Friday, 21 March 2014

We got the power

Given that title, I could use this platform to give a talk about multiculturalism in which Loreen, a powerful singer with Moroccan parents, living in Sweden, sang a song in English at the opening of Eurovision 2013.  But I won't. Instead I'll make small-small addendum to my piece about the ethics of using lab animals in research and teaching.  I am in the middle of a year long subscription to the general science journal Nature, which I got for an unrefusable €32.  It comes through the door in a plastic wrapper every week and I wish I had time to read them as they come in.  But I don't.  I've only read about 2 books in the last 12 months, down from a pre-Institute norm of >2 each week, so all too often the next Nature arrives before the last one has been unpacked.  During a long weekend, I've been trawling through some of the backlog and came across an editorial from 12/12/13 about how the Animal House at Imperial College London had, despite reams of legislation, carefully written standard operating procedures, and dedicated staff, been found wanting in its duty of care for the animals that it used in its research programmes.

The editorial points out two interesting things.  First is that the core legislation in England dates back more than 100 years to 1911.  In that statute, it is an offense for doctors and scientists to 'infuriate' their laboratory animals.  That requires a much higher standard of welfare than merely ensuring that the animals are adequately fed and watered.  Such a minimum would probably be adequately taken care of by any institution that wanted to protect its investment.  It may be efficient for disease control to house rats in separate cages, but that would surely contravene the 1911 Act - rats are sociable by nature and keeping them in solitary confinement is, if not infuriating, at least deeply depressing and engendering of psychosis if human response to such treatment is anything to go by.

The damning report was commissioned by Imperial College itself on foot of allegations made by BUAV - the British Union for the Abolition of Vivisection who had infiltrated the darker corners of the College in an under-cover operation.  That's heartening, no? Imperial have taken on board a long list of recommendations to change their practice and channels of communication which will benefit the rodents and primates but also ensure that better science is done.  It is also deeply worrying because the failures of care were of the sort that would not have been caught by an official inspection by, say, the RSPCA.  And nobody imagines that Imperial are alone in these things - their errors are probably systemic wherever animals are kept for research.

The second significant point emerges from these recommendations: that statisticians must be brought in before any grant application is written.  They will help with the experimental design, especially in the form of a power analysis.  Power analysis helps you determine, given the expected size of the difference between the two groups, how many cases and and how many controls (hereinafter called 'rats') are required to have a reasonable stab at getting a valid answer.  You don't want to spend €1million housing and feeding hundreds of rats to show that the presence of a Y chromosome in an embryo is a reliable indicator of the development of testes later on.  €1000 would be plenty.  On the other hand, there may be statistically significant associations that only show up if you have many hundreds of observations to compare.  Maybe a good example is the association between passive smoking and lung-cancer.  In any such study there will be plenty of rats exposed to gaspers 24 hours a day who never develop tumours and some which aren't exposed but get cancer anyway.

There . is . no . point in carrying out a study that is under-resourced.  You may get a significant result but it won't be reliable; as Motyl and Nosek found last year and Begley and Ellis exposed on a much larger scale before that.  That's an utter waste of your time and tax-payers dollars.  If you're honest about the expected difference, it may turn out to require such a large sample size that every science dollar in a small country might be required to answer your questions.  Now that may be a sensible way to spend those dollars, but it's never going to happen in any real country because of the Buggins' Turn mentality that ekes out the available cash by spreading it across institutions and disciplines in a way that seems fair but gives None of the successful applicants enough money to get reliable and reproducible results.  I've been saying this for the last decade to anyone who will listen, and although people have been a little uncomfortable at my rant, they have invariably gone on to write grant applications that will a) fund the excellent existing post-doc for another couple of years b) add a brick to the wall of the lab's previous research and c) maintain credibility with the PI's statistically innumerate pay-masters.  Getting such studies published will require statistically innumerate editors and referees, but there are plenty of them.

We not got the power! There I've said it.  You may call me Bob the Thunderer.

No comments:

Post a Comment