Intended for healthcare professionals

Editor's Choice

Goodbye PubMed, hello raw data

BMJ 2011; 342 doi: (Published 12 January 2011) Cite this as: BMJ 2011;342:d212
  1. Fiona Godlee, editor, BMJ
  1. fgodlee{at}

This time last year the H1N1 influenza pandemic was burning itself out, having caused, thankfully, far less sickness and death than predicted. Now this year’s seasonal flu epidemic is doing its rounds in the northern hemisphere (doi:10.1136/bmj.d203, doi:10.1136/bmj.d190). The UK’s problems with uptake and availability of the flu vaccine seem to have been sorted out, but what interests me is this year’s low key approach to antivirals.

You will remember that neuraminidase inhibitors were promoted by WHO as a key part of influenza prevention and treatment, and that oseltamivir was stockpiled at vast expense by most governments around the world. The drug was made widely and easily available, but even so, huge amounts were left unused. You may also remember that serious doubts were raised about its effectiveness.

At the end of 2009 we published an update of the Cochrane review of antivirals as treatment for flu in otherwise healthy adults (BMJ 2009;339:b5106). As reported in a BMJ/Channel 4 investigation, the reviewers had found that, despite repeated requests to the drug company, Roche, they were unable to obtain the trial data necessary to validate their earlier conclusion that oseltamivir reduced complications (BMJ 2009;339:b5374).

This week the Cochrane team explains why their experience with Roche blows a hole in the systematic review enterprise (doi:10.1136/bmj.c7258). The incomplete information they obtained from Roche merely proved how inadequate the published record on oseltamivir was. The two main published trials don’t mention any adverse events, but the partial study reports from Roche listed 10 serious events, three of which were classified as possibly due to oseltamivir. By laboriously compiling a full list of industry and non-industry trials, they found one large trial by Roche Shanghai that Roche headquarters in Basel hadn’t got on their list. By looking at the regulatory documents, they found that the largest phase III trial of oseltamivir (unpublished) is hardly mentioned in regulatory documents.

From now on, they say, reviewers must have access to all unpublished data, not only from unpublished trials—the usual focus of concern about publication bias—but also from those that have been published in peer reviewed journals. Reviewers must assess entire trial programmes, and so new tools and methods are needed. If the trial reports are incomplete, reviewers should turn to reports from the drug regulators. As Tom Jefferson, the lead author for the Cochrane review, told me, “it’s goodbye PubMed, goodbye Embase.”

The reviewers have posted their new style protocol for this review on the Cochrane site and, recognising the enormity of the task, they are recording how much work is involved. But it must be clear to everyone that such a heroic approach is unsustainable across the whole of healthcare, given the resource constraints on academics and regulators. Which brings us back to what seems to be the only real solution—that the raw data from trials must be made freely available. Journals clearly have a role to play in making this happen, as An-Wen Chan agrees in his editorial (doi:10.1136/bmj.d80). The International Committee of Medical Journal Editors meets in a few months’ time. This will be on the agenda.


Cite this as: BMJ 2011;342:d212