Statistical methods in cancer research

With the pervasive presence of computers, the application of sound statistical theory is now expected in any medical study. The Consort statement published by Journal of the American Medical Association and the British Medical Journal makes mandatory a standard checklist with any submitted publication [1,2]. However, this is just one of many guidelines [3±7] that have recently appeared in the medical literature addressing both design and analysis of clinical studies. Indeed, there are many guides for studies with more general outcomes than the effect of a drug, such as cost-bene®t ratios or quality of life. Developments both in the theory and the ability to apply theory have made many innovational studies possible. There are many computationally intensive exploratory techniques that enable a more complete investigation of data. Allied to the development in software and hardware is the ease in which data may be collected and accessed. This is why it is even more necessary now that attention be given to basic precautions against invalid data and incorrect analyses. Our discussion will focus on issues in cancer study design and analysis. In this context we will be referring to the many different types of cancer studies: