|
|
|
|
Scientific Program Recent advances and trends in statistics applied to clinical trials Download the Scientific Program : pdf document
Missing data Missing data are a recurring problem in clinical trials and handling missing data is a critical component of the analysis. Basic "naïve" methods for data imputation may lead to substantial bias in the results or to inefficient analyses but the direction of bias is relatively quantifiable and, provided a range of sensitivity analyses are also conducted, such approaches may be acceptable for regulatory submissions. Statistical methods for addressing missing data have been actively developed in recent years including likelihood, multiple imputation, EM algorithm, simulation and weighted approaches. The main focus will be on these methods and their applications to the analysis of clinical trial data under various mechanisms governing missingness and data imputation. The session will also address whether the assumptions behind these "new" methods are appropriate for confirmatory clinical trials.
Flexible designs Flexible designs for clinical trials allow for mid-course adaptations based either on interim data or information from outside the trial, without compromising the overall type I error rate. Adaptations can be sample size re-estimation, dropping or adding treatment arms, changing the study population, the primary endpoint or the test statistics, combining different phases into a single ("seamless" designs) and so on. The statistical aspects and the practical implementation (i.e. how to minimize operational biases?) of these methods has received much attention in statistical recent literature and will be addressed. Examples from real trials will be used to illustrate these approaches and the associated "pros" and "cons" to their use.
Multiplicity Issues Multiplicity issues commonly arise in clinical trials with multiple endpoints, multiple objectives and/or multiple dose levels or treatment groups. New methods, such as gate-keeping testing procedures and related strategies, have been developed in the last decade for addressing them. Other specific scenarios related to multiplicity that are receiving increasing attention are drawing specific conclusions from lists of secondary endpoints and interpreting pre-specified subgroup analyses in addition to a primary analysis based on all patients. A particularly complex scenario arises when it is desirable to consider different endpoints as primary for the interim and for the final analysis. Interim analyses still remain a topical source of multiplicity issues. All these various situations will be addressed in this session.
Meta-analyses There has been a growing interest in new approaches to meta-analyses in the last decade. These include meta-regressions, Bayesian meta-analyses, methods for identifying and/or reporting biases. Complex evidence synthesis which can be considered as extended meta-analyses is also a growing field, including "Chain-of-evidence" models and mixed treatment comparisons meta-analysis. This session will mainly focus on recent developments in the meta-analysis field illustrated through clinical trials with efficacy or safety endpoints.
|
| Last update: 04 October 2009 |
|
Home Scientific committee Organising committee Sponsors Scientific program Downloads Contacts |