Statistical Methods in Experimental Pathology: A Review and Primer.

Abstract

Correct use of statistical methods is important to ensure the reliability and value of the published experimental pathology literature. Considering increasing interest in the quality of statistical reporting in pathology, the statistical methods used in ten recent issues of the American Journal of Pathology were reviewed. The statistical tests performed in the articles were summarized with attention to their implications for contemporary pathology research and practice. Among 195 articles identified, 93% reported using one or more statistical tests. Retrospective statistical review of the articles revealed several key findings. First, tests for normality were infrequently reported and parametric hypothesis tests were overutilized. Second, studies reporting multi-sample hypothesis tests (e.g., ANOVA) infrequently performed post hoc tests to explore differences between study groups. Third, correlation, regression, and survival analysis techniques were underutilized. Based on these findings, a primer on relevant statistical concepts and tests is presented, including issues related to optimal study design, descriptive and comparative statistics, and regression, correlation, survival, and genetic data analysis.