The Use and Misuse of Statistical Methods Reported in Clinical Cancer Research

• Activity Number: 40
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Education
• Author(s): Diane Liu*+ and J. Jack Lee
• Companies: MD Anderson Cancer Center and MD Anderson Cancer Center
• Keywords: correlated data ; multiplicity ; statistical education

Abstract

We survey the recent 5 years of publications in the journal "Clinical Cancer Research" to identify the statistical methods used in study design and data analysis in cancer research. Commonly used statistical methods such as t-test, Wilcoxon test, chi-square test, and log-rank tests are usually done correctly. On the other hand, data of correlated nature such as repeated measurements, longitudinal data, and competing risks are less well presented and analyzed. Multiplicity issues are seldom addressed. More advance statistical techniques are rarely used. We will report the findings and point out the main areas that require more attention in statistical education and practice.