Rocio Lledo-Garcia, Foteini Strimenopoulou, Ruth Oliver, Miren Zamacona
Pharmacometrics Department , Global Exploratory Development, UCB Pharma, Slough, UK
Objectives: During first in human- dose escalations studies, the pharmacokinetic (PK) data may be required to compare the exposure to the toxicological exposure limits or to relate to the pharmacodynamic findings and to adjust the doses to be evaluated accordingly. However, at the time of dose escalation decision, only a truncated PK profile is available – usually less than one half life – making difficult to predict accurately exposure variables such as AUC0-inf. The human PK of monoclonal antibodies (mAb) is in general well predicted from animal data. Therefore, a Bayesian framework with the use of prior information could be useful to support dose escalation studies. The aim of this work was i) to define the prior distributions and ii) to evaluate the performance of the NONMEM prior functionality compared to a full Bayesian method when applied to PK analysis of mAb.
Methods: A two compartment PK model with linear and nonlinear elimination was used as structural PK model.. Prior distributions for the population mean PK parameters and interindividual variabilities were selected based on several sources of information such as scaled PK parameters from monkey and clinical PK data from similar type antibodies, reflecting the knowledge and uncertainty around parameter distributions. The PK model was implemented using the prior functionality in NONMEM and in WinBUGS. Truncated PK data available at the time of each dose escalation were analysed using both approaches and predictions for the next dose level were performed. A comparison in terms of parameter estimates, parameters precision and run time between NONMEM and WINGBUGS results was performed.
Results: The use of prior distributions to inform PK model integrated together with data generated in the study allowed a good estimation of the model parameters. PK parameters estimated in NONMEM using the prior functionality were in close agreement with those produced by the full Bayesian analysis.
Conclusions: A Bayesian framework for dose escalation in FIM studies for mAb has been established, allowing a reliable prediction of the exposure at the next dose level based on limited PK data. The use of the prior functionality in NONMEM yielded similar results and it showed to reduce run times.
Reference: PAGE 21 (2012) Abstr 2438 [www.page-meeting.org/?abstract=2438]
Poster: Estimation methods