2004
Uppsala, Sweden
The Value of Priors and Prior Uncertainty in Clinical Trial Simulation: Case Study with Actinomycin-D in Children with Cancer
JS Barrett, J Skolnik, MR Gastonguay and PC Adamson
Pediatrics Department, College of Medicine, University of Pennsylvania and The Children's Hospital, Philadelphia, PA, USA
Introduction:
Actinomycin–D (AMD) is an antineoplastic agent used for the treatment of childhood cancers since the 1960’s. Despite its longstanding and widespread use, there is very little pharmacokinetic information from which safe and appropriate pediatric dosing can be derived.Objectives:
To assemble and review prior information about AMD’s activity, PK/PD, and clinical experience in adults and children to construct a drug and study model to explore outcomes and evaluate trial designs. To examine the sensitivity of outcomes (PK, PD, safety and efficacy) to model assumptions and various parameterizations.Methods:
Compartmental and physiologically-based PK models were created based on prior information derived primarily from the literature. Likewise, clinical event rates for drug response and toxicity measures were obtained from the literature and incorporated into the outcome model as categorical data with assigned probabilities. Model evaluation was conducted using NONMEM (version V) and simulation models were built using Pharsight Trial Simulator (version 2.1.2). Available priors from which the drug model was created include the following: DNA binding and inhibition of cell proliferation data [1, 2]; 3H-AMD animal disposition data[3,4]; AMD-3H pilot PK study in 3 patients [5].Results:
Structural models (2 and 3 CPM and physiologically-based models with allometric scaling of drug clearance) for AMD agreed well with limited adult human and animal data. The appropriateness of the various model expressions is extremely dependent on the assumption that the 3H-AMD accurately reflects the free, parent AMD in children, variability is modest and that typical scaling of drug clearance with body weight is appropriate. Assignment of the toxicity correlation is dependent on the assumption that other co-administered chemotherapeutic agents do not contribute to the AMD-associate toxicities and that we have appropriately aligned systemic versus peripheral toxicities based on the presumed exposure-response relationship. Based on our assumption of potency and composite PK model, we can discriminate dose groups independent of study design.Conclusions:
Compartmental and physiologic models used to define the underlying structural PK model of AMD are based on limited prior information. Despite the degree of parameter uncertainty, projected outcomes based on constructed models still permit the association of exposures to observed toxicities and confirms the utility of the approach. We plan to examine the validity of the proposed simulation model in an upcoming prospective, clinical trial. We further intend to construct an appropriate evidence-based pop-PK/PD model from which dosing guidance for children with cancer can be derived.References:
2. Yung BYM, Bor AMS, Chan PK. Short exposure to actinomycin D induces reversible translocation of protein B23 as well as reversible inhibition of cell growth and RNA synthesis in HeLa cells. Cancer Res. 50: 5987-5991 (1990)
3. Galbraith WM and Mellett B. Tissue distribution of [3H]actinomycin D (NSC-3035) in the rat, monkey, and dog. Cancer Chemother. Rep. Part 1, 1975; 59: 1061-1069.
4. Terasaki T, Iga T, Sugiyama Y, Sawada Y, Hanano M. Nuclear binding as a determinant of tissue distribution of adriamycin, daunomycin, adriamycinol, daunorubicinol and actinomycin D. J Pharmacobiodyn. 1984; 7(5):269-77.
5. Tattersall, M.H.M., Sodergren, J.E., Sengupta, S.K. et al. Pharmacokinetics of actinomycin D in patients with malignant melanoma. Clin. Pharmacol. Ther., 17: 701-708 (1975).