Erschienen in:
01.12.2007 | Editorial
Authors should publish their raw data
verfasst von:
Nicolaas J. D. Nagelkerke, Roos M. D. Bernsen, Diaa E. E. Rizk
Erschienen in:
International Urogynecology Journal
|
Ausgabe 12/2007
Einloggen, um Zugang zu erhalten
Excerpt
Most publications in today’s biomedical journals are original research articles that, by definition, are based on novel data. Exceptions are hypothesis papers, systematic reviews, meta-analyses, clinical guidelines, editorials, clinical opinions/updates, case reports, etc. Normally, the procedure of producing these original publications is fairly standard: Authors write a research protocol, obtain ethical approval, apply for funding, and carry out the interventions and observations or measurements. The data thus generated are analyzed in a way that seems appropriate for the research questions at hand and often involves professional statisticians, and conclusions are subsequently written in the form of a manuscript that is submitted to a journal. Peer reviewers then assess the content and judge whether the work is original, of sufficient merit, and scientifically relevant. If accepted for publication, the manuscript will be published as an article that can be read and used by the medical community at large for further research or patient care. However, reviewers and journal editors normally base their decisions on the data provided in the submitted manuscript alone. With rare exceptions, neither sees the raw data collected and very rarely will the readers also have access to these data after publication. Instead, all that is provided in the manuscript are aggregate statistical measures such as means, standard deviations, correlation coefficients, etc. or graphs such as histograms, box-and-whisker plots, etc. In addition, findings from more intricate statistical analysis such as adjusted odds ratios are also increasingly reported in recent publications. These statistical measures, graphs, and analyses are usually chosen to support the conclusions drawn by the authors and are therefore
intrinsically selective. As a direct result, it is normally difficult for editors, reviewers, and readers to objectively gauge whether the conclusions drawn by the authors are fully justified by the data that they have collected [
1]. Alternative analyses that would validate or test those conclusions cannot also be carried out by independent or extrinsic scientists. It is actually strange that we, the scientific community, are so comfortable with this submission/peer review/publication cycle that we rarely question it. This is unique because regulatory authorities such as the FDA would not normally license a new drug without full access to the primary data of the studies carried out by the pharmaceutical industry in support of their application [
2]. Despite the fact that admission of incorrect findings to the scientific body of knowledge can be just as detrimental as permitting the use of ineffective drugs and that the whole scientific process can be fallible, we continue to fail to apply the same standards [
3‐
5]. …