Tuesday, 30 April 2013

Publication Bias - Research Transparency & Accessibility

JAMA August 8th 2012 Vol 308

I’m not sure how many psychiatrists subscribe to the Jiscmail list on Evidence Based-Health but one of the topics that comes up regularly there is the subject of publication bias. This is the phenomena that studies with positive findings are more likely to get published than studies that don’t show a difference. 

The reason this matters is that the treatments we prescribe may not be as effective as we think. This has been a hot topic for antidepressants in recent years with some Systematic Reviews concluding that antidepressants are no better than placebos except for the most severe forms of depression. One way to get round this is to require all treatment studies to be Registered before they start so that at least we can see if a study has not been reported. However this only applies to studies about treatment and it takes some effort to find out what hasn’t been published.

The next generation of research transparency comes with making the results of all clinical studies available online for everyone to access. An interesting twist in the accessibility argument comes from John Ioannidis writing in JAMA. He argues that data sets should be registered – including those from non randomised controlled trials – to enable authors to ask questions about studies which haven’t been done but should be. In other words the data that will help answer important clinical questions is already out there if only the datasets were accessible.

No comments:

Post a Comment