Please activate JavaScript!
Please install Adobe Flash Player, click here for download

ISCB2014_abstract_book

ISCB 2014 Vienna, Austria • Conference Courses 9 Monday25thAugustTuesday26thAugustThursday28thAugustAuthorIndexPostersWednesday27thAugustSunday24thAugust Sunday, 24th August 2014 – Pre-conference Courses - Full-day Conference courses Course 1 Handling missing outcome data in clinical trials I White1 , S Seaman1 1 MRC Biostatistics Unit, Cambridge, United Kingdom   This course aims to provide practising statisticians with the necessary practical skills to handle missing data in their analyses, and in particular to move beyond the use of complete‐case analysis and last observation carried forward analysis. Four sessions will each consist of a lecture followed by a short discussion exercise: 1. An introduction to missing data in randomised trials: popular ways to analyse trials with missing data, focussing on their assumptions; the inten- tion‐to‐treat principle. 2. Mixed models analysis of incomplete data: how it should be imple- mented in randomised trials, including issues arising from missing base- line data. 3. Multiple imputation: a brief introduction, and how to use it in ran- domised trials. 4. Sensitivity analysis to departures from assumptions: principled sensitivity analysis, and a suggestion of how to implement it. We focus on trials with quantitative outcomes, and also consider binary outcomes but not time‐to-event outcomes. Course 2 Data analysis with competing risks and multiple states R Geskus1 , H Putter2 1 Academic Medical Center, Amsterdam, The Netherlands, 2 Leiden University Medical Center, Leiden, The Netherlands   In the end we all die, but not all from the same cause, nor with the same life histories. This course will be devoted to the analysis of different types of events that can occur either exclusively (competing risks) or sequen- tially (multi‐state models). The morning session is devoted to competing risks analysis. Competing risks take the spectrum of event types into account. The main difficulty is the choice of the correct quantity to be estimated. When do we need a competing risks analysis? Do we want to estimate cause‐specific or subdis- tribution hazards?What do we need to assume with respect to the censor- ing by the competing event? The actual analysis is much easier because software to perform the computations has become readily available. Multi‐state models are the topic of the afternoon session. They extend competing risks models to the analysis of what happens beyond some first event, by allowing individuals to progress through different states. Estimation in multi‐state models is reasonably straightforward, at least as long as all transitions are observed without uncertainty. Arguably the most interesting use of multi‐state modles is for dynamic prediction. This aspect will be discussed in some detail; dynamic prediction using multi‐ state models will be contrasted with more recent developments such as landmarking.   Sunday, 24th August 2014 – Pre-conference Courses - Half-day - Morning Conference courses Course 3 Extension of frailty models for recurrent or clustered survival data with prediction V Rondeau1 1 Department of Biostatistics at the National Health and Medical Research Institute of the University Bordeaux Segalen, Bordeaux, France   Simple shared frailty models have been largely developed and applied for recurrent or clustered survival data in the literature. However, extensions of frailty models are less common in publications and are not well repre- sented in classical software. We are aiming at filling this gap by consider- ing extensions of frailty models (such as additive frailty models, nested frailty models or joint frailty models) and by presenting an implementation of these models using the R package frailtypack. Particular interest will be given to joint frailty mod- els in order to jointly analyse recurrent events such as cancer relapses and a dependent terminal event (death or loss to follow‐up). Prediction tools associated with this package will be presented, too. The first part of this course will introduce general frailty models, the esti- mation methods and the research questions they may address. The second part of this course will be dedicated to the joint frailty models with illustration on real data. The estimation and the predictive dynamic tools that can be derived from them will be exposed, with methods to evaluate their performance. Emphasis is given, via examples on real data, of the ability of extended frailty models to describe a very broad range of practical situations. Each concept will be illustrated through implementation of these models using the R package frailtypack. Course 4 Statistical methods in Systems Medicine H Fröhlich1 1 University of Bonn, Bonn, Germany   Discovery of prognostic and diagnostic biomarker signatures for diseas- es, such as cancer, is seen as a major step towards a better personalized medicine. During the last decade various methods have been proposed for inferring such signatures from high dimensional molecular data (e.g. genomics, transcriptomics, proteomics and metabolomics profiles). However, one important obstacle for making molecular signatures a stan- dard tool in clinical diagnosis is the typical low reproducibility of these sig- natures combined with the difficulty to achieve a clear biological interpre- tation. For that purpose in the last years there has been a growing interest in approaches that employ biological background knowledge. In addition, the increasing availability of different -omics profiles for the same patient now raises the question on how to integrate these data. The purpose of this course is to shed light on current integrative model- ing efforts that combine different ‐omics entities and/or biological back- ground knowledge in order to achieve higher robustness, stability and interpretability of molecular biomarker signatures.   Sunday, 24th August 2014 • Pre-conference Courses - Full-day

Pages Overview