Research integrity has two aspects that partially overlap. Firstly, there is integrity of the scientific literature, namely its freedom from errors, and secondly, the integrity of the researchers themselves, i.e. whether they act fairly and honestly. Failures of either can reduce the reproducibility of science, reduce its efficiency, waste funding, and have negative impacts on the scientists themselves, as well as, in some cases, their research subjects.
In the life sciences, statistician John Ioannidis has shown that study design and external pressures mean most published research findings are likely to be false, and drug makers Bayer and Amgen reported that 75% and 89% of preclinical findings could not be repeated. The number of published retractions is growing at an accelerating rate, and in most cases, the cause for the retraction is attributed to misconduct.
Three reasons are commonly evoked to explain how the errors that prevent reproducibility arise. They can be caused by bad luck or statistical flukes; by sloppy science, such as using inappropriate statistics, accidentally mixed up samples, or mistakes in data entry; and by deliberately falsifying results, i.e. research misconduct.
While nothing can be done to stop anomalous results, it should be possible to improve the rigor with which research is conducted, and to minimize the incidence of research misconduct.
Improving research integrity will require a bi-directional approach. Early researchers should be provided with training in use of equipment and analytical tools, and also on the ethics and norms of their chosen fields. However, this “bottom up” approach is unlikely to achieve much unless it is combined with “top down” measures to promote compliance. Such measures include improving the quality of pre-publication peer review (to prevent sloppy or dishonest papers being published); improving post-publication peer review (to allow papers with errors to be detected and corrected rapidly); and establishment of systems to enable those who commit misconduct to be held responsible, and to be given appropriate penalties.
Improving the integrity of science, its reproducibility and its efficiency, is not only the responsibility of individual researchers. Vital contributions are also needed from research institutions and their administrators, journal editors and publishers, peer reviewers, whistle-blowers, charitable and governmental funding bodies, industry, learned associations and academies, and national offices of research integrity.
This post was first published on the Royal Society’s In Verba blog on 17 April 2015.