How to assess the scientific integrity of the work of one author or author group?
2Amsterdam UMC, Netherlands
3University of Melbourne, Australia
4University of Nottingham, United Kingdom
Background: Data-fabrication is a considerable problem in randomised controlled trials (RCTs), and might affect 20% of the published RCTs. It is likely that when an author has published fabricated RCTs, his other work also might be affected, a phenomenon that has been confirmed empirically.
Objective: Since methods to assess the work of one author on trustworthiness are lacking, we propose a method to investigate integrity issues in randomised clinical trials of one author or author-group.
Design: Based on practical experience while investigating problematic RCTs, and based on a recent published scoping review on methods to assess research misconduct, we propose a series of tools to investigate the published work of one author or author-group.
Results: In the investigation of the work of one author, we recommend a systematic search of the work of the involved author in Pubmed and other databases, as well as a search of trial registries for unpublished RCTs. Studies should be tabulated and graphed in order to assess consistency between registration, execution and submission. Comparison of baseline and outcome tables can reveal copying, particularly between studies on a similar topic. Studies on similar interventions in the same time period with similar participants raise concern, as are large differences in baseline characteristics or effect size between studies on similar types of patients and interventions. Assessment of baseline characteristics from multiple randomised clinical trials in Monte Carlo analysis assesses whether the data were generated by true randomisation. If serious concerns are raised, a more thorough investigation should be performed, including assessment of original data and ethics and other governance documents.
Conclusions: The proposed methods can help to assess the work of one author. Assessment of the integrity of all randomised clinical trials of one author or author-group is more powerful than assessment of single trials. Definite conclusions on fabrication can usually only be drawn when original data are assessed. Finally, there is need for discussion if, how and by whom the further work of an author with fabricated RCTs should be assessed, as well as the governance under which action should be taken. No patient involvement.