The main tasks of the Biostatistics Division are service and research in the field of biostatistical methods and their application in cancer research. We provide statistical support for all scientific activities of the DKFZ, from in vitro and animal studies to human subject research including clinical trials, thus linking the methodical research and biomedical disciplines. Our support covers experimental design, sample size estimation, data analysis and preparation of results for publication. It ranges from brief statistical consultations to longterm collaborations and covers standard statistical analysis approaches as well as the development of complex statistical methods tailored to specific questions. The ongoing evolution of novel measurement techniques and platforms, and the development of new research questions makes it necessary to continuously refine biostatistical and biomathematical methodology and to develop and implement new methods for analysis. Our current research areas reflect the requests we are confronted with from collaborators. We develop and assess efficient and valid methods for visualizing, integrating and analyzing data, in particular high-dimensional molecular data. We develop optimized biometrical designs in experimental cancer research and in clinical studies.

FUTURE OUTLOOK
One focus of interest of our research is the evaluation of molecular data in biomarker studies. We will extend our research to find associations between clinico-pathological factors, prognosis or response to therapy with the aim of identifying diagnostic, prognostic or predictive factors. We will develop and validate statistical methods for classification and prediction using high-dimensional data. Another area of research is the development and application of statistical and stochastic models for dose-response relationships and optimal design of studies for drug combinations. In the era of personalized medicine we investigate designs for (early) clinical trials for biomarker-targeted drugs since biomarker-guided treatment decisions for individual patients are expected to improve the effectiveness of these treatments. In this context, we will develop Bayesian designs which are flexible and able to borrow information within trials or from outside sources.