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

ISCB2014_abstract_book

68 ISCB 2014 Vienna, Austria • Abstracts - Oral PresentationsWednesday, 27th August 2014 • 14:00-15:30 Monday25thAugustTuesday26thAugustThursday28thAugustAuthorIndexPostersWednesday27thAugustSunday24thAugust Wednesday, 27th August 2014 – 14:00-15:30 Invited session S2 The power of data sharing: advancing research for everyone’s benefit? (Panel discussion) Organizers: Martin Posch and Franz König Panelists: Hans-Georg Eichler, Senior Medical Director, EMA; Simon Day, CTCT; Ulrich Burger, Roche; Trish Groves, BMJ; Stephen Senn, CRP-Santé   It is widely recognized that the current publication practice of clinical trial results is deficient. A large number of trials are never published, and if they are published, essential information is often not included in the published manuscripts. Potential consequences are impaired meta-analyses, clinical trials that are unnecessarily repeated and a lack of transparency regarding the statistical analysis. To address these issues, recently, academia, journal editors, regulators and the pharmaceutical industry proposed a variety of approaches to data sharing and the European Medicines Agency plans to finalise its draft policy on the publication and access to clinical-trial data by June 2014. Of particular interest is wether and how access to patient level data is granted. In this panel discussion key stakeholders from academia, regulators and industry will explore the wide range of opportunities and discuss the op- portunities and risks that arise with the implementation of comprehensive data sharing commitments. Contributed sessions C37 Causal inference from observational studies I C37.1 Adjusted survival curves by using inverse probability of treatment weighting: the comparison of three adapted log-rank tests F Le Borgne1,2,3 , M Giral2 , A-H Querard1,2,4 , Y Foucher1,2 1 Department of Biostatistics EA 4275, Nantes, France, 2 Transplantation, Urology and Nephrology Institute (ITUN), Nantes, France, 3 IDBC/A2com, Pace, France, 4 Departmental Hospital Center of Vendée, La Roche sur Yon, France   In observational studies, the presence of confounding factors is common and the comparison of different groups of subjects requires adjustment. In presence of survival data, this adjustment can be achieved with a mul- tivariate model (usually a Cox model) allowing to validate a difference ob- served from crude survival curves. However, by using such type of regres- sion, the effect of the factor under interest is often summarized by the haz- ard ratio (HR). This loss of information is so damaging that most research projects in biology or medicine present both crude survival curves (biased but illustrating precisely the differences in survival) and adjusted HR (not biased but too synthetic). A recent solution is the use of adjusted survival curves and log-rank test based on inverse probability of treatment weight- ing (IPTW). However, three adaptations of the log-rank test are found in the literature without any comparison of the performances in terms of type I and II errors.We performed a simulation study in order to (i) evaluate if the performances of these adjusted log-rank tests are acceptable com- pared to the Cox model used classically, and (ii) choose the most powerful of these three IPTW approaches. For illustration, we also propose to study the patient and graft survival of kidney transplant recipients according to the expanded donor criteria (ECD). Among the three approaches, the one proposed by Xie and Liu (2005) should be preferred in future studies to compare adjusted survival curves. Nevertheless, the results show that the Cox model remains the most efficient approach.   C37.2 Inverse probability weighting of overmatched nested case-control data to enable estimation of main effects and interactions B Delcoigne1 , E Colzani1 , K Czene1 , M Reilly1 1 Karolinska Institutet, Stockholm, Sweden   Introduction: Matched nested case-control designs are generally ana- lyzed with conditional logistic regression. However, breaking the match- ing in such data and using inverse probability weighting offers a way to exploit the data for additional research questions. Our aim is to use this method for nested case-control data to overcome a problem generated by overmatching and enable us to address a research hypothesis involving interaction terms. Methods: A nested case-control study was conducted of lung cancer in women who had radiation therapy for a previous breast cancer diagnosed in Sweden during 1958-2001. Cases were individually matched to controls on age, calendar period of diagnosis and region. We broke the matching and analyzed the data with weighted Cox proportional hazards regres- sion, investigating the role of radiotherapy as a potential risk factor for lung cancer and its effect modification by smoking. Results: The study included 1525 breast cancer patients, of whom 731 were lung cancer cases. Overmatching was apparent, with 75% of the

Pages Overview