Considering LDL to be 'bad' and HDL to be 'good' is a vast oversimplification. LDL, if it's a high-particle-count (so small/dense) has a tendency to invade the epithelium of artery walls (maybe as deep as the intima - the next layer down). This causes inflamation, drawing macrophages. Those macrophages fill up with lipids - to the point of becoming dysfunctional, and become part of the problem. (lipid filled macrophages look like foam, hence the name 'foam cells'). LDL doesn't have to be small/dense, but it is often enough that, yeah 'bad' works.
HDL as 'good' is another story. We can raise HDL levels via niacin (if you can stand the 'flush') but that doesn't change the CVD risk profile. It *can* be good indeed; small HDL particles can slip into the epithelium or evevn intima and pull lipids out of those poor macrophages (AKA delipidation). Once they've filled up they either bring that back to the liver (or enterocyte) or hand it off to a LDL particle, which brings it back to the liver.
It's absurdly complicated with lipid exchanges going on between LDL-HDL, HDL-HDL and both of these with, well, lots of other kinds of cells. Even thinking of it as 'forward' and 'reverse' transport is an oversimplification - and we know nearly nothing about what HDL is otherwise doing.
I've been trying to wrap my head around the lipid transport system for the last year at least. The parts that broke my brain is that more lipids are carried by ethrocytes (RBCs) than anything else - and that all particles (including chylomicrons) carry cholesterol because the proteins need it to form a spherical shape. Research often breaks complex things down in order to first understand each part, before trying to understand the whole, but the lipids system is a good example of 'everything everywhere all at once'.
Sorry to have gone on, but please at least stop calling HDL 'good'.
[/rant]
Oh, good analysis of that report's atttempt to pull causality out of correlation. Sometimes there's an obvious 'natural experiment' that allows that (e.g. cutoff dates for shingles vaccine elegibility) but I don't think there have been many of those (I can think of maybe three). Otherwise, all correlation gets you is hypothesis generation.