Sunday, September 20, 2009

Mortality and cholesterol

People might enjoy O Primitivo's latest excel plot here. I did. I also suspect the effort involved in this is stupendous. Thanks!

Peter

6 comments:

Adam said...

Unbelievably thorough and reassuring information. I'm a competitive athlete and evangelical hyperlipider and this confirms the justification for many of the choices I've made. Awesomely powerful, and probably beyond the mitigation and denial of causation of low cholesterol and disease that the statin purveyors will attempt.

Ed said...

I thought we knew that oxidized LDL led to atherosclerosis, and that there are 3 major influences on oxidized LDL levels: LDL counts (not weight), proportion of LDL that is PUFA, and key anti-oxidant status.

The correlational data put forward (close your eyes, take a deep breath, and repeat after me, "correlation does not prove causation") uses a measure of LDL weight, not LDL count, and further does not control for the other known confounders (granted, that would be very difficult across the broad data set that he found).

I would have to think that there are yet more extremely powerful confounders, like, say, health care access and stress levels. Eyeballing the chart, I see Sierra Leone doing poorly, hrmm is that caused by low cholesterol?

I am very sympathetic to the argument that cholesterol is unnecessarily vilified, but I still desire a more nuanced view and interpretation of known data.

The question this data seems to answer is "if I know absolutely nothing about you except for your total cholesterol density, to what degree of confidence can I predict your likely mortality rate from a number of diseases?" and the answer seems to be very interesting.

The question this data does NOT answer is "would I live longer if I were to raise or lower my LDL density or count?" (yeah that's like 4 questions in one, sorry :-)

Anyway, this is very interesting stuff and I plan on sharing it, I would just caution people from over-interpreting what it tells us.

Ed said...

Err I'm re-reading my comment, and I have to correct myself, this presentation doesn't even use LDL weight, it uses total cholesterol weight. *shrug* That's what I get for not previewing before I post.

Peter said...

Hi Ed,

I would agree, this is observational and should be kept as hypothesis generating fodder. But it certainly does not suggest to me an hypothesis that low cholesterol is beneficial on a mortality basis.

I'm also very open to the idea that oxLDL, like HbA1c, is a marker of a set of conditions (especially hyperglycaemia) which are also likely to generate multiple abnormalities in the clotting cascade. We are talking thrombosis in a significant number of coronary events. Of course it is equally possible oxLDL may be the villain, but with the track record of the lipid hypothesis to date I'm not holding my breath on that one.

The other problem with O Primitivo's observations is that it is damned difficult to raise you TC. And if we get a drug to do it we might well just end up with another torcetrapib... My wife has a cardiologically Nirvanous TC, presumably LDL too, and gets trashed by every bug going. Obviously eating very close to the same diet as me. Short of filling her with sugar (which precipitates very interesting palpitations, no thanks) I don't see any way of raising her LDL, diet wise.... There has been so much talk on the paleo LC front about high TC. I'm just hoping there is no problem for those who do develop low cholesterol on LC eating. I doubt this is a real problem but, if it is, no one is worrying about it (except me, as my wife coughs next to me!).

And of course the LDL values won't have been measured, they're almost certainly calculated.

So lots of holes, but as you say, still interesting stuff.

Also, as an add on, the curves do not suggest an hypothesis which includes a particularly long life expectancy for someone with a TC of 8.6mmol/l..... Two ends to every U curve!

Peter

. said...

"Which cholesterol level is related to the lowest mortality in a population with low mean cholesterol level: a 6.4-year follow-up study of 482,472 Korean men." - http://aje.oxfordjournals.org/cgi/reprint/151/8/739.pdf

Peter said...

Ta Ricardo

P