout, Charoin, Jorga. PAGE Meeting 2005 Poster n°765.
[4] Retout, Comets, Samson, Mentré. PAGE Meeting 2005. Poster n°775.