Design Evaluation for PK Substudies in Phase III

From kgkowa@searle.monsanto.com Thu Aug 1 09:36:52 1996 Subject: Design Evaluation for PK Substudies in Phase III Dear NONMEM users, I have developed a preliminary PK model based on data from phase I and II studies and I now want to design a PK substudy for a pivotal phase III trial. I'm considering various alternative sampling designs varying the number of patients, number of blood draws/patient and timing of blood draws. I'm also considering designs where timing of the blood draw is random based on the time of arrival to the clinic and the time of last dose (I have some information regarding the distribution of sampling and dosing times from phase II trials that I was thinking of building into the simulation). With a random sampling design I am also considering augmenting the design to have a few patients stay in the clinic for an extended period of time in which a predose sample and several postdose samples at prespecified times will be taken to ensure that we have adequate coverage of the complete dosing interval. An approach I was considering for evaluating the candidate designs was to perform say N simulations of the clinical trial for each design based on my preliminary PK model from the phase I and II studies. I would then fit the model to the simulated data for each clinical trial and summarize the model fits for the N simulated trials for each design. Here are my questions: What is the best way to summarize the results of the model fits for the N simulated trials for making comparisons among the candidate designs? Should I summarize the point estimates (thetas and omegas) and compare them to the theoretical values used in the simulations by calculating say a %bias and %rmse for each parameter from the N fits and then compare these estimates for each of the candidate designs? In summarizing the results for each trial, should I incorporate the precision of the estimates (standard errors)? If so, how? Are there any published papers on the use of simulations in evaluating study designs? Your comments and suggestions are greatly appreciated. Thanks in advance, Ken Kowalski (kgkowa@searle.monsanto.com)