Hi Sujit,
Thanks for posting. For others who want to listen - start at about 20 minutes 40 seconds into the recording for this piece.
1. What was the study design?
This is a case control study. Cases were women with ovarian cancer and loyalty cards. Controls were other women with loyalty cards (am not sure what other criteria were used to select them)
2. What epi measurements would you use to assess the usefulness of the prediction?
Odds ratios - the exposure was purchase of analgesia and antacid medication. I would imagine that these two exposures would be tested separately.
If possible it would be useful to have data from all loyalty card users, ideally stratified by age and sex, to then separately test the sensitivity and specificity of analgesic and antacid purchases to see how useful these could be as a predictor. It might be possible to add analgesic and antacid purchases together to form a composite exposure measure to try to increase specificity for ovarian cancer diagnosis.
3. Could we test this prediction in an RCT? How?s
If women with loyalty cards agreed to be contacted (or agreed for their GPs to be contacted) and agreed for researchers to retrieve information about them from the national cancer registry, then the researchers could analyse data from loyalty cards owned by women over 50 years. Half of these women would be randomly selected as the intervention group and half as the controls. Women in the intervention group with increasing analgesic and antacid use would be contacted by researchers who would draw their attention to their changing use of medicaton and ask them to consider visiting their GP if they were unsure what the underlying cause of this was. Some information about ovarian cancer would be included. Women in the control group would not be contacted. Follow up would involve analysing stage at ovarian cancer diagnosis comparing women in the intervention to those in the control group.
There would be several problems with this study. Firstly, by recruiting the women, both cases and controls, and telling them about ovarian cancer, their awareness would be raised and they would be more likely to seek help early than women who had not been advised of the possibility of ovarian cancer. Secondly, women might become very anxious if receiving letters from researchers advising them to consult their GP. Most women buying analgesics and antacids will not have ovarian cancer. Thirdly the primary care physicians involved would be likely to react in very different ways to the women presenting, increasing the sample size needed to detect any difference. They also might feel that extra work and patient anxiety was being created for women with, for example, knee pain and indigestion for years, who are no more likely to have ovarian cancer than the general population.
Best wishes
Judith
