Pharmacodynamic assessment

A. General Considerations Safety information and adequate and well-controlled clinical studies that establish a drug’s effectiveness are the basis for approval of new drugs. Exposure-response data can be derived from these clinical studies, as well as from other preclinical and clinical studies, and provide a basis for integrated model-based analysis and simulation (Machado et al. 2000; Sheiner and Steimer 2000). Simulation is a way of predicting expected relationships between exposure and response in situations where real data are sparse or absent. There are many different types of models for the analysis of exposure-response data (e.g., descriptive PD models (Emax model for exposure-response relationships) or empirical models that link a PK model (dose-concentration relationship) and a PD model (concentration-response relationship)). Descriptive or empirical model-based analysis does not necessarily establish causality or provide a mechanistic understanding of a drug’s effect and would not ordinarily be a basis for approval of a new drug. Nevertheless, dose-response or PK-PD modeling can help in understanding the nature of exposure-response relationships and can be used to analyze adequate and well-controlled trials to extract additional insights from treatment responses. Adequate and well-controlled clinical studies that investigate several fixed doses and/or measure systemic exposure levels, when analyzed using scientifically reasonable causal models, can predict exposure-response relationships for safety and/or efficacy and provide plausible hypotheses about the effects of alternative doses and dosage regimens not actually tested. This can suggest ways to optimize dosage regimens and to individualize treatment in specific patient subsets for which there are limited data. Creating a theory or rationale to explain exposure-response relationships through modeling and simulation allows interpolation and extrapolation to better doses and responses in the general population and to subpopulations defined by certain intrinsic and extrinsic factors. B. Modeling Strategy In the process of PK-PD modeling, it is important to describe the following prospectively: 1. Statement of the Problem The objectives of the modeling, the study design, and the available PK and PD data; 2. Statement of Assumptions The assumptions of the model that can be related to dose-response, PK, PD, and/or one or more of the following: • The mechanism of the drug actions for efficacy and adverse effects • Immediate or cumulative clinical effects • Development of tolerance or absence of tolerance • Drug-induced inhibition or induction of PK processes • Disease state progression • Response in a placebo group Contains Nonbinding Recommendations 18 • Circadian variations in basal conditions • Influential covariates • Absence or presence of an effect compartment • Presence or absence of active metabolites and their contribution to clinical effects • The PK model of absorption and disposition and the parameters to be estimated • The PD model of effect and the parameters to be estimated • Distribution of PK and PD measures and parameters • Distributions of intra- and inter-individual variability in parameters • Inclusion and/or exclusion of specific patient data The assumptions can be justified based on previous data or from the results of the current analysis. 3. Selection of the Model The answer to the question of what constitutes an appropriate model is complex. In general, the model selected will be based on the mechanism of action of the drug, the assumptions made, and the intended use of the model in decision making. If the assumptions do not lead to a mechanistic model, an empirical model can be selected. In this case, the validation of the model predictability becomes especially important. The available data can also govern the types of models that can be used. The model selection process can be a series of trial and error steps. Different model structures or newly added or dropped components to an existing model can be assessed by visual inspection and tested using one of several objective criteria. New assumptions can be added when emerging data indicates that this is appropriate. The final selection of the model will usually be based on the simplest model possible that has reasonable goodness of fit, and that provides a level of predictability appropriate for its use in decision making. 4. Validation of the Model The issue of model validation is not totally resolved. Generally, we recommend that the predictive power of a model be dealt with during the study design as well as in the data analysis stages and that the study be designed to yield a predictive model. When plausible exposure-response models are identified based on prior knowledge of the drug before conducting an exposure-response study, the predictive power of the final models derived from the study results becomes a function of study design factors, such as number of subjects and sampling plan. The predictive power can be estimated through simulation, by considering distributions of pharmacokinetic, pharmacodynamic, and study design variables. A robust study design will provide accurate and precise model parameter estimations that are insensitive to model assumptions. During the analysis stage of a study, models can be validated based on internal and/or external data. The ultimate test of a model is its predictive power and the data used to Contains Nonbinding Recommendations 19 estimate predictability could come from exposure-response studies designed for such a purpose. A common method for estimating predictability is to split the data set into two parts, build the model based on one set of data, and test the predictability of the resulting model on the second set of data. The predictability is especially important when the model is intended to (1) provide supportive evidence for primary efficacy studies, (2) address safety issues, or (3) support new doses and dosing regimens in new target populations or subpopulations defined by intrinsic and extrinsic factors or when there is a change in dosage form and/or route of administration. VII. SUBMISSION INFORMATION: EXPOSURE-RESPONSE STUDY REPORT It is advisable for the general format and content of a clinical study report to be based on that presented in the ICH E3 guidance on the Structure and Content of Clinical Study Reports, modified to include measurements of exposure and response and planned or actual modeling and simulation. It is helpful to include a description of the assay methods used in quantifying drug concentrations (if they are components of the exposure measure) as well as assay performance (quality control samples), sample chromatograms, standard curves used, where applicable, and a description of the validity of the methodologies. The report could also contain: • The response variable and all covariate information • An explanation of how they were obtained • A description of the sampling design used to collect the PK and PD measures • A description of the covariates, including their distributions and, where appropriate, the accuracy and precision with which the responses were measured • Data quality control and editing procedures • A detailed description of the criteria and procedures for model building and reduction, including exploratory data analysis The following components of the data analysis method used in the study would also ordinarily be described: (1) the chosen dose-response or PK-PD model, (2) the assumptions and underlying rationale for model components (e.g., parameterization, error models), (3) the chosen modelfitting method, (4) a description of the treatment of outliers and missing data, where applicable, and (5) diagrams, if possible, of the analysis performed and representative control/command files for each significant model building and/or reduction step. In presenting results, complete output of results obtained for the final dose-response, or PK-PD model, and important intermediate steps can be included. A complete report would include a comprehensive statement of the rationale for model building and reduction procedures, interpretation of the results, impact of protocol violations, discussion and presentation of supporting graphs, and the ability of the model to predict performance. It is helpful if an appendix is provided containing the data set used in the dose-response or PKPD analysis, the programming codes along with the printouts of the results of the final model, and any additional important plots.