Statistical fundamentals on cancer research for clinicians: Working with your statisticians
Wei Xu, Shao Hui Huang, Jie Su, Shivakumar Gudi, Brian O’Sullivan
Abstract
PURPOSE: To facilitate understanding statistical principles and methods for clinicians involved in cancer research. METHODS: An overview of study design is provided on cancer research for both observational and clinical trials addressing study objectives and endpoints, superiority tests, non-inferiority and equivalence design, and sample size calculation. The principles of statistical models and tests including contemporary standard methods of analysis and evaluation are discussed. Finally, some statistical pitfalls frequently evident in clinical and translational studies in cancer are discussed. RESULTS: We emphasize the practical aspects of study design (superiority vs non-inferiority vs equivalence study) and assumptions underpinning power calculations and sample size estimation. The differences between relative risk, odds ratio, and hazard ratio, understanding outcome endpoints, purposes of interim analysis, and statistical modeling to minimize confounding effects and bias are also discussed. CONCLUSION: Proper design and correctly constructed statistical models are critical for the success of cancer research studies. Most statistical inaccuracies can be minimized by following essential statistical principles and guidelines to improve quality in research studies.