Item — HN Reader

Have a citation for a risk model that uses triglycerides?

"Predicting the Thirty-year Risk of Cardiovascular Disease: The Framingham Heart Study" (http://circ.ahajournals.org/content/119/24/3078.abstract) reported that triglycerides were not statistically significant when added to a model that already contains total and HDL cholesterol:

"Standard CVD risk factors (male sex, age, SBP, antihypertensive treatment, total and HDL cholesterol, smoking and diabetes) were highly significant (0.01 level) in the multivariable model. DBP and triglycerides were not statistically significant and inclusion of LDL in place of total cholesterol did not improve model performance."

1 comments

Interesting, thanks for the citation I hadn't seen it before. It does look like pretty solid evidence, so I'm less sure of the triglyceride hypothesis than I was before.

A couple cites of interest:

http://cpr.sagepub.com/content/3/2/213.short

http://www.ahjonline.com/article/0002-8703(86)90296-6/abstra...

http://circ.ahajournals.org/content/85/1/37.short

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2664115/

There is significant overlap in the explanatory power of these variables because they are not independent. So it is possible that "high" cholesterol has covariance with triglyceride:HDL ratio in the majority of the population sample (seems likely). In that case we should see people with "high" TC still have better mortality outcomes. Which we do see: http://www.ncbi.nlm.nih.gov/pubmed/11502313

But like I said, the Framingham study does seem like reasonable evidence. I'll look over the entire thing later.

There is also this interesting development, but I don't know enough to speak intelligently about it: http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2796.2006....

It is interesting to note that apo-b levels VERY closely track HDL:Trig ratio.

Sounds good. What I've mostly seen is that LDL, total cholesterol, and triglycerides seem to be highly correlated, and adding in any one of those three (assuming you also have HDL) gives the same accuracy, so the risk modelers usually just stick with total cholesterol since more patients know it.

For ratios, this particular model is log-linear: risk = b^(w0 + w1log(cholesterol) + w2log(hdl) + ...)

So adding in variables to represent ratios would change the coefficients, but not the final risks.

For that last one, the "low cholesterol" bucket they allude to is people whose total cholesterol is <180 mg/dL. 180 mg/dL is exceptionally low, particularly for an older population. There's also a potential confounding effect when you study only total cholesterol (ignoring HDL)--your "low cholesterol" group of people includes a disproportionate number whose HDL ("good") cholesterol is also low, and who thus are actually at a high risk of heart disease according to models like the Framingham one.