Dull and Hill would probably disagree with you, and their work was quite impressive. If the response to survey is less than 20%, it is very difficult to infer causality. I do not think that is
Contentious.
There are no randomised trials showing a relationship between cigarettes smoked and lung cancer - these are based on surveys. A survey is a reasonable way to see a duration-response effect and it was strong and backed up by other studies. odd to argue against really