Building a lag period into the analysis can correct for treatment of cancer symptoms with a drug before the cancer is diagnosed.  The substantial association of iron medication with cancer of the large bowel was reduced with a 1-year lag and totally removed with a 2-year lag.