Frequency and diversity of use of statistical techniques in oncology journals.

We describe the frequency with which various statistical techniques are reported in almost 5,000 articles published in five major American oncology journals during 1983 and 1984. A reader familiar with about a dozen techniques can expect to understand approximately 90% of the quantitative concepts cited in these journals. With the exception of survival analysis, these were typically the techniques encountered in many introductory statistical texts. Oncology training program preceptors concerned about making their graduates more effective consumers of the literature can use these findings in structuring exposure to quantitative skills. In four of the five journals reviewed, failure to identify the statistical methodology was among the ten most commonly encountered "techniques." This made it impossible to identify the analytical procedures used in many articles, and hence to judge their scientific validity. We encourage more editorial boards to adopt guidelines requiring a standardized format for presenting results based on a statistical analysis.