There is evidence, but it is all of the soft kind, the statistical correlations on non-randomized sample (just like web polls). Such 'soft' evidence only indicates that smoking is a factor within some complex web of causes and effects with 'smoking related diseases'. It doesn't tell you what role smoking played in that web of causes and effects.
For example, tobacco smoke nearly doubles three key antioxidant and detox enzymes in human body: glutathione, catalase and SOD (no one knows which of 10,000+ chemical components of tobacco smoke do that). All industrial or environmental toxins you are exposed to are neutralized and removed by these enzymes. Since tobacco smoke will nearly double your detox rates, if you are exposed to industrial toxins or if you are sensitive to them at typical exposure levels, you will feel relief from smoking and have greater motivation to smoke than someone not exposed or less sensitive. Consequently, smoking becomes statistical marker for exposure (or genetic sensitivity) to industrial or environmental toxins (or similarly for other hardships), which themselves cause 'smoking related' diseases.
Hence, this association is no different than observing that people taking blood pressure and cholesterol lowering medications will have more heart attacks in coming years than those not taking them, or that people wearing sunglasses will have more sunburns than those not wearing them. While sunglasses and sunburns are in the same web of causes and effects, the sunglasses not only do not cause sunburns but they protect your eyes and skin around them from sun exposure. It is precisely because of their protective role against the cause of sunburns (sun rays exposure) that they associate with any harm caused by sun rays exposure, such as sunburns. Whenever you see media or pharma drug peddlers leap from such associations to causal relation, you better hold onto your wallet, since are a target of a junk science scam.
The statistical correlations on non-randomized (self-selected) samples of subjects (smoker, ex-smoker, never-smoker) by themselves, no matter how strong, consistent and universal still only imply that smoking is node in some complex, largely unknown web of causes and effects in which smoking related diseases are terminal nodes. That's all it tells you. It doesn't tell you anything about where in that complex web of causes and effects (largely unknown, uncharted and unquantified by researchers) the smoking node fits, let alone that smoking node must have a positive causal feed into the disease node. None of that follows from any observed correlations on non-randomized (self-selected) samples of subjects {smoker, ex-smoker, never-smoker}.
There are at least three models which could give rise to such correlations (i.e. there are at least 3 configurations in this web of causes and effects):
A) Smoking causes 'smoking related' diseases.
B) Smoking is protective/therapeutic against these diseases or their causes
C) There is some common factor(s) CF which causally contributes to these diseases and to smoking
Antismoking "science" simply declares that only model (A) exists, hence that is the explanation, period, debate over. For normal science the debate just begins at this point. The findings of correlations, implying models (A), (B) or (C) are a mere hint indicating that one or more of these three models is the mechanism behind the correlations. The correlations on non-randomized samples simply lack resolution to discern the picture beyond this whole set and to single out any of them as "The Explanation". It is the task of the much sharper instruments of hard science (experiments, randomized intervention trials) to zoom in and find out how is the web of causes and effects, underlying the observed correlations, laid out, what nodes connect to which others and how do causal effects flow between them.
Indeed, that's how the research proceeded in the very early phase of antismoking "science", back in early 1950s (the real start was actually in 1930s Germany, motivated by Hitler's antismoking hysteria; he is the spiritual father and the role model of our present antismokers). Hard science projects were funded, researchers got busy, smoking machines were humming, puffing hard on all their pipes, tens of thousands of lucky mice, rats, hamsters, dogs, monkeys,... got free cigarettes, all they can smoke, plus several times more,... All went as it should in normal science.
But then, as the first results started coming back, the sudden dead silence fell over the lands of antismoking -- all the results went the "wrong" way, the smoking animals did better, much better, in every way and from every angle they looked at. Nothing worked. Whatever poor scientist did, even pushing the smoke concentrations right up to the edge of asphyxiation, day in day out for the full lifespan of the test animals, they simply could not cause harm by the inhaled tobacco smoke. The smoking critters, always ended up living longer, staying thinner, sharper, tougher, happier... however they twisted it and tossed it around.
This ancient medicinal miracle plant, the 'most precious gift of gods' as shamans understood it in ages past, like a kind of Asimov's robot software programmed into the biochemical networks of the tobacco plant by 'ancient gods' of sun, earth and fire eons ago, simply would not do harm to its master, no matter what. It's just that good. Well, heads were rolling, new teams were brought in, still no luck.
By 1958, the mathematical genius and father of modern scientific statistics R. A. Fisher, noticed the dead silence and subsequent squirming in the land of antismoking "science" and called their bluff (pdf):
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 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...
Well, here we are, half a century later since Fisher threw the gauntlet, and antismoking "science" is still stuck in that same statistical hint loop, churning more and more correlations on non-randomized samples.
As for hard science, the experiments, that's still a no go region, a part of the town you just don't go to. See the 2005 survey of experimental field by the authority in this field, S. S. Hecht (discussed also here), with his funny euphemism for the persistently "wrong" key outcome that he refuses to name (increased longevity in smoking animals) "which complicates the interpretation of data." (p. 1489), insisting that to avoid the above "complication" that "The 4 month recovery period is absolutely necessary", i.e. what he is trying to say, but just can't get it out, is that to make sure smoking animals don't end up living longer, forcible abrupt quitting must be imposed. As to why the forcible quit makes it "work", he said "the reason for this is not clear" (it is quite clear, but he can't say it otherwise outraged smokers would string him and his paymasters on the nearest pole). If you look the dramatic survival graphs of NCI experiments brought up earlier, or this one on mice from 2005 (and similar one from 2004 on rats), the reason is pretty clear -- otherwise, the smoking test animals live longer and stay thinner, hence you can't claim with straight face that smoking caused harm to their health.
