Authors: LM Schwartz et al.

Statistician's report by Hazel Inskip
 

This is a very interesting paper and I enjoyed reading it. It has the strong merit of having actually asked women what they think rather than making assumptions about their reactions. I do, however, have a few comments that perhaps the authors could consider.

1. It would be helpful for the non-American reader to have details of the screening system in the States. In UK, I understand that screening does not begin until 50 years of age, it is repeated every three years and there is little if any counselling about false positives. Such different practices in different countries mean that the results are hard to generalise, but it is important that the reader understands the background of this particular survey.

2. I am afraid that the description of the weighted analyses was not clear to me and I wonder if this could be dropped. Adjusting for the demographics of the sample seems to worsen the sample's representation of the socio-economic variables (and presumably vice versa). Given that this paper has been submitted to the BMJ (rather than an American journal) it may be better to present the crude results and comment on the sample (and its weaknesses). Again, the lack of generalisability in the point above re-inforces this point that the weighted analysis adds little.

3. I have a slight problem with the last sentence on page 6. I think that it is trying to explain confidence intervals, but it does not seem quite correct. The strict definition is that if the survey was repeated 100 times then in 95 of them the given confidence interval should include the true mean of the population (assuming that the sample is representative). I would suggest that the error (due to random variation only) of 4-6% should be given but not the explanation, as there is little point when the sample is not representative of the population as a whole anyway.