May 30, 2012

Cardiovascular Disease

The straight dope on cholesterol – Part VI

In this post we’ll address the following concept: Why is it necessary to measure LDL-P, instead of just LDL-C?

Read Time 13 minutes

Previously, in Part I, Part II, Part III, Part IV and Part V of this series, we addressed these 7 concepts:

     #1What is cholesterol?

     #2What is the relationship between the cholesterol we eat and the cholesterol in our body?

     #3Is cholesterol bad?

     #4 How does cholesterol move around our body?

     #5 How do we measure cholesterol?

     #6How does cholesterol actually cause problems?

     #7Does the size of an LDL particle matter?


(Not so) quick refresher on take-away points from previous posts, should you need it:

  1. Cholesterol is “just” another fancy organic molecule in our body but with an interesting distinction: we eat it, we make it, we store it, and we excrete it – all in different amounts.
  2. The pool of cholesterol in our body is essential for life.  No cholesterol = no life.
  3. Cholesterol exists in 2 formsunesterified or “free” (UC) and esterified (CE) – and the form determines if we can absorb it or not, or store it or not (among other things).
  4. Much of the cholesterol we eat is in the form of CE. It is not absorbed and is excreted by our gut (i.e., leaves our body in stool). The reason this occurs is that CE not only has to be de-esterified, but it competes for absorption with the vastly larger amounts of UC supplied by the biliary route.
  5. Re-absorption of the cholesterol we synthesize in our body (i.e., endogenous produced cholesterol) is the dominant source of the cholesterol in our body. That is, most of the cholesterol in our body was made by our body.
  6. The process of regulating cholesterol is very complex and multifaceted with multiple layers of control.  I’ve only touched on the absorption side, but the synthesis side is also complex and highly regulated. You will discover that synthesis and absorption are very interrelated.
  7. Eating cholesterol has very little impact on the cholesterol levels in your body. This is a fact, not my opinion.  Anyone who tells you different is, at best, ignorant of this topic.  At worst, they are a deliberate charlatan. Years ago the Canadian Guidelines removed the limitation of dietary cholesterol. The rest of the world, especially the United States, needs to catch up.  To see an important reference on this topic, please look here.
  8. Cholesterol and triglycerides are not soluble in plasma (i.e., they can’t dissolve in water) and are therefore said to be hydrophobic.
  9. To be carried anywhere in our body, say from your liver to your coronary artery, they need to be carried by a special protein-wrapped transport vessel called a lipoprotein.
  10. As these “ships” called lipoproteins leave the liver they undergo a process of maturation where they shed much of their triglyceride “cargo” in the form of free fatty acid, and doing so makes them smaller and richer in cholesterol.
  11. Special proteins, apoproteins, play an important role in moving lipoproteins around the body and facilitating their interactions with other cells.  The most important of these are the apoB class, residing on VLDL, IDL, and LDL particles, and the apoA-I class, residing for the most part on the HDL particles.
  12. Cholesterol transport in plasma occurs in both directions, from the liver and small intestine towards the periphery and back to the liver and small intestine (the “gut”).
  13. The major function of the apoB-containing particles is to traffic energy (triglycerides) to muscles and phospholipids to all cells. Their cholesterol is trafficked back to the liver. The apoA-I containing particles traffic cholesterol to steroidogenic tissues, adipocytes (a storage organ for cholesterol ester) and ultimately back to the liver, gut, or steroidogenic tissue.
  14. All lipoproteins are part of the human lipid transportation system and work harmoniously together to efficiently traffic lipids. As you are probably starting to appreciate, the trafficking pattern is highly complex and the lipoproteins constantly exchange their core and surface lipids.
  15. The measurement of cholesterol has undergone a dramatic evolution over the past 70 years with technology at the heart of the advance.
  16. Currently, most people in the United States (and the world for that matter) undergo a “standard” lipid panel, which only directly measures TC, TG, and HDL-C.  LDL-C is measured or most often estimated.
  17. More advanced cholesterol measuring tests do exist to directly measure LDL-C (though none are standardized), along with the cholesterol content of other lipoproteins (e.g., VLDL, IDL) or lipoprotein subparticles.
  18. The most frequently used and guideline-recommended test that can count the number of LDL particles is either apolipoprotein B or LDL-P NMR, which is part of the NMR LipoProfile.  NMR can also measure the size of LDL and other lipoprotein particles, which is valuable for predicting insulin resistance in drug naïve patients, before changes are noted in glucose or insulin levels.
  19. The progression from a completely normal artery to a “clogged” or atherosclerotic one follows a very clear path: an apoB containing particle gets past the endothelial layer into the subendothelial space, the particle and its cholesterol content is retained, immune cells arrive, an inflammatory response ensues “fixing” the apoB containing particles in place AND making more space for more of them.
  20. While inflammation plays a key role in this process, it’s the penetration of the endothelium and retention within the endothelium that drive the process.
  21. The most common apoB containing lipoprotein in this process is certainly the LDL particle. However, Lp(a) and apoB containing lipoproteins play a role also, especially in the insulin resistant person.
  22. If you want to stop atherosclerosis, you must lower the LDL particle number.
  23. At first glance it would seem that patients with smaller LDL particles are at greater risk for atherosclerosis than patients with large LDL particles, all things equal.
  24. “A particle is a particle is a particle.”  If you don’t know the number, you don’t know the risk.
  25. To address this question, however, one must look at changes in cardiovascular events or direct markers of atherosclerosis (e.g., IMT) while holding LDL-P constant and then again holding LDL size constant.  Only when you do this can you see that the relationship between size and event vanishes.  The only thing that matters is the number of LDL particles – large, small, or mixed.


Concept #8 – Why is it necessary to measure LDL-P, instead of just LDL-C?

In the growing list of reasons why I used to refer to myself as “chick-repellant” in college, I have a confession to make: I find the topic of statistical concordance and discordance to be so exciting, I sometimes have a hard time containing myself. This may explain the paucity of girlfriends in college. Let me use an example to illustrate the distinction between these terms.  Let’s say you want to predict the change in home prices in the following year (I used to model this for a living).   There are at least a dozen parameters linked to this, including: GDP growth, unemployment, interest rates (both short term and long term, though to different degrees), housing inventory (i.e., how many houses are on the market), housing absorption (i.e., how quickly houses go from being on the market to being sold), major stock indices, and consumer confidence.  Historically, from the mid-1990’s until about the fourth quarter of 2006, this worked like clockwork.  While each of these variables had differing strengths of predicting changes in home prices, they all moved together.  For example, when GDP growth was robust, unemployment was low, interest rates were modest, housing inventories were about 60 to 90 days, etcetera.  All of these variables pointed to a predictable change in home values.

Around Q42006 (i.e., last 3 months of 2006), one of these variables began to deviate from the others.   The details aren’t important, but the point is one variable began to suggest home prices would fall while the others all pointed to a continued rise.   Prior to Q42006 these parameters were said to be concordant – they all predicted the same thing – either up or down.  By 2007, they became discordant – one variable said the sky was falling while others said everything was fine.

This was true on the “micro” level, too. [What I described above is called “macro” level.]  As a lender, it should be very important to know the risk of each and every loan you make (clearly this was part of the root problem in the age of mass securitization).  Will this person pay the loan back or will they default?

Same game here, but now a new set of even greater variables.  As a lender, if I want to know if YOU will default, I will want to know a lot of things about you, such as your agency credit risk scores, your bank account activity, payroll activity, how much you’re borrowing relative to the value of your house, where your house is located, and about 50 other things (literally).

Not surprisingly, the same thing that happened on the macro side happened on the micro side.  It became difficult to predict who would default and would not default because there were so many variables to consider and lenders didn’t know which ones were still predictive.  The models that predict default are very sensitive to the balance of these inputs.  When all of the variables are concordant, their accuracy is prophetic, as was the case from the mid-1990s until late 2006.  When some variables become discordant with each other, especially variables that were historically concordant with each other, really bad stuff happens, as became evident to me, personally, one Thursday afternoon in November 2007.  It became clear the sky was about to fall.  And, of course, it did.


What does real estate have to do with atherosclerosis?

Fortunately, predicting heart disease is a little easier than predicting changes in home prices.  It’s not perfect, of course, but it’s pretty good.  Why is it not perfect? For one thing, we can’t do the “perfect” experiment.  The “perfect” experiment would look something like this:

Take 100,000 people and randomize them into four matched groups, A, B, C, and D.  Wave a magic wand (you can see why this experiment hasn’t been and won’t be done) and give the folks in Group A an LDL particle concentration of, say, 700 nmol/L; those in Group B you give 1,200 nmol/L; those in Group C you give 1,600 nmol/L; and those in Group D get 2,000 nmol/L.

In our dream world, due to the randomization process, these four groups would be statistically identical in every way except one – they would, thanks to our magic wand, have a different number of LDL particles.  We would follow them without further intervention for 10 years and then compare their rates of heart disease, stroke, and death.

There are some areas in medicine where we can do such experiments.  But, we can’t do this experiment for this question.  Even when we do the next best thing — give people a drug that lowers their LDL-P and measure the impact of this intervention — there is always a chance we’ve done something in addition to “just” lowering LDL-P.   If you’ve been reading this series, you no doubt know my thoughts on this: while other factors are likely to be involved the pathogenesis of atherosclerosis (e.g., endothelial “health”, normal versus abnormal inflammatory response) the primary driver of atherosclerosis is the number of apoB trafficking lipoproteins in circulation, of which LDL particles are the vast majority.

The data below should further clarify this association.


What do concordant LDL-C and LDL-P values look like?

Among the two largest studies tracking the association between cholesterol and atherosclerotic mortality are the Framingham study and the MESA trial (the two largest trials were AMORIS and INTERHEART).  The figure below, which I’ve graciously borrowed from Jim Otvos, shows the risk stratification of LDL-C (top) and LDL-P (bottom) from the Framingham study and MESA trial, respectively. As you can see, conveniently, LDL-C values in mg/dL are about 10x off from LDL-P values in nmol/L.   In other words, in the Framingham population, the 20th percentile value of LDL-C was 100 mg/dL, while the MESA trial found the 20th percentile of the population to have an LDL-P concentration of 1,000 nmol/L.  As you will see by the end of this post, this “rule of the thumb” should never be used to infer LDL-P from LDL-C.


Cut-off points for LDL-C and LDL-P
Image credit: Jim Otvos

If this were always the case – that is, if LDL-C and LDL-P were always concordant – we could conclude that LDL-C and LDL-P would be of equal value in predicting heart disease.  Obviously this is not the case, or I wouldn’t be making such a fuss over the distinction.  But how bad is it?


What do discordant LDL-C and LDL-P values look like?

The figure below, from the Journal of Clinical Lipidology, shows the cumulative incidence of cardiovascular events (e.g., myocardial infarction, death) over time in three sub-populations:

  1. Those with concordant LDL-P and LDL-C (black line);
  2. Those with discordant LDL-P and LDL-C (LDL-P>LDL-C, shown by the red line);
  3. Those with discordant LDL-P and LDL-C (LDL-P<LDL-C, shown by the blue line).

This analysis was done using a Cox proportional hazard model and was adjusted for age, sex, and race.  The steeper the line the more people in that sub-population died or experienced adverse cardiac events relative to other sub-populations.  In other words, the folks in the red group had the worst outcomes, followed by the folks in the black group, followed by the folks in the blue group.



What can we infer from these data?

First, we confirm what I alluded to above.  Namely, that a non-zero percent of the population do not have LDL-C and LDL-P values that predict the same level of risk.  However, and perhaps more importantly, we get another look at an important theme of this series: LDL-P is driving atherosclerotic risk, not LDL-C.   If LDL-P and LDL-C were equally “bad” – even when discordant – you would expect the blue line to be as steep as the red line (and both to be steeper than the black line). But this is not the case.

Let’s look at these data parsed out another way.  Below we see the four possible subgroups, from the top:

  1. Not low LDL-P, low LDL-C (red line);
  2. Not low LDL-P, not low LDL-C (yellow line);
  3. Low LDL-P, low LDL-C (black line); and
  4. Low LDL-P, not low LDL-C (blue line).

Note that “low” is defined below the 30th percentile and “not low” is defined as greater than 30th percentile for each variable.   This figure is even more revealing than the one above.  Again, it demonstrates the frequency of discordance (about 20% in this population with these cut-off points), and it shows the importance of LDL-P’s predictive power, relative to that of LDL-C.

In fact, though not statistically significant, the highest risk group has high LDL-P and actually has low LDL-C (I’ll give you a hint of why, below) while the lowest risk group has low LDL-P and not-low LDL-C.  *This is not a typo.


MESA LDLp vs LDLc 4 groups

The highest risk and lowest risk groups are those with discordant LDL-C and LDL-P.  The high risk group has high LDL-P and low LDL-C, while the lowest risk group has high LDL-C with low LDL-P. Only a minority of physicians would know that there is a segment of the population with elevated LDL-C who are at low risk! The same conclusion will be drawn from the next study.

Let’s look at an even longer-term follow up study, below.  This study followed a Framingham offspring cohort of about 2,500 patients over a median time period of almost 15 years in each of the four possible groups (i.e., high-high, high-low, low-high, and low-low) and tracked event-free survival.  In this analysis the cut-off points for LDL-P and LDL-C were the median population values of 1,414 nmol/L and131 mg/dL, respectively. So “high” implies above these values; “low” implies below these values.  Kaplan-Meier survival curves are displayed over a 16 year period – the steeper the slope of the line the worse the outcome (survival).


Survival curve
Image credit: Cromwell et al., 2007

The same patterns are observed:

  1. LDL-P is the best predictor of adverse cardiac events.
  2. LDL-C is only a good predictor of adverse cardiac events when it is concordant with LDL-P; otherwise it is a poor predictor of risk.

Amazingly the persons with the worst survival had low (below median) LDL-C but high LDL-P.  The patients most likely to have high LDL-P with unremarkable or low LDL-C are those with either small LDL particles, or TG-rich / cholesterol poor LDL particles, or both (e.g., insulin resistant patients, metabolic syndrome patients, T2DM patients).   This explains why small LDL particles, while no more atherogenic on a per particle basis than large particles, are a marker for something sinister.


Populations where LDL-P and LDL-C discordance are even more prevalent

As I described above, the discordance between LDL-P and LDL-C is exacerbated in patients with metabolic syndrome. The figure below, MESA data, again borrowed from Jim Otvos, presents this difference in an elegant way.  The horizontal axes show LDL-P concentration in the usual units, nmol/L.


Otvos ADA
Image credit: Jim Otvos

Patients with LDL-C between 100 and 118 mg/dL (i.e., second quartile of risk: 25th to 50th percentile) are shown without metabolic syndrome (top) and with metabolic syndrome (bottom).  In the patients without metabolic syndrome, LDL-C under-predicts cardiac risk 22% of the time, consistent with the population data I have shown you earlier.  However, when you look at the patients with metabolic syndrome, you can see that 63% of the time their risk of cardiac disease is under-predicted.  Again, not a typo.

There are so many subsets and cut-off points that I could devote ten more posts to showing you every one of these analyses.  Let me finish this point with the most recent, hot-off-the-press (actually, still in press in the American Journal of Cardiology, though you can get a preprint here) analysis of which Tom Dayspring is one of the authors.


Evaluation of Low-Density Lipoprotein Particle Number Distribution
Image credit: Malave et al., 2012

These data were collected from nearly 2,000 patients with diabetes who presented with “perfect” standard cholesterol numbers: LDL-C < 70 mg/dL; HDL-C > 40 mg/dL; TG <150 mg/dL.  However, only in 22% of cases were their LDL-P concordant with LDL-C.  That is, in only 22% of cases did these patients have an LDL-P level below 700 nmol/L.

Remember, LDL-C < 70 mg/dL is considered VERY low risk – the 5th percentile.  Yet, by LDL-P, the real marker of risk, 35% of these patients had more than 1,000 nmol/L and 7% were high risk.  When you do this analysis with the same group of patients stratified by less stringent LDL-C criteria (e.g., <100 mg/dL) the number of patients in the high risk group is even higher.

The real world tragedy: 90-95% of physicians, including cardiologists, would bet their own lives that persons with an LDL-C < 70 mg/dL have no atherosclerotic risk. 

Tim Russert, shortly before his death, had his LDL-C level checked.  It was less than 70 mg/dL.  Sadly, his doctors didn’t realize they should also have been checking his LDL-P or apoB.  The figure below, which is from one of Tom Dayspring’s presentations, shows data from this study of nearly 137,000 patients hospitalized for coronary artery disease between 2000 and 2006.  As you can see, LDL-C fails to even reasonably predict cardiovascular disease in a patient population sick enough to show up in the hospital with chest pain or outright myocardial infarction.


Insulin Resistance Lipids & Lipoproteins

Why are LDL-C and LDL-P so often discordant?

Think back to what you learned in a previous post in this series.  LDL particles traffic not only cholesterol ester but also triglycerides.  Each and every LDL particle has a variable number of cholesterol molecules which, because of constant particle remodeling, is constantly changing.  In other words, of the several quadrillion LDL particles floating in your plasma, no two are carrying the exact same number of cholesterol molecules. It takes many more cholesterol-depleted LDL particles than cholesterol-rich LDL particles to traffic a given cholesterol mass (i.e., number of cholesterol molecules) per volume of plasma (i.e., per dL).  Core cholesterol mass is related to both LDL particle size (the volume of a sphere is a third power of the radius — it can take 40-70% more small particles than large LDL particles to traffic a given cholesterol mass) and the number of TG molecules per LDL particle.

TG molecules are larger than cholesterol ester molecules, so as the number of TG molecules per particle increases, the number of cholesterol molecules will be less – in a very non-linear manner. Regardless of size it takes many more TG-rich LDL particles (which are necessarily cholesterol-depleted) to traffic a given cholesterol mass than TG-poor LDL particles.  The persons with the highest LDL particles typically (though not always) have small LDL particles that are TG-rich.  These are incredibly cholesterol-depleted LDL particles.



Take a look at this figure below from the 2011 Otvos et al. paper I referenced above.  It’s a scatterplot of each data point (i.e., patient) in the study. The solid red line shows perfect concordance between LDL-P and LDL-C.  The dashed red lines show a +/- 12% margin on each side.  Look at how many dots (remember: each dot represents a person) lie OUTSIDE of the dashed red lines. Now look again.


MESA LDLp vs LDLc J Clin Lip2011

When people argue with me about why it’s unnecessary to check LDL-P or apoB because it’s much easier and cheaper to check LDL-C, I like to remind them of what Clint Eastwood would probably say in such a situation:  You’ve got to ask yourself one question: Do I feel lucky? Well, do ya, punk?”


  1. With respect to laboratory medicine, two markers that have a high correlation with a given outcome are concordant – they equally predict the same outcome. However, when the two tests do not correlate with each other they are said to be discordant.
  2. LDL-P (or apoB) is the best predictor of adverse cardiac events, which has been documented repeatedly in every major cardiovascular risk study.
  3. LDL-C is only a good predictor of adverse cardiac events when it is concordant with LDL-P; otherwise it is a poor predictor of risk.
  4. There is no way of determining which individual patient may have discordant LDL-C and LDL-P without measuring both markers.
  5. Discordance between LDL-C and LDL-P is even greater in populations with metabolic syndrome, including patients with diabetes.  Given the ubiquity of these conditions in the U.S. population, and the special risk such patients carry for cardiovascular disease, it is difficult to justify use of LDL-C, HDL-C, and TG alone for risk stratification in all but the most select patients.
  6. This raises the question: if indeed LDL-P is always as good and in most cases better than LDL-C at predicting cardiovascular risk, why do we continue to measure (or calculate) LDL-C at all?

Photo by Aldric RIVAT on Unsplash

Disclaimer: This blog is for general informational purposes only and does not constitute the practice of medicine, nursing or other professional health care services, including the giving of medical advice, and no doctor/patient relationship is formed. The use of information on this blog or materials linked from this blog is at the user's own risk. The content of this blog is not intended to be a substitute for professional medical advice, diagnosis, or treatment. Users should not disregard, or delay in obtaining, medical advice for any medical condition they may have, and should seek the assistance of their health care professionals for any such conditions.


  1. I got here first, so I better post.
    Beautifully laid out. Thanks for the reminder about the power of concordant/ discordant variables. Grateful to have gutted through this series! Thank you for the time to pound this out.

  2. So I guess the upshot so far, Peter, is that 1 – everybody should get themselves the advanced lipid test; 2 – if you are discordant, double down on the low-carb diet and start drug therapy ASAP.

    • Almost. Yes, everyone needs an NMR, at least once. If you are discordant, NMR should be the preferred method for following you. We’ll get into treatment subsequently.

    • > Everyone needs an NMR, at least once.
      > If you are discordant, NMR should be
      > the preferred method for following you.

      Really astounded by this series, Dr. Attia, and I look forward to rereading the whole thing and taking notes. I feel like I only absorbed a fraction of the material you presented.

      When it comes to tests, I hear two mentioned quite frequently in the context of, “Get either an NMR or a VAP test. The standard panel is useless…” You do not mention VAP (that I’ve seen). Do you feel it is inferior to the NMR? I’ve not seen a good explanation for one test over the other, so I’m curious to hear your thoughts on the matter.

      • I address this in part II or III. To be clear, VAP is NOT a good test to get, despite the fact that I used to track my progress until last year. It estimates apoB. Your best bet is NMR (many companies can send out to LipoScience, the only company able to do this with FDA approval), or go with one of the companies can can actually measure apoB directly.

    • I look forward to subsequent post on treatment. I hope you can include some thoughts on the larger view as well – i.e. recognizing that the cardiovascular system is part of a larger system. Given the increasing awareness of potential side-effects of statins (Ref Golomb studies and as recently acknowledged by FDA), how does one weave treatment strategies that maximize total health?

  3. Good presentation again Dr. Attia. Question: A lot of interesting studies do comparisons of pattern A (big fluffy particles) vs pattern B (small dense particles) LDL-p without looking at total particle #.

    Can a person make assumptions about particle number based off of total cholesterol and pattern shifts? For example, say a study examined one group of people with pattern B (small dense LDL-P) with a total cholesterol of X. They now give a dietary intervention and the people with pattern B now change to pattern A but with the SAME X amount of cholesterol as before. Can we assume their particle # went down since their cholesterol stayed the same but now the cholesterol is being housed by larger LDL-P which means total particle # should have theoretically gone down?

    I ask because a lot of studies investigate particle size but not total particle number.

    • You’re right, most studies are fixated on LDL size, not count. I’d like to say that I can always “guess” correctly which way LDL-P moves with a change in LDL-C, TG, and particle size, but I can’t. There are certainly clues, as I explain in this post. But at best, that’s all they are.

    • If you shift particle size upwards it will have a larger volume and thus will have to be carrying more lipid molecules. Sotypically the larger LDL will carry more cholesterol molecules and it takes less cholestrerol-rich LDLs to traffic a fiven LDL-C value (mass). However if the particles become larger because they are TG-rich, they will be cholesterool-depleted particles and it takes many more chol-poor than chol-rich LDLs to traffoc a given cholesterol mass

  4. It’s all coming together now! Will you talk more about Lp(a)? Any evidence that it might be even more important than LDL-P (both atherogenic and thrombogenic?). Any data on discordance between LDL-P and Lp(a)?

    Thanks again for your work!

    PS: How is the cookbook coming along?

    • I spoke a bit about Lp(a) in an earlier post. It’s a particle like LDL, but with an apoprotein called apo(a). For most folks the number of Lp(a) particles is a fraction of LDL-P, though we can currently (i.e., for about another year) only estimate Lp(a)-P from 2 other features. Cookbook is a work in progress. NuSI is priority #1, 2, and 3 right now.

  5. We have a Health for Life program at work where everyone is encouraged (bribed) to get their blook work done every year. I asked them a few months ago why we don’t measure LDL-P in this program. I am going to have another go at it and link this post as justification. Lots of great data rich graphs in this one.

    Maybe I missed it but are you saying that Metabolic syndrome is driving the diconcordance between LDL-P and LDL-C? Did you state whey High LDP-P/Low LDL-C was worse than High Hight? Is it because there is something slightly more protective about LDL-C?

    • Discordance exists at a baseline for reasons I don’t entirely understand, probably 20-30% of the population. However, in MS, the discordance goes WAY up, almost certainly related to the TG-CE swap in LDL particles. In my personal experience, with hundreds of patients, I’m seeing about 30-40% discordance.

    • Well fatty liver is very common eating fructose and high carbohydrates diets. It can’t be cured in a few weeks or months, maybe that why some still have problem with cholesterol.

      The VCU findings may provide researchers with potential new targets for treatment and also allow clinicians to further refine how they assess cardiovascular risk and develop ways to reduce it in individuals with a more aggressive form of nonalcoholic fatty liver disease called nonalcoholic steatohepatitis, or NASH.

      In the study, published in the May issue of Cell Metabolism, the team has shown that there is not only increased production of cholesterol but a decreased expression of the receptor that takes up cholesterol from the blood. This would be expected to both enhance cholesterol output from the liver and reduce its removal, thereby making it more available to enter blood vessels and contribute to cardiovascular disease. The liver not only makes cholesterol, but also takes up cholesterol from the blood.

      “This indicates that there is excessive cholesterol production in the liver when one develops fatty liver disease,” said lead investigator Arun Sanyal, M.D., professor and chair in the Division of Gastroenterology, Hepatology and Nutrition in the VCU School of Medicine.

  6. Again – brilliant and much needed comprehensive reporting on the truth of Cholesterol science. Your effort and time put into propagating the real concepts behind “what people believe” is much appreciated.

    Unfortunately, widely dispersed productions like that recent HBO documentary about the “Weight of America” (or whatever it was called), overlooks the most important parts.

    You are an asset to our community, Peter!

    • I appreciate your kind words. Change takes time. I’m focused on making sure our kids grow up in world where people can decide what to eat to based on real data, in a way few of us can do today. Sort of like smoking today. All I ask is that folks get a fair shot at making a choice for themselves — based on real data. Kick-off scientific meeting for NuSI is Friday…

  7. Great stuff.

    “Regardless of size it takes many more TG-rich LDL particles (which are necessarily cholesterol-depleted) to traffic a given cholesterol mass than TG-poor LDL particles.”

    Is this where diet/lifestyle comes in?

  8. 10/10 points for style, substance, and clarity. 😉
    Thank you once again for compiling and filtering the information, Peter.
    It takes a lot of time and effort to do so, I’m painfully aware. 🙂

    So, is this the punchline?… or are we getting more on this subject?

    • Thank you, Vasco. Yes, I think we have a few topics left in this series before we can hand out the graduation caps and call ourselves certified lipidologists. I hope folks can hang on for a few more rounds.

  9. Awesome, as usual.

    One thing you could add for clarity is to explain the figure “MESA data, again borrowed from Jim Otvos” has LDL-P on the horizontal axis. This is not noted on the figure and might make it easier for readers to see why you say LDL-C under-predicts risk.

    For my annual physical, my doctor runs some standard blood work. If, the next time I see him, I say I want my LDL-P checked, would he be likely to comply? How can I convince him, and what other atypical measures should I request while I’m at it?


    • Great. Just updated. I hope you can get him to comply. Everyone deserves at least one NMR. If they are concordant, they can probably get away with subsequent LDL-C, provided no major metabolic changes take place.

  10. OK, LDL-P is the best current predictor of CVD risk. But how good is it? How much of the total risk for any individual is “explained” by his or her LDL-P number? That gets to how important it is for any individual to worry about.

    Perhaps an even better way to ask this question would be to ask how much life expectancy is sacrificed by having high LDL-P to varying degrees? IIRC, except at severe extremes life expectancy is decreased by at most several months for those with unfavorable conventional lipid panels, which is one reason why people like Dr. Nortin Hadler advise well people not to even get the tests. I’m curious how many lost months of expected lifespan high LDL-P predicts, and how confidently we can know that.

    This sort of information seems important if people are to rationally decide how much to worry about this, and if it’s worth it to them to incur the potential problems and risks of treatment to lower LDL-P. I know you’ll be covering treatment later, and I’ll be curious to see what you say about the NNT (number needed to treat) for LDL-P lowering treatment in various populations, a related question.

    Thanks for the excellent series.

    • Fortunately, we have pretty decent data for these questions, but it’s deserving on an entire post, not a short response. Fret not, though. Reducing a person’s LDL-P from 2,000 to 1,000 does a lot more than buy them a few months. We’re not talking about esoteric (morally suspect) chemotherapy. This is real primary and secondary prevention.

    • There is so much good here, I feel guilty about the fact that I am struggling with this series. But . . .

      Bill asks an important question. Reading the summary of previously established points, I thought that Dr. A would have submitted significant evidence that LDL-P is the best predictor in Part IV, but the text there is thin on the point. Yes LDL-C is not a great predictor, and yes Dr. A has now overwhelmed us with data showing that LDL-P is better than LDL-C as a predictor and that LDL-P size does not matter. The assertions are mostly only relative. The case will be harder to make that LDL-P is the be-all and end-all predictor. I believe that a ratio of LDL-P to HDL-P would be better than LDL-P alone. I bet APO A-1 / APO B would beat LDL-P alone (a test I got for around $40 not long ago). Or Dr. William Feeman’s CRF (colesterol retention fraction), which is (LDL-C minus HDL-C)/LDL-C plotted on a graph with systolic BP (see Bowling Green study). Or smoking plus most anything. I guess I don’t see why it is significant to know that LDL-P is by itself the best predictor compared to any single other predictor, if in fact it is not that great. It’s like knowing that Green Bay gained the most yards in 2012 but, oh yeah, they did not win the Super Bowl or even get there.

      I am anxious for the grand train wreck that comes when Dr. A tells Jimmy Moore and the low-carb/ paleo crowd that have achieved > 2000 LDL-Ps (yes, I am in that club, thanks to FH) that they have to get to an 1100 LDL-P, and good luck getting there with anything but statins or unproven crap like zetia (or maybe an Ornish diet – me, I choose to eat food). As TO says, get your popcorn.

      • Lots in here…can’t address this in a short response, which seems all I have time for these days. Certainly ratios carry predictive power, and apoB to apoA1 is possible the best (except for possibly LDL-P to HDL-P). It’s pretty rare (read: I have never seen a case) to see someone with a LDL-P of 2,000 (95% percentile), yet ratio of LDL-P to HDL-P of 20 (probably 10th percentile). It’s possible, but boy, it’s an outlier. What would it mean if you were that fellow? Who knows…
        Maybe you’re right. Maybe in such a case you have so many HDL particles that RCT can fix all the problems of too many LDL particles. We just don’t have the kind of data to make anything other than a best guess. So, I see your point about “offense” vs. “defense” but I hope you realize that atherosclerosis is more complex that football. Think about this way, does the team with the greatest ratio of points for/against always win? Of course not. It probably predicts more success, but it’s not a guarantee.
        I’m not sure where the assertion comes that Jimmy Moore et al. all have LDL-P > 2,000 just because you might? My guess is very folks out there even know their LDL-P.

    • EPIC:


      Hence they assessed the relationship of LDL-P and particle size as measured by a relatively new technique—nuclear magnetic resonance spectroscopy—with the risk of future CAD in more than 25 000 subjects with moderately elevated LDL-C in the European Prospective Investigation into Cancer and Nutrition (EPIC)-Norfolk.

      They matched 1003 cases of individuals who developed CAD during a six-year follow-up with 1885 control subjects and calculated the odds ratios (ORs) for future CAD. They also evaluated whether LDL-P could improve upon the Framingham risk score to predict CAD.

      LDL-P (OR 2.00 for top vs bottom quartiles) was more closely associated with the occurrence of future CAD than LDL-C (OR 1.73) and was related to CAD on top of the Framingham risk score even after adjustment for LDL-C.

      But after adjustment for HDL-C and triglycerides, LDL-P was no longer more predictive than LDL-C. Nor was LDL particle size associated with CAD risk, following adjustment for LDL-P.

      end snip.

      Reviewing EPIC data, information from the standard lipid panel, if interpreted in a sophisticated way, yielded better predicitive value thal LDL-P alone. More interesting to me, top quartile LDL-P had a 2 to 1 odds ratio, vs. a 1.73 to 1 odds ratio for top quartile LDL-C vs bottom quartile. OK, LDL-P was a little better, but not really that great — bottom quartile LDL-P still had significant residual risk. My Framingham score would be half of what it is if I did not have high LDL-C. Dr. A, the absence of a clear statement as to the degree to which LDL-P improves on the predicitive value of LDL-C makes these analyses misleading. LDL-C sucks, and LDL-P is a bit better. You are putting LDL-P on a high pedestal it does not deserve.

    • Thanks for replying and thanks for this series.

      Jimmy Moore has gotten an NMR, and reported LDL-P of 2130. I see a significant number of low-carbers seem to have spikes in their LDL-C and LDL-P that would appear to be driven by diet. They are not necessarily FHers like me (but I have seen wide ranging LDL-P, from 1500 to 2800, and cannot yet tie it out to diet). Commenter MacKillop below refers to a double-digit percentage of folks on low-carb diets who see very high LDL-C. Mr. and Mrs. Jaminet have blogged at some length on the issue.

      I see the view expressed by some that this phenomenon is due to ApoE4, but I don’t buy it (I am a 3/3).

      • The question we don’t know the answer to is if an LDL-P of 2,000 in someone who eats no carbs is the same as an LDL-P of 2,000 in someone who does. I had breakfast with Eric Westman today and we discussed this topic. Eric makes a pretty compelling case that these 2 states are not, in fact, the same thing. I think we can safely say we don’t know the answer. At least I don’t. I’ll keep looking for clues, though.

    • Glad to hear of your conversation w Eric re implications of high scores on high vs low carb. Isn’t this something that a study could be constructed on? Hard to do double blind, but might be able to at least get clues from population studies? Not sure diet itself is the variable to correlate with, maybe hsCRP, maybe something else?

  11. Great article! So this is begging for a sequel which answers the question: What kind of food makes LDL-P and LDL-C discordant?

    I suppose given that even most researchers don’t know jack about what you revealed to us, they use LDL(-C) as one of their primary markers in studies. The number of nutritional studies that use LDL-P and LDL-C there I assume to be really small.

    • To your last point, exactly. To your first question, the type of foods that lead to metabolic syndrome are exactly the type of foods that increase the proportion of discordance.

    • What I would be interested in is where oxidized LDL fits in… I’m doing my homework on pubmed right now 🙂

    • What would be the draw back of treating a no carb person with LDL-P in the higher levels(assuming no adverse physical reactions to statins)and later on finding out that higher levels of LDL-P in no carb dieters is different and not harmful… For every ying there is a yang. I could be off base here but I think the goal should be fixing any accumulated problems caused by SAD diets and trying to reset, then strictly avoid refined, processed and unnatural “food products” and assume that whatever the body does from there is what it is intended to do. Our bodies have come millions of years figuring out what is the right state and in the absence of SAD should be able to figure that state out again

      • Gabe, I can’t argue with you. I really wish I could say I KNEW the answer to this question. If I told you I did, I’d be misleading you.

    • Russ,
      There is much room for research. Like Peter says we just don’t know right now. My plan of attack is rid myself of Insulin Resistance and lower LDL-P to under 1000 (with statins if necessary) once both those goals are reached i’ll consider myself “reset” and hopefully stop medications. I’ll continue low carb paleo lifestyle and what my body does from it’s “reset” state i’ll assume is what it’s supposed to. The question is once I reach my goal does the body really get to start over or will there always be some underlining bit of damage that doesn’t get reversed. Will the statins that helped me “reset” cause issues. No way to tell for now..I guess time will tell.

  12. Follow up question on Lp(a). Standard tests of Lp(a) measure cholesterol not particle number. What can we say about a high Lp(a) cholesterol? Is it the same principle as LDL-P vs LDL-C?

    • Currently it’s only possible to measure Lp(a) cholesterol content and mass, but not particle number. If both are low, Lp(a)-P is low; if both are high, Lp(a)-P is high; if mixed it’s harder to tell. Folks are working on this right now. Assays expected to be able to count them directly by next year.

  13. When can we expect the part that says sugar and other “crap” increases LDL-P? Awesome work!

    • Sounds like you’ve already figured it out! Actually, you’ve already got a big hint in that direction based at least one typical cause of discordance: LDL particles carrying too much TG and not enough CE. Guess what causes that?

  14. This was my father’s story. Had “perfect” labs and resting ECG two months before having an early spring round of golf followed by a bagel (!!) and then dropping dead on the living room floor at the age of 64.

    • Very sorry to hear this. I can almost assure you his LDL-P, had it been measured, would have been very high, despite normal LDL-C.

  15. Okay, I’m all in convinced. A few installments back I bristled a bit that our history of testing for lipids has sent us barking up some seriously wrong trees, but this evidence for particle testing is compelling.

    Sincere gratitude for the series, both to you and to Drs. Dayspring and Otvos for permission to share their slides in it. I can imagine the work to do the series, but the additional time and energy to be so present here in the comments is really above and beyond.

    I’m also grateful to have been turned onto the work of Drs. Dayspring, Otvos and Dall, and through your direction, have become a regular at Lipidaholics and other features at Lecturepad. But your synthesis of their work here has added a level of analysis that’s been so very helpful. I can’t believe I got it all for free as I’d expect to pay many hundreds (or thousands) for something like this to either of the institutions that educated me.

    • I’m glad you can see why I consider my so fortunate to be able to call these folks mentors and teachers. Their generosity with their time humbles me and it’s my pleasure to “pay it forward” as best I can.

  16. Oh boy, great post! I can hardly wait for the next ones. Thanks again for putting this together. Can I infer that insulin resistant individuals should shoot for TG levels even lower than the 150 currently used as the standard for “optimum”? What should be the desired TG levels for this population?

    • Absolutely. In fact, the most strict lipidologists would argue that a “physiologic” TG level is closer to about 50 mg/dL, well below the 150 touted as “normal.” I would absolutely concur with this.

  17. First, best post ever. I saw some of the graphs on one of the testing web sites and almost posted it last week. Good thing I didn’t steal your thunder but it makes a ton of sense. Especially pictorially.

    One question – if you get an NMR and find out your relationship between LDL-C and LPL-P is concordant (or discordant) will that relationship continue to exist in the same way over time? i.e. you will always be discordant or concordant as far as the two variables hence need an NMR test or rely on LDL-C respectively?

    • This is a really good question, Tom. I don’t have enough longitudinal data to really comment. Clearly the relationship between LDL-P and LDL-C changes over time. No doubt about this. Constant re-modeling of particles, changing levels of TG, changing levels of CE. The question, of course, is if you are concordant today, under what circumstances can I be confident you are next year? On a personal level, it’s not worth even risking it. I check LDL-P every single time. At the “policy” level, this is a different matter and probably warrants further investigation.

    • Dr. Lipid, thanks for the insight and I have viewed a couple of your video interviews online. Fascinating and well done.

      One question, IF (capitals on puspose) I have a High LDL-P and a low LDL-C why would I want that discordant relationship corrected? Seems like that is the best place to be. I am assuming you mean the converse of low LDL-P and high LDL-C?

  18. A couple of comments in the last couple of segments spoke of a “damaged” particle as the possible instigator of vessel wall penetration and retention, the implication being that the LDL particle was damaged previous to penetration and retention. Just yesterday I listened to an interview with Chris Masterjohn who seemed to be saying a similar thing. I’ve been wondering about what this “damage” might be and what the evidence is for it.

    Again, the way I heard it, the particle is damaged while out in the circulation, previous to its penetration and retention in the endothelium, and it’s the damage that accounts for its penetration.

    Any input from anybody?

Facebook icon Twitter icon Instagram icon Pinterest icon Google+ icon YouTube icon LinkedIn icon Contact icon