Influence of statistician involvement on reporting of randomized clinical trials in medical oncology.

Ideally, statisticians should be involved in the design, analysis, and reporting of randomized clinical trials (RCTs). This study assessed the impact of a statistician involvement in published medical oncology RCTs between 2005 and 2009. The reporting quality of each publication was rated using the Overall Reporting Quality Score on the basis of either 2001 or 2010 Consolidated Standards of Reporting Trials criteria. A four-question email survey on the statistical design and analysis was sent to the corresponding authors of each trial. Nonresponders were approached a maximum of three times. Overall, 107 responses were received from 357 solicited authors (30%). Corresponding authors from industry-funded RCTs were less likely to respond (51 vs. 65%, P=0.013). The same person was responsible for statistical design and analyses in 47% of cases. Overall, the statistician involved held a PhD (or equivalent) in statistics in most cases. The statisticians responsible for the statistical design and analysis were listed as coauthors in 68 and 81% of RCT manuscripts. There was no statistically significant impact on manuscript reporting quality of the degree of statistician involvement in manuscript preparation. Fewer trials were reported as positive when the responsible statistician was listed as a coauthor. It is possible that RCTs included in this review are in general of higher quality or were more likely to have a greater level of statistician involvement than smaller, single-arm, or unpublished studies. This imbalance could explain the lack of significant difference observed in the Overall Reporting Quality Score between trials where statisticians were listed as coauthors or not.