Intended for healthcare professionals

Rapid response to:

Papers Research pointers

Taking folate in pregnancy and risk of maternal breast cancer

BMJ 2004; 329 doi: (Published 09 December 2004) Cite this as: BMJ 2004;329:1375

Rapid Response:

What's in a name?

In this interesting follow-up of an old trial, the authors tell us that "this randomised trial was of high quality" and that "the trial was double blind". They also tell us that "Tablets were supplied in six colours, two of which contained folate in 0.2 mg and 5 mg daily doses. The tablets were kept in numbered drawers and distributed in sequence." If the tablets had different colours for different treatments the trial wasn't double blind, as any trialists seeing the tablets would know the treatment. If treatments were given sequentially, they were not random. Also, there was no allocation concealment, an important indicator of high quality, if treatments were given sequentially, as the treatment for the next subject would be known.

I would not criticise the authors because a trial carried out in the 1960s does not meet current standards, but we should use technical terms such as "double blind" and "randomised" to mean what it is agreed that they mean, not something else.

Another curiosity of their report is that there appear to be four times as many subjects on placebo as on either active treatment. I cannot get access to the original paper, but this seems a rather extravagant design, even for the 1960s. Were some of these “placebos” other treatments? If so, is it possible that these other treatments could reduce the risk of death, rather than folate increasing it?

There are 30 tests of significance here, including the confidence intervals. One is statistically significant at the 0.05 level, P=0.02. A simple Bonferroni correction would suggest that the P value for the composite hypothesis that folate increases risk of death is 30 times 0.02 = 0.6. The authors do not tell us why of all the possible causes of death they picked breast cancer, indeed, they tell us they "had no prespecified hypothesis that taking folate supplements in pregnancy would increase the risk of cancer." One wonders how many other causes they looked at and which have not been mentioned. Of course, the Bonferroni method is crude and the tests are not independent, but in the absence of any more appropriate analysis by the authors it is all the reader can do.

Competing interests: None declared

Competing interests: No competing interests

10 December 2004
J Martin Bland
Prof. of Health Statistics
University of York, YO10 5DD, UK