I'm having trouble looking up the exact study, the link they provide doesn't actually link the study.
But if I were to structure an observational smoking study, I basically track smokers vs non-smokers from the same population. Hope that my n is big enough and then simply track years non-smokers lived longer vs number cigs smoked.
Thus linear.
Help me - because I can't imagine a different way to run this.
I can not imagine that someone who smokes a few cigarettes in their 20s loses the same per cigarette as someone who smokes for years.
For example kidney injury is progressive, and by the time the kidney's functional reserve is exhausted, things will remain bad until death or kidney transplant.
Also for diseases like HIV/AIDS, the cognitive reserve influences how long it takes until HIV associated dementia sets in.
Realistically even in a large enough sample set you won't see a "jump" or "cliff". It'll just look like blurry data.
At some point, you'll be like obviously this is making a difference, but it'll be this very wide band and putting drawing a line will never be right. You'll end up saying something like >500-12,000 smoked means X minutes per cig.
The band will be so sloppy it's better to just go linear and say an overly simplistic answer.