Editor's Choice

Data transparency is the only way

BMJ 2016; 352 doi: https://doi.org/10.1136/bmj.i1261 (Published 03 March 2016) Cite this as: BMJ 2016;352:i1261
  1. Fiona Godlee, editor in chief
  1. The BMJ
  1. fgodlee{at}bmj.com

The BMJ has been accused, with some justification, of being against the drug industry. The industry’s record on subverting medical education, practice, and research in its own rather than the public interest has been a key object of study for the past 20 years and more. But once in a while it’s nice to be able to show that we can be equally critical of academia. Our aim in all cases is to support unconflicted and open science.

When it comes to hiding and misreporting clinical trial data, the industry has a long string of inglorious scandals to its name: oseltamivir (Roche), reboxetine (Pfizer), and paroxetine (GlaxoSmithKline), to name but three (doi:10.1136/bmj.e7303; doi:10.1136/bmj.c4942; doi:10.1136/bmj.h4629). But as a class, academic researchers are not markedly better. Recent examples of extended delay in publishing trial results include the sagas of sentinel node biopsy and vitamin A (doi:10.1136/bmj.e8645; doi:10.1136/bmj.e8558). Many academic institutions are still failing in their responsibilities to disseminate trial results (doi:10.1136/bmj.i637). And there has been no progress that I am aware of on the availability of the trial data on statins for independent scrutiny (doi:10.1136/bmj.h3908).

The field of fluid resuscitation is particularly muddied by controversy, with supporters of colloid and crystalloid solutions still battling it out. Joachim Boldt’s breathtaking fraud—94 papers retracted at the last count—has done little to help.

So it was intriguing to be approached two years ago by the manufacturer of a hydroxyethyl starch solution, unhappy that trialists it had funded to evaluate its product were refusing to release the raw data. On the face of it, the trialists’ position seemed reasonable. The company had signed away its rights in a commendable commitment to independence. But when the trial, published in the New England Journal of Medicine, found against its product, it wanted a reanalysis. There were, it told us, enough errors in what was published to raise concerns.

As Peter Doshi explains (doi:10.1136/bmj.i1027), we sought advice from Harlan Krumholz, whose YODA project has now brokered the independent analysis of industry funded trials. But fearing that once the data were in the public domain the company could not be trusted to act in good faith, the trialists have held their position. “We do research to answer questions. They do analyses to prove a point.”

Is it possible that some data are, in Doshi’s words “too important to share”? I don’t think so, and nor does Ian Roberts, whose 2013 Cochrane review found that colloids could do more harm than good (doi:10.1002/14651858.CD000567.pub6). “I don’t think it should be [the authors’] decision,” he said. “They have valid concerns but it would be better if they made [the data] open.”

Doshi finds that there is enough doubt over the analysis and reporting of the data, including from one of the authors, to warrant reanalysis by people other than the trialists. Such independent reanalysis and public access to anonymised data should anyway be the rule, not the exception, whoever funds the trial.

Is it possible that some data are “too important to share”?

View Abstract