[greatpatton]

"how effective are these sorts of statistical methods?"

Exactly they are so many variables to take into account that each population subset is becoming very small. I would love to have the insight of a stats nerd on the study.

Btw if you look at the raw data, they are for instance less cases per subject in the high consumer than in the low or none consumer (table 2), but they get a higher hazard ration, meaning that other correction were taken into account. It is very difficult to interpret such results.