Just a statistician's comment on research using small samples.

Small sample studies are not always suspect just because the sample size is small. For example, if the sample can be said to represent the population in all regards and if they find a statistically significant difference between two groups (e.g. drinkers and nondrinkers), that isn't necessarily suspect, in fact it's more impressive that they found the difference with a small sample--the difference must be a more sizable one to be detected (although NO finding should be taken too seriously until there is some form of replication--repitition-- of the results in another study).

Where small sample studies really fall down is when they find there is NO effect of something (eg. oral cancer screening) because with a small sample size, there can actually be an effect that doesn't come out to be statistically significant because of a lack of power (power grows as sample size grows) to detect the difference. One should always be suspicious of concluding that there is defniutely NO effect of somehting and basing policy on that unless the sample sizes are quite large.

There is a statistical method called meta-analysis which statistically combines results of small sample studies to increase power to detect an effect overall, and it's often used when there is controversy about whether an effect exists or not (such as there seems to be with the effect alcohol on OC). Was this study a meta-analysis? Because, if so, that is a little more convincing than a simple review of the literature, which really can't tell you much.

But even metaanalyses are only as good as the data that goes into them. If some of the studies they used were flawed, for other reasons than small sample size, then even if it is a meta-analysis there can't be any strong conclusions. Then of course there is the fact that none of these variables can be randomly assigned. It is indisputable that people who drink more than a small amount are also less likely to take care of themselves in other ways, as Brian said. Although these things can be statistically covaried out if they are measured.

But I'm not 100% convinced that alcohol only acts directly on oral cell memebranes in being a co-factor for oral cancer. It seems like the truth is we just don't know enough to know all the ways it could affect things. I just read an article that found that alcohol consumption was also related to recurrences of breast cancer, at least for post-menopausal women. So maybe there's something else at work there too.

Brian, do you happen to know a citation for the relation of HPV to breast cancer that you mentioned? I'd be interested in reading more about this.

Nelie