We should be very nervous about any results that depend on this many adjustments. Remember: If their measure of smoking history is crude or inaccurate, this basically means they only "half control" for it.
What I would love to see in this kind of a study is what the strength of the association is WITHOUT adjusting for anything and then show how it changes as more adjustments are added. If the association drops a lot with each adjustment, then we should be suspicious that enough controls haven't been included or the controls are too noisy.
Of course, at the end of the day, nothing substitutes for an RCT. These kinds of observational studies are worth thinking about even though they are much less informative, given that they are orders of magnitude cheaper. It's just worrisome that most popular reporting doesn't seem to recognize how much less they should be trusted.