I can attest to that. I had weaker health a few years ago (lack of activity mainly) but I could kick a regular schedule of activity to get back on track. An accident broke my health below the self help threshold. When you're worried you can't jog, even walk, it's hard to bootstrap the exercice-health cycle again.

The same could be said of almost any study that tried to show this. It is unethical to design the study in the way you're suggesting (i.e. stop people exercising to see if they get cancer).

They did however:

> Body mass index adjustment modestly attenuated associations for several cancers, but 10 of 13 inverse associations remained statistically significant after this adjustment.

So even for two thin people, exercise seems to have helped. I legitimately don't see how you could design a study that you'd be satisfied with.

The correct study is one where some participants get an intervention to make them exercise more, and some are left alone. There's no ethical issue, but running such a study would be difficult and expensive.

Adjusting for BMI does not solve the problem, because there are many things besides BMI (unmeasured confounders) which affect both cancer risk and willingness to exercise.

And BMI is a pseudo science measure anyway. Its been proven to be a poor indicator in about every disease many times before.

the problem is that on noisy data, that adjustment of the variables isn't really as statistically clean as maths would say.

it's the same with red meat causes cancer studies. they say they remove smoking and lack of exercise and still wave the statistical significance.

it's very questionable how they removed the influence of these variables, since meat eaters do smoke way more, and do not exercise, and care about their health less.

Actually, I think noise is a red herring, and that worrying about noise (instead of worrying about unmeasured confounders and incorrect causal structures) is how epidemiology got into this mess in the first place.

Statistically significant does not prove anything, and it does not mean the difference would be large either.

An interventional study is ideal, but unlikely to happen.

I would have been happy if the authors just tried to factor in other characteristics like socioeconomic status, education level, etc. It looks like they didn't do that but instead just observed an "association" which isn't terribly interesting or informative.

I will add that this is the sort of thing that peer review theoretically ought to stop, or at least force authors to include caveats for. Unfortunately, studies with this sort of methodology make up most of the so-called studies published, so they can't really do that.

I haven't read the paper itself, but even if it didn't explicitly note these caveats, it would be assumed by any reader in its intended audience.

The problem is the press representation of the papers as something they're not.

These types of papers should be published, because they are necessary to decide which hypotheses to test with interventional studies.

John Oliver explains?

https://www.youtube.com/watch?v=0Rnq1NpHdmw

John Oliver pointed out that the media usually distorts what the studies say. He failed to flag the problem of junk science in high-prestige journals.

(Edit: I just re-watched his segment. He does talk about P-hacking and perverse incentives, which is at least a reasonable-size chunk of the problem, just not the failure mode for this particular study.)

I still agree with you, he did not place enough of the blame on scientists who are unable or unwilling to properly communicate the implications of their work.