Thanks for all the suggestions. I actually have the Framingham dataset, but I'm specifically looking for case-control data. It doesn't have to be big, just real.
Susana! A happy new year to you too, my friend. I will ask Guy. This sounds like a very useful teaching dataset, so I hope there aren't legal problems getting the data.
European Open Data portal: http://open-data.europa.eu
Datasets from Breslow and Day http://faculty.washington.edu/norm/datasets.html
I decided to use the Ille-et-Vilaine dataset as a starter. It shows graded risk for both alcohol and tobacco which are independent. And also a rather curious age gradient in risk factor levels not shown in Breslow and Day…