The UK National Health Service has a recent overview of the study that was used as evidence associating talc with ovarian cancer. They come to the conclusion that while there is some evidence of an association, there is as yet no firm evidence proving causation: http://www.nhs.uk/news/2016/03March/Pages/Talc-and-ovarian-c...
Glancing at it, it seems like a very professional writeup, with many appropriate disclaimers about possible confounders. The key statistic that I have not been able to find in the paper (although I presume it's in there somewhere) is the incidence rate of ovarian cancer among the 1000 control subjects. SEER says (http://seer.cancer.gov/statfacts/html/ovary.html) that 1.3% of women will be diagnosed with ovary cancer during their lives.
Since the control subjects haven't lived their entire lives, and since diagnosis is mostly in older women (age 63), I'd guess that they'd have only something like fewer than 10 subjects on the control side with ovarian cancer. I think this means that if by chance there happened to 2 or 3 more subjects with cancer on the control side, the 1.3 increase in risk would be reversed, and thus the evidence is extremely weak! Or am I wrong, and there is some way of determining the odds ratio without knowing the true incidence rate among the controls?
That article helpfully links the paper itself, which is happily is available to read in full: http://journals.lww.com/epidem/Citation/2016/05000/The_Assoc...
Glancing at it, it seems like a very professional writeup, with many appropriate disclaimers about possible confounders. The key statistic that I have not been able to find in the paper (although I presume it's in there somewhere) is the incidence rate of ovarian cancer among the 1000 control subjects. SEER says (http://seer.cancer.gov/statfacts/html/ovary.html) that 1.3% of women will be diagnosed with ovary cancer during their lives.
Since the control subjects haven't lived their entire lives, and since diagnosis is mostly in older women (age 63), I'd guess that they'd have only something like fewer than 10 subjects on the control side with ovarian cancer. I think this means that if by chance there happened to 2 or 3 more subjects with cancer on the control side, the 1.3 increase in risk would be reversed, and thus the evidence is extremely weak! Or am I wrong, and there is some way of determining the odds ratio without knowing the true incidence rate among the controls?