In the same study, there were considerably more exposed cases when we looked at all cancers combined.  Here the odds ratios and confidence intervals for 50 vs. 10 controls per case were virtually identical.