“any claim coming from an observational study is most likely to be wrong”