Here is a summary of the responses that I received to my original query
regarding animal experiments.
I have not looked through all the references yet so I have cut and pasted
the individual responses.
Many thanks to
Chris Metcalfe, Stephen Senn, Jon Wright, Phil O'Neill, Robert Newcombe,
Jane Hutton, Harry Southworth, K Mcconoway, Simon Day, C Farrington, Robert
Nemeth, Nigel Stallard, Miland Joshi, Ted Harding, Andrew kenwright, Ina
Wellman and Nick Fay
Websites
papers available through the Centre for Medicines Research
http://www.cmr.org/papers.htm
The ICH S4 guideline
http://www.ich.org/ich5s.html.
The FDA has lots of guidance documents that can be accessed
from their website. There might be something on there that you'd
be interested in.
http://www.fda.gov/opacom/morechoices/industry/guidedc.htm
There are lots of welfare guidelines for the conduct of animal studies,
which might be useful to you because they are especially concerned with
minimising sample sizes whilst still maintaining statistical power in
any tests.
Such issues are dealt with as part of courses run for
scientists to qualify for Home Office licences to do animal experiments,
and these courses are run by companies such as Bioscientific Events
([log in to unmask]). Their booklets (especially regarding Module
5) contain lots of useful information and statistical techniques (e.g.
formula and tables for the calculation of sample sizes for various types
of statistical comparison, and references, such as Campbell, Julious &
Altman 1998 BMJ 311, 1145-1148).
The Association for the Study of Animal Behaviour, publishers
of the journal Animal Behaviour, have produced a standard 'Ethics in
Research on Animal Behaviour'. This contains various useful guideline
articles, including one on sample sizes, which was originally published
as: Still (1982) Anim. Behav. 30, 873-880). This journal often returns
to welfare and statistical issues, such as a recent article on sample
sizes by Ruxton (1998 Anim. Behav. 56, 511-512).
Association for the Study of Animal Behaviour and the Animal Behavior
Society have a set of guidelines for animal studies in their field,
including something brief (and some references) on sample size. It's
published every year in the January issue of the ASAB journal `Animal
Behaviour'. There are several references in there to web sites and so on
where other sources of information on the subject can be found
I've never seen any guidelines, but I believe that the Home
Office has set out statutory requirements, which impose very severe
limitations on the possibility of animal work at all, and which
severely limit numbers of animals that can be used. I suspect you're
up against a major problem here: for most of the issues that one
would want to address, either the permissible study size i.e. number
of animals would be much too small to yield any realistic power to
detect an effect of plausible size, or else non-statisticians would
succumb to the temptation of analysing data on repeated measurements
or replicated anatomical sites as if different animals, disregarding
independence. I have come across the argument that because highly
inbred strains of animals are generally used, one can be quite
cavalier about applying what is axiomatic in human studies, namely
that inter-subject variation is expected to be a dominant source of
variation. I find this argument very hard to believe - and besides,
if that is the situation, then logically any inferences drawn from
the study are no more generalisable to other animals of the same
species (let alone other species) than a single case report on a
human patient would be.
There is an EU Note for Guidance on "Good Clinical Practice for the Conduct
of Clinical Trials for Veterinary Medicinal Products". I think it has used
many
of the ideas from the human version of Good Clinical Practice. It says
nothing
about sample size (I'm afraid) but I could send you a copy if it would be
useful.
There was a piece in a recent-ish issue of Stats in Medicine on sample
sizes for small animal studies (ie, small studies, not small animals !),
the emphasis being on keeping them down to an absolute minimum. Sorry I
can't be more specific !
Broadly speaking, issues of design and sample size are much the same
whatever the experimental material, especially where sample size is
concerned.
You need to be quite specific on (at least) the following questions:
1. What type of data will be recorded (e.g. 1 observation per animal;
a time series; multivariate data; survival data; ... )?
2. What is the basic type of design envisaged (e.g. randomised control
vs. treatment; comparison of several treatments; dose/response; ... )?
3. What is the purpose of the study (e.g. investigate whether there is
a difference of effect between treatments; estimate a difference;
estimate parameters in a response curve; ... )?
4. In the case of "whether there is a difference", what degree of
difference is considered to be worth establishing, and with
what confidence do you want to establish it if it exists?
5. In the case of estimation, what degree of precision in the estimate
do you seek and what confidence level do you want to attach to that?
After that, one can begin to discuss sample size.
Gad, S.C. and Weil, C.S. (1982). Statistics for Toxicologist.
In:"Principles
and Methods of Toxicology". Ed. A.W. Hayes. Raven Press, New York is a
commonly used lookup.
Guidelines on carcinogenicity testing are in the IARC index
Statistical Methods in Cancer Research. Volume III: The Design and Analysis
of Long-term Animal Experiments.
By J.J.Gart, D.Krewski, P.N.Lee, R.E.Tarone and J.Wahrendorf
IARC-Press No.79, 1986; 213 pages
I am sure that you have already been pointed in the
direction of Michael Festing's (MRC Toxicology Unit,
University of Leicester) work. Papers I have found useful
are:
Festing, MFW (1992) The scope for improving the design of
laboratory animal experiments. Laboratory Animals 26,
pp256-267
Festing, MFW (1994) Reduction of animal use: experimental
design and quality of experiments. Laboratory Animals 28,
pp212-221
He has also published:
Festing, MFW and Lovell, DP (1996)Reducing the use of
laboratory animals in toxicological research and testing by
better experimental design. Journal of the Royal
Statistical Society Series B 58, pp127-140. Plus
discussion: pp141-153.
Which looks excellent but I've only just got hold of a
copy. Worth a look if you are collaborating in animal
studies is:
Laboratory Animals Science Association (London) /
Universities Federation for Animal Welfare (South Mimms,
Herts) (1994) Guidelines on the care of laboratory animals
and their use for scientific purposes IV: Planning and
design of experiments.
And if you are offering stats consultancy to animal
researchers an overview of some general issues is given in:
Engeman, RM and Shumake, SA (1993) Animal Welfare and the
Statistical Consultant. The American Statistician 47,
pp229-233.
Jill Mollison
Lecturer in Medical Statistics
Department of Public Health
University of Aberdeen
Medical School
Foresterhill
ABERDEEN
Tel: 01224 663123 x54089
email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|