Avoiding biostatistical pitfalls in the design and analysis of head and neck cancer clinical trials

The ultimate objective of any clinical trial is to obtain the correct answer to an important medical question. The science of biostatistics plays an important role in helping to meet this fundamental objective. By properly applying sound statistical principles to clinical research in oncology, one can insure that results of completed studies are valid and convincing to others in the scientific community. To this end, the biomedical literature contains several excellent and thorough discussions regarding the design, execution, and analysis of cancer clinical trials and methods of data acquisition [1–3]. These references draw attention to certain basic components of a successful clinical trial, including: (1) a clear, unambiguous protocol which addresses a significant medical question, (2) well-defined conditions for entry of patients on-study, (3) sample sizes sufficiently large and duration of follow-up adequate to detect treatment effects if they are present, (4) a clear description of treatment regimens and experimental design, (5) explicit definition of endpoints used for efficacy and safety evaluation, (6) patient record forms and data management procedures which enhance data quality, (7) appropriate methodology for data monitoring and statistical analysis to account for incomplete data, and (8) appropriate consideration of ethical issues. While there may be a variety of equally plausible ways to satisfy each of these criteria, their precise specification will depend on the goals of the particular trial, its administrative structure, and the nature of the treatment and disease under study.