Epidemiology by itself isn't a junk science. It crosses into junk science when someone leaps from observed statistical associations on non-randomized samples to wishfully ($$$) declaring causal relations. Such associations are at best a hint that there may be causal relation, but one needs hard science, such as randomized trials or animal/human experiments to find out what kind of links (e.g. causal or protective/therapeutic) connect those correlated variables.
That's how it is done in normal science, you use statistical hint to make hypothesis that is followed up with hard science. But antismoking "science" is stuck on the same hint since 1950.
And it is not for lack of trying hard science. There were thousands of experiments done since then. The problem was that they all went the "wrong" way -- the smoking animals live longer, perform better on cognitive tasks, get cancers less often, etc. What can poor scientific mercenaries do, when their bosses want the opposite result, but stick with what works, parrot the statistical hints disguised as "science." This was so unusual pattern that already in 1958, the father of modern statistical methods, famous British mathematician R. A. Fisher noticed it and wrote (pdf; this article also contains a very readable exposition of the sample randomization topic):
"Most of us thought at the time, on hearing the nature of evidence, which I hope to make clear a little later, that a good prima facie case had been made for further investigation. But the time has passed, and although further investigation, in a sense, has taken place, it has consisted largely of the repetition of observations of the same kind as those which Hill and his colleagues called attention to several years ago. I read a recent article to the effect that nineteen different investigations in different parts of the world had all concurred in confirming Dr. Hill's findings. I think they had concurred, but I think they were mere repetitions of evidence of the same kind..."
Yet, the antismoking "science" still rests its case squarely on the same kind of soft/junk science that Fisher objected to over half a century ago.