by Kathy Archibald,
says that reducing publication bias in animal research would ensure a sound
basis to move from animal studies into clinical trials (5 June, p 22). This
would be true if the results of animal studies translate directly to humans.
They do not, which is a far more important problem than publication bias.
Full publication of every animal study of the immunomodulatory drug TGN1412,
for example, would still have suggested that it was safe to proceed to clinical
trials, since the devastating response to the drug is unique to humans.
reviews of the applicability of animal results to human medicine - such
as those by Pablo Perel and others (BMJ, vol 334, p 197) and by Daniel Hackam
and Donald Redelmeier (The Journal of the American Medical Association,
vol 296, p 1731) show consistently that animal studies predict human response
incorrectly a majority of the time. In the case of stroke, is anyone seriously
suggesting that more than 150 treatments successful in animals have failed
in humans because of publication bias? Of course, this bias should be addressed.
There should unquestionably be a registration system for animal studies,
as there is for human studies. As chief executive of Understanding Animal
Research, surely Festing should be calling for this, rather than merely
commenting that "it is not inconceivable that we might move towards
a similar system".