Summary of Cholesterol research

So this post aims to provide evidence for the skeptical reader (Larry) as to why the status quo on both the intake and measurement of cholesterol has been woefully off the mark. For those who are satisfied with summaries from those with a pedigree, I’d skip this document and click on the links provided in this blog post.

My own education on this topic started with a video I saw by Gary Taubes on the history of this controversy. I then endeavoured to try and follow his citations, which is difficult because many of his citations aren’t easily available online… and I’m too cheap to buy them or even his book. ūüôā

However, we’ll save his research for later (scroll down to the bottom)… because we’ll agree that although he cites his data, it isn’t the data itself. So let’s continue with primary data available. It is interesting to note that Gary’s approach and mine are different in focus. Gary prefers to retrace a historical account of how current medical institutions got it wrong and then doubled down on their error. I prefer to look at how current data throws confusion on the status quo.

As always in science, it’s the data that doesn’t fit that should cause the most alarm.

Epidemiological contradictions

We’ll start with what is generally the weakest form of data (as Gary points out beautifully), that of epidemiology. This data focusses on the outliers of Ancel Keys’ 7 country study, the landmark that made fat out to be an evil in the first place. Incidentally the best summary of interpreting Ancel’s work, given later published critiques by Yerushalmy and Hilleboe, is beautifully cited in this blog post. She gets my full respect for a wonderful piece.

The first outlier that should be considered while we’re fighting epidemiology with epidemiology is summarized by this paper about the French paradox:

Although several explanations for the paradox are explored (wine drinking, diet-heart hypothesis), the authors conclusions are admittedly by their own writing inconclusive.

The French outlier isn’t the only important outlier. However it’s important to note that that these authors and many others wish to preserve the notion that we should be allowed to interpret the data WITHOUT the outliers, which would indicate that high fat diets (high in cholesterol intake) should be bad for the heart. No-where in their writing do they acknowledge the potential that their primary hypothesis may be measuring confounders.

Let’s continue, but before we leave epidemiology, let’s just note another contradiction since this debate started with an assessment of the notion that¬†reducing caloric intake reduces one’s risk of heart disease. These studies find exactly the opposite correlation:

“Men who had a greater caloric intake or a greater caloric intake per kilogram of body weight were less likely to develop CHD manifest as myocardial infarction (MI) or CHD death, even though men of greater weight were more likely to develop CHD.”

This supports the notion that if you can find calorie-neutral ways of keeping off the weight, you’re in a good position.

Of course the study also finds that 2 out of the 3 populations in the study found benefit from CHD in consuming more starches (a mixed result). But let’s leave these studies be, as unless epidemiological studies have the kind of statistical significance of the smoking data, survey food data should be considered extremely unreliable. Tit for tat, and lets move on.

Moving on to non-survey data.

Chipping away at the status quo

Part of this debate is about the contention of the usefulness of total cholesterol.

The next contradiction comes from Petah Tikva medical centre confirming that higher cholesterol levels were associated with a decrease in morbidity for the elderly. Although they make reference to the status quo knowledge relating to middle aged men, their own data presents a departure from convention.
“In very elderly hospitalized subjects, increased levels of serum total cholesterol and albumin may be associated with reduced mortality risk.”

So maybe this presents two hypotheses: either the elderly who survive are a selected population of strong hearts who benefit from increased cholesterol intake OR yet again the correlation we are observing in the middle-aged here is yet again the result of confounders.

The subjects of the study were a mix of men and women with a mean age of 81 years deviating by about 6 years either side.

What about the men of say 50-70? This says nothing about them. Well according to Taubes:

Researchers involved with the Framingham Heart Study found that in men and women 50 and older, ‚Äútotal cholesterol per se is not a risk factor for coronary heart disease at all.‚ÄĚ

Unfortunately it’s difficult to find the original study, so if you don’t want to take Taubes’s word (and the NYT editorial team), the best I can do is the 30 year follow up to framingham which is written with some caveats.
“After age 50 years there is no increased overall mortality with either high or low serum cholesterol levels.”

Update: Found Framingham.

What should be clear by now is that taking the advice of trusting in the predictive powers of total cholesterol numbers alone is starting to look unwise. One should also be clear that what we’re investigating is the predictive power of total cholesterol for incidents and mortality relating to CHD (coronary heart disease). As can be seen from the following study, total cholesterol has no predictive powers when it comes to strokes.
“This analysis of the EUROSTROKE project could not disclose an association of total cholesterol with fatal, non-fatal, haemorrhagic or ischaemic stroke. HDL cholesterol however, seemed to be related to stroke differently in men than in women.”

and as we shall see in the next section, HDL is the starting clue to the prediction conundrum (thanx Gary).

Just to give food for thought, it’s worthwhile reviewing the results of this paper.

“Although it correlated with the number of severely narrowed coronary arteries per subject, the serum total cholesterol level did not correlate with the percentage of 5 mm segments of coronary artery with severe narrowing.”

In other words, the narrowness of arteries due to plaque build up did not correlate with serum total cholesterol. Of course, although many segments were analysed, this was only 40 patients.

Multivariate Regression Analysis.

The following data from the Israel Ischemic Heart Disease Study is key to sandboxing serum total cholesterol.
“Multivariate analysis of the data to adjust for possible confounding effects of additional mortality risk factors demonstrated that total cholesterol made no independent contribution to total mortality, but that the contribution of low HDL to mortality persisted after adjustment. CHD mortality consistently increased with rising concentrations of total cholesterol.”

So let’s unpack this. In this study, they did in fact link total cholesterol to mortality risk of CHD. However, despite CHD related deaths representing 37% of all deaths in the study, total cholesterol fails to have any predictive benefit on total mortality when adjusted for confounders! Low HDL on the other hand remained a predictive indicator of risk (conversely high HDL remains an indicator of health).

This helps corroborate Gary’s assertion that HDL is 4 times the predictor compared to that of LDL, which serum total cholesterol is a proxy for. Of course it may be that HDL predicts many ailments unrelated to CHD, all the more case not to primarily focus on total cholesterol.

I actually find the wording of the conclusions of this paper overly timid: “These results indicate that, particularly in older age-groups, measures designed to increase HDL cholesterol may prove as valuable in preventing CHD as those designed to reduce low density lipoprotein cholesterol.”

This, despite the fact that HDL performed significantly better! Perhaps it’s because they didn’t want to rock the boat with one of it’s funders?? Conspiracy theory anyone? hehe.

I don’t have access to Gary’s citations from the original HDL studies conducted under NIH auspices, but if we extend him a professional courtesy for the moment (and he corrects himself when he’s wrong), he cites that the original studies came to the conclusion that total cholesterol numbers were meaningless (see his video below).

How do we reconcile the fact that some studies attribute weak prediction of total cholesterol (compared to HDL) and some studies suggest no predictive power what-so-ever?

Reconciling Total Cholesterol, HDL & LDL

The problem (as highlighted by Gary’s video), is that high HDL levels are associated with decreased risk. However, HDL is a component of total cholesterol… so if you act to increase your HDL levels, you are also increasing your total cholesterol levels!?! So how should total cholesterol be interpreted?

If increasing your total cholesterol levels means increasing equal rates of HDL and LDL, or only increasing HDL, then not much can be said about the risk factors associated with total cholesterol.

In populations where increased total cholesterol is correlated with increased levels of LDL then this may be a problem. However, it’s more complicated by the invention of new tests and data. As the Mayo Clinic writes in 2011 about the new LDL particle size test, the extent of increases in one’s LDL is relatively dangerous in proportion to the increase in one’s small LDL particle size count. Again, if one’s total cholesterol has been increasing due to an increase in ones large LDL particle count, then this is significantly less risky.

Update: Although HDL is a good predictor compared to say total cholesterol, studies have not managed to record a benefit of increasing one’s HDL-C without lowering the other stuff.¬†

This is only the basic story, as the role of triglycerides and VLDL etc has not been discussed.

However, if the assertions follow, then it would be silly for anyone with the means of tracking small LDL particles and HDL levels, to instead be tracking serum total cholesterol.

Updated, particle size no longer relevant

Even the distinction between large and small LDL particle count, based on research from the top people in the field is potentially no longer the relevant indicator. According to famed Thomas Dayspring, the counts of various LDL particles can differ from total counts of LDL and the composition of whether LDL is cholesterol depleted and carrying triglycerides can be of critical importance.

Update: This man’s eloquence is backed up by the following study¬†¬†and¬†

Additionally we are learning the roles of HDL particle counts also.

And as an aside, it’s worth listening to this interview, for fuller explanations, as well as his first interview on this channel.

Gary’s research

So the thing that set me off on my cholesterol research was Taube’s video here:

A more formally written account can be found in his New York Times article here:

Although I can’t find the original framingham stuff, some of the contentiousness of the original data, even before it was assumed settled by policy makers can be found here:

…and diet

Given how all this got started with comments about the sustainability, effectiveness and health of atkins style diet, I thought I’d post some of the latest data comparing the most popular diets out there with a view to optimizing for sustainability and health for the most at risk group: Those displaying insulin resistance and hence most highly at risk for metabolic syndrome.

From Stanford medicine.

This talks cites a number of papers highlighting the most interesting parts of the new literature.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: