Swedish Agency for Health Technology Assessment and Assessment of Social Services
Reliable findings are within reach
Many research findings regarding effects and associations are misleading or exaggerated. The responsibility weighs heavily on those who pay for and conduct the studies, according to Professor John Ioannidis at Stanford University in California. In a recent article, he urges his research colleagues and financial backers to adopt an array of measures.
Collaborate across the board with other researchers. Genetic epidemiology is one example of a field in which such collaboration between research groups has paid off.
Accept that findings must be replicated in new studies before the results can be considered correct, especially results from laboratory studies and small clinical trials.
Submit ongoing studies, protocols and data collections to registers in order to avoid unnecessary duplication of work and to ensure that all relevant information is accessible.
Share data, protocols and tools with other researchers. Then others can verify that the results are correct. This approach is already in use within biomedicine.
Fend off the influence of conflicted sponsors or authors, even when research findings are to be used in health economics, meta-analysis and guidelines.
Use more appropriate statistical methods and apply standardised definitions and analyses. False positive results must be minimised within fields such as epidemiology, psychology and economics. Be skeptical of alleged ‘discoveries’ or ‘successes’.
Tighten requirements for study design and follow up using checklists for good research practices. Randomisation and blinding of investigators should be applied even in animal studies.
Improve peer review, reporting and dissemination of research. There are many suggestions for improvements, such as how to report various types of studies.
(www.equator-network.org).
Provide researchers with better training in methodology and statistics. One example is the Clearinghouse for Training Modules to Enhance Data Reproducibility*, at the US National Institutes of Health, NIH. • RL
Reference
Paraphrased from Ioannidis JPA. How to make more published research true. PLoS Med 2014;10:e1001747
