My first thought was: "I wish I understood Cox proportional hazards better. I should know this stuff."
My second thought was to spend far too much time looking at the risk adjustments, and comparing their (off the top of my head) positive and negative associations with both statin use and cancer mortality. In particular spent far too much effort thinking about diabetes/heart disease, which is the last mentioned, but which drives statin use through the roof and cancer mortality down (since there is a rather important competing cause of death there).
My third thought was the devil is in the detail that current statin users are at reduced cancer mortality risk compared to past or never users. And why would you bother treating a terminal cancer patient with statins? You'd take them off ASAP.t
In other words, this is a pretty special example of volunteer bias. We came across something similar some years ago, in a study which showed people prescribed sleep aids were more likely to die within 3? (6? dunno but it doesn't really matter) months of the prescription than people on the rolls at the same docs who didn't get prescriptions.
Of course there the control group is wrong, since 100% of people bothering a doctor for sleep aids have something wrong with them (at a minimum they can't sleep, but many of them will have something quite seriously wrong with them). In the control group, chances are 90% of them go one year to the next without ever bothering a doctor about anything at all.
In this case, current statin users are basically healthy, other than having (assuming the statins are prescribe correctly), an elevated risk of (often sudden) death from a non-cancer cause. Cancer takes longer to kill and you know about it, and if you have it stuff like statins will be stopped for futility.
OK. Strap yourselves in. This is a roller-coaster of confounding factors.
Michael D. Shaw on Junkscience.com writes on statins… over-prescribed and over-hyped.
In this case the substantial confounding factor is that high cholesterol is protective against cancer. Those most likely to be prescribed statins are, if the statins have little effect (as they often do/don't) then the patients are already more resistant to cancer.