Simulation Methodology for Quantitative Study Decision Making in a Dose Response Setting
Meg Bennetts
Pfizer Ltd, Sandwich, Kent, UK
Objectives:
Tools & Methodology: Develop a framework for simulating study data based on current longitudinal models using R and NONMEM6. Obtain a table of Trial Performance Metrics that quantify study success in terms of how many times a correct decision is made.
Study: Quantify the expected performance of a designed Phase IIb dose-ranging study and probable Phase III dose by evaluating Decision Criteria for FEV1 and Heart Rate endpoints.
Background: Decision criteria are rules to decide the course of action (a decision) prior to initiation or after completion of a clinical study. The performance of the drug compound and study are assessed against Target Values defined by the project team based on specific issues pertinent to the stage of drug development, such as clinical relevance or commercial viability. The true treatment effect DELTA is unknown, but can be estimated from the predictive distribution of DELTA, for a given model fitting the current state of knowledge about the drug effect, integrated across fixed and random effects. This estimate of the true effect, DELTA^, can be used to assess the technical success of a drug compound irrespective of study design. In contrast, T is an estimate of DELTA using data analytic models (the formal prospectively defined study data analysis methods). This estimate of the trial effect can be used to assess the success of the study design including elements such as sample size. In a simulation context Quantitative Decision Criteria allow us to evaluate the probability of DELTA^ and T achieving their Target Values for a given set of model parameters common to both estimates within paired replicates. The resulting table of Trial Performance Metrics enable the project team to assess the probability of Technical Success for the drug compound and the probability of making a correct Go/No Go decision given the study design. [1, 2, 3, 4]
The designed, 4 week, study consisted of 5 treatments (three dose levels of novel drug, placebo and active control) with 80 subjects/treatment group. Longitudinal FEV1, PK and Heart Rate endpoints were to be measured on days 1 and 29.
The NONMEM Models for the endpoints were based on previous studies in the drug program. FEV1 was a one-compartment, bio-phase kinetic/sigmoid Emax PD (KPD) model with baseline and circadian variation. Heart rate consisted of a PK model with complex absorption profile and high accumulation, and an Emax PD model including covariates for gender, baseline and circadian variation. [5]
Methods: Model parameter estimates, parameter uncertainty and random effects were extracted from the final NONMEM6 model estimation output files.
Trial input datasets specifying the structure of the study design were constructed for NONMEM6 simulation of FEV1 using R. Truth input datasets were constructed with an identical design with only one subject per dose. Replicate-specific paired parameter Truth/Trial simulation control files were constructed by overwriting $THETA parameter values with ones drawn from the underlying multivariate normal distribution centred on the final model parameter estimates with associated variance/covariance. The resulting study specific longitudinal FEV1 tables read back into R for derivation of endpoints, analysis and application of Decision Criteria.
Once all simulations were complete, Truth and Trial success proportions were summarised to form Tables of Trial Performance Metrics.
Conclusions: A fit for purpose process was developed for simulations using R and NONMEM6 enabling the study team to make quantitative decisions regarding choice of study Decision Criteria and probable doses to manufacture for Phase III.
Discussion: The poster will discuss simulation and application of quantitative decision criteria in a dose response setting and issues encountered regarding appropriate simulation of Truth replicates for the FEV1 and Heart Rate models.
References:
[1] Lalonde, R.L., et al. "Model-Based Drug Development". Clin Pharm Ther 2007;82:21-32.
[2] Kenneth G. Kowalski, Jonathan L. French, Mike K. Smith, Matthew M. Hutmacher. "A Model-Based Framework for Quantitative Decision-Making in Drug Development". ACoP, Tucson, AZ, March 2008.
[3] Kowalski, K.G., Ewy, W., Hutmacher, M.M., Miller, R., and Krishnaswami, S. "Model-Based Drug Development - A New Paradigm for Efficient Drug Development". Biopharmaceutical Report 2007;15:2-22.
[4] Tracy Higgins, Meg Bennetts, Patrick Johnson. "Simulation and Design Considerations for Noninferiority Trials in Phase II". PAGE, Marseilles, France, June 2008.
[5] Jakob Ribbing, Gai Ling Li, Erno Van Schaick, Lutz Harnisch, "Assessing the therapeutic benefit of a new inhaled QD LABA through population PK-safety & efficacy modelling". ERS, Vienna, Austria, September 2009.
