Turning the tide on conflicts of interest

Three cheers for the editor of the BMJ! What vision, what clarity of
thought and above all what courage, criticising the errors of others.

But hang on a minute? It seems the FDA is to be blamed because drugs
are sometimes withdrawn. "Only last month, an FDA panel voted to withdraw
the breast cancer drug bevacizumab, which had been given "accelerated
approval" in 2008 (BMJ 2011;343:d4244)." Is this the same mechanism that
led to Gleevec/Glivec being approved, a drug that is so vitally
important, apparently, that Novartis is to be blamed for trying to stem
the generic copycats? And fast track approval, is that something that is
foreign to the medical press? Well, 'pon my soul what did I find here
http://resources.bmj.com/bmj/authors/fast-track-publication and publishing
faulty and mistaken articles is that something that the BMJ never does?

Far from it! And leaving a job to take up a commercial post, is that
something that only regulators do or journal editors also? Perhaps the
editor of the BMJ might like to engage in a little local historical
reflection.

So what are the real differences between the FDA and the BMJ? I can
think of at least two. First, the agency will require you to submit your
data files and your computer code with your application so that the agency
statisticians can spend weeks and months checking your claims. The BMJ may
pass it out for statistical review in which case an overworked
statistician will give it the once over and report back in a few days. But
let's not forget the wonderful weekly BMJ hanging committee. (Are coffee
and biscuits served I wonder?) Secondly, the FDA does, indeed, reverse its
decisions. On the other hand, I seem to recall a particular editor
admitting that 'half of what we publish is wrong; the problem is we don't
know which half' to which could be added 'but when we do know it's wrong
we certainly don't withdraw it.'

By all means, Dr Godlee, ban whomever you like from publishing in
your journal but pardon me for thinking that this is merely a side show.
The truth is that when it comes to what the FDA does to ensure quality and
what you do, you are not even at square one. If you want to improve the
quality of your journal, let me give you a positive suggestion. Why don't
you require that all data and all computer code required to support any
claim in any article be lodged on a special BMJ website created for this
purpose. Instead of a policy of 'Original data if you think it will help
our reviewers or if we specifically request it', a policy of 'no data and
code no BMJ article'. (I appreciate, of course, that requiring code is
going to upset your very many authors who rely on Excel for their analysis
but everybody has to make sacrifices.) Of course the growing disciplines
of forensic biostatistics and bioinformatics
http://www.ndns.nl/static/files/sls/presentations/Baggerly-AnnalsApplied... will be much easier to practice and rather less
fun but surely that will be a small price to pay.