Study design aspects and inter-subject variability in longitudinal clinical phase II dose-finding trials.

MCP-Mod has been proposed as analysis method for investigating the dose-response (DR) relationship and dose finding in clinical Phase II trials. The original methodology as well as several generalizations are well accepted by Health Authorities. In this paper, we investigate the application of the generalized MCP-Mod approach in chronic pain, a challenging therapeutic area with primary endpoints based on patient reported outcomes, repeatedly measured over time, with high inter-subject variability and potentially high premature discontinuation rates. In a simulation study, we investigate the impact of several study design factors on the ability to establish proof of concept, to derive information on the functional DR relationship, and to estimate target doses of interest (eg, ED80). Furthermore, different methods to derive confidence intervals for the target doses of interest are compared regarding their coverage rates and widths. Proof of concept is well established by MCP-Mod even in longitudinal trials with high inter-subject variability. While the most reasonable DR model is also selected in most cases, estimated target doses and their bootstrap confidence intervals have to be treated with some caution, if trial data are strongly affected by individual heterogeneity as observed in two recent chronic pain trials. We describe encountered challenges and provide recommendations for designing future dose ranging longitudinal trials with high inter-subject variability under model uncertainty.

[1]  Yahong Peng,et al.  Recommendations for the Primary Analysis of Continuous Endpoints in Longitudinal Clinical Trials , 2008 .

[2]  F Bretz,et al.  Combining Multiple Comparisons and Modeling Techniques in Dose‐Response Studies , 2005, Biometrics.

[3]  N. Hjort,et al.  Frequentist Model Average Estimators , 2003 .

[4]  A. Hartford,et al.  Dose-ranging design and analysis methods to identify the minimum effective dose (MED). , 2017, Contemporary clinical trials.

[5]  H. Dette,et al.  Model selection versus model averaging in dose finding studies , 2015, Statistics in medicine.

[6]  Holger Dette,et al.  Practical considerations for optimal designs in clinical dose finding studies , 2010, Statistics in medicine.

[7]  Björn Bornkamp,et al.  Viewpoint: model selection uncertainty, pre‐specification, and model averaging , 2015, Pharmaceutical statistics.

[8]  Bjoern Bornkamp,et al.  Characterization of dose‐response for count data using a generalized MCP‐Mod approach in an adaptive dose‐ranging trial , 2015, Pharmaceutical statistics.

[9]  Frank Bretz,et al.  Design and Analysis of Dose-Finding Studies Combining Multiple Comparisons and Modeling Procedures , 2006, Journal of biopharmaceutical statistics.

[10]  Frank Bretz,et al.  Model‐based dose finding under model uncertainty using general parametric models , 2013, Statistics in medicine.

[11]  D. Verrier,et al.  Dose-finding studies, MCP-Mod, model selection, and model averaging: Two applications in the real world , 2014, Clinical trials.