Raw correlation here is very weak evidence; it might have meant more with data from 80 years ago, but people who smoke these days either take it up while knowing the extent of the health risks, or are incredibly uniformed on basic health issues. Either way, we can expect smoking to be a very good indicator of people predisposed to high-risk behavior in general. So we should expect results similar to this even if smoking doesn't in itself cause these health effects. We would need a lot more information to control for confounding variables, and a reasonable model of unknown confounding variables, in order for this data to become meaningful evidence. Granted, they do purpose likely mechanisms for some of the effects. But their data tell us very little about the magnitude of any such effects.