Let’s work togetherBMJ 2013; 347 doi: https://doi.org/10.1136/bmj.f5111 (Published 14 August 2013) Cite this as: BMJ 2013;347:f5111
- Trish Groves, deputy editor, BMJ
John Ioannidis, in a long and thoughtful interview in The Atlantic, was recently called a “dashing nerd” for his work as a meta-researcher: it was meant as a compliment (www.theatlantic.com/magazine/archive/2010/11/lies-damned-lies-and-medical-science/308269). His work couldn’t get much more meta than his current BMJ paper with US colleagues Konstantinos C Siontis and Tina Hernandez-Boussard (doi:10.1136/bmj.f4501).
They searched 2010’s PubMed for meta-analyses of randomised controlled trials on the effectiveness of interventions, randomly sampled 5% of them, searched with no date restriction for other potentially overlapping meta-analyses, and homed in on 20 clinical topics to assess the degree of redundancy. Two thirds of index meta-analyses had at least one partially or fully overlapping meta-analysis, few of them real updates. For instance, on statins for atrial fibrillation after cardiac surgery, there were 11 meta-analyses with similar eligibility criteria and overall results, albeit with a variety of included trials. The authors’ cautious conclusion is that “there might be some redundancy in published meta-analyses and waste of effort.” Indeed.
David Moher’s linked editorial argues that, while there’s no magic number of meta-analyses, some judicious replication is required for drug regulation in different territories, and can increase clinicians’ and other decision makers’ confidence in the results (doi:10.1136/bmj.f5040). If all meta-analysts registered their protocols prospectively, ideally at the international registry PROSPERO (www.crd.york.ac.uk/NIHR_PROSPERO), and then followed the PRISMA reporting statement (www.prisma-statement.org), the evidence base would be in much better shape, says Moher. That’s why the BMJ asks authors to do both (www.bmj.com/about-bmj/resources-authors/article-types/research). Meanwhile, in rapid responses to the paper, Ingrid Mühlhauser and colleagues suggest a creative solution: design and perform each meta-analysis as a project by two independent working groups, planning replication right from the beginning (www.bmj.com/content/347/bmj.f4501/rr/656302). Indeed, that’s what the Multiparty Group for Advice on Science and the Cochrane Acute Respiratory Infections Group are doing right now with the individual participant data from Roche’s trials of oseltamivir—although the two groups got there by a much less collaborative route (http://blogs.bmj.com/bmj/2013/06/26/trish-groves-what-does-tamiflu-do-and-how-will-we-know).
Richard J Wassersug, cancer researcher, has been an individual participant in a trial (doi:10.1136/bmj.f4879). If we are serious about getting more patients to join clinical trials, he warns, we need to improve the language used in consent forms. The standard format includes pages where every paragraph begins with “You will,” “You can,” “You may,” and so on, reviving for him the memory of a dressing down at school. “You realise how much trouble you are in? You hear me? You better not do that again! You understand? You better watch out!” Wassersug suggests much more inviting and friendly language, such as “If you join this study, we’ll be asking you to fill in a questionnaire on the side effects you experience,” and signs off with “I hope you agree … You better.”
Cite this as: BMJ 2013;347:f5111