402 ‒ NMR blood analysis: how mortality risk and more can be assessed from a single blood sample

1. 0:00 – 0:11

   Intro

   1. PAPeter Attia0:00

      [upbeat music] Hey, everyone. Welcome to the "Drive" podcast. I'm your host, Peter Attia. [upbeat music]

2. 0:11 – 12:50

   How investigating a flawed 1986 cancer test led to the development of NMR (nuclear magnetic resonance) lipoprotein testing

   1. PAPeter Attia0:11

      Jim, so great to be with you again. We were just talking a minute ago that w- I, I had forgotten briefly that we were together once in 2013. But, um, in many ways, it, you know, th- this is, this is a pretty wonderful opportunity for me to sit down with someone whose work I've been following for, uh, literally 15 years. It was, it was, uh, May. I still remember, it was May of 2011 when Tom Dayspring introduced me to your work, and I began voraciously consuming everything you had written, um, in my personal obsession to better understand the fields of, of lipidology. Um, again, I think there, there can't be that many people listening to us that haven't at some point probably had an LDL-P or HDL-P test done, and yet most of them will never realize until now, presumably, that that test goes all the way back to you, and you are the, the creator of that. So maybe give us a little bit of a s- a, a, a story, your, your journey. Uh, you did a PhD in physical chemistry or?

   2. JOJim Otvos1:17

      In, in chemistry.

   3. PAPeter Attia1:18

      Chemistry, okay.

   4. JOJim Otvos1:19

      And, and biochemistry. I, uh, was in academia for 20 years doing the kinds of things that people did in academia and do in academia using NMR spectroscopy as a structural tool. So NMR is a very common, uh, structural tool. You can find NMR machines in every chemistry department in the country. And, uh, so I had, I had a- appointments in chemistry at the University of Wisconsin-Milwaukee, and then moved in 1990 to North Carolina State University. So I was, uh, basically doing my thing, m- minding my own business, using NMR for the, the usual purposes.

   5. PAPeter Attia1:58

      And for the listener, we are going to explain how NMR works because for people who didn't take, you know, chemistry and might not remember it, we'll, we'll come back to it, but I don't wanna interrupt-

   6. JOJim Otvos2:07

      Okay

   7. PAPeter Attia2:07

      ... now to do that.

   8. JOJim Otvos2:08

      [chuckles]

   9. PAPeter Attia2:08

      Yeah, we'll come back to it.

   10. JOJim Otvos2:09

       Yeah, I'm not sure that that's terribly relevant, but we can, we can-

   11. PAPeter Attia2:12

       Yeah, yeah

   12. JOJim Otvos2:12

       ... talk about it. But, but, um, anyway, the point is that, that NMR was and is very useful for a particular purpose, which is helping organic chemists determine the structure of molecules that they synthesize, for example. Um, and I was using it to study biomolecules, so it was more challenging than small organic molecules, and, um, trying to get at understanding what was going on at the active site of zinc-contain- zinc, zinc metalloenzymes. But anyway, uh, and so I was funded to do that and, um, had an NMR machine in Milwaukee to do that research. And then in 1986, um, there was a paper published in The New England Journal with a lot of hoopla, and in particular in the NMR, uh, field, um, people paid attention to this. I didn't normally read The New England Journal [chuckles] of Medicine ever. Um, but it claimed that a very simple NMR test could tell whether somebody had cancer or not, plus or minus, irrespective of whether it was this cancer or that cancer. Uh, nothing really in the paper that laid a, a, a mechanistic foundation for why this relationship should be. It simply measured a couple prominent signals in the NMR spectrum of blood plasma and measured how wide the signal was halfway up the signal. If it was narrow, uh, you had cancer. If it was not narrow, [chuckles] you didn't have cancer. And, and, you know, normally this wouldn't be given much attention, but it was published in The New England Journal.

   13. PAPeter Attia3:50

       Yeah.

   14. JOJim Otvos3:51

       And so everybody, uh, who had NMR machines [chuckles] was interested in, in, in seeing if they could replicate this. Uh, I was in a chemistry department not associated with a medical school, so I had no idea how to get my hands on plasma if I wanted to play around with this. But, um, [chuckles] I went across the street to a hospital and talked the people in the lab to giving me six leftover plasma samples from healthy people and, um, popped it into my NMR machine. And sure enough, half of the signals were narrow and half were broad. Did half of these people have cancer? No. These were-- These three people were women who had just given birth, so pregnancy was one false positive that was, uh, given in this, in New England Journal paper. So that, if it wasn't for that sort of linkage to something that seemed consistent with what was published, I probably never would have taken another spectrum of, of, of plasma. But, um, just out of scientific curiosity, uh, we started measuring, uh, plasma and noticed that the signal that was this supposed cancer diagnostic didn't look like a nice symmetrical NMR signal. It had lumps and bumps and shoulders and, and so what was up with that? And pretty quickly, when we started to ask where does this signal show up and what, what's the, what are the molecules that are giving rise to these signals, it was clear that these were signals from the lipids and lipoprotein particles. And so we then, uh, act very serendipitously got funding from Siemens Medical Systems, uh, basically $100,000 after me giving a one-hour presentation for what, what I might learn with $100,000. So that was a pretty cool opportunity for a professor who, [chuckles] you know, had to go through a lot more hoops to get funding. Um, so we had the wherewithal to get, uh, samples from people with and without cancer and then separate the major lipoproteins, VLDL, LDL, and HDL. And we looked at those signals and noticed that the VLDL signals were always to the left of the LDL signals. The LDL signals were always to the left of the HDL signals. And it was the superposition of these signals and bec- and their relative concentrations differing that gave rise to the different shapes of this composite mixture signal that you would see in a plasma sample. So it was obvious that, that the signal was coming from lipoproteins. So what's up with the narrow signal meaning cancer? Well, that turns out to be due to the fact that those sig- those signals from those people were from people with higher triglycerides and lower HDL cholesterol, and the two-- combination of those two things made the signal narrower. [chuckles] So this was nothing to do with cancer per se. It was to do with the l-lipoproteins, the lipid levels of people with cancer. Twenty years before that, people had published that people with cancer on the average have higher triglycerides and lower HDL cholesterol. So, um, anyway, we, we published in, in, in 1990 or '91 a paper in Clinical Chemistry that showed that the NMR signals from isolated VLDL, LDL, and HDL from people with and without cancer didn't differ at all. So there was nothing distinct about whether the sample came from a cancer patient or not. But what was very reproducible, and we didn't understand why, was this phenomenology of the VLDL signals not showing up in exactly the same place as the LDL signals or the HDL signals. So, um, we, we got the brilliant idea, which I didn't think was very brilliant, I thought it was obvious, to use this putative cancer signal as a source of information about the concentrations of lipoproteins in the blood. So with a fairly simple low-tech NMR spectrum that anybody could do with any NMR machine, uh, you could generate this signal whose shape and amplitude could be used to deduce the concentrations of the VLDL, LDL, and HDL that were giving rise to that composite signal. Okay? And, and so we had been funded by Siemens, and w-we published a paper, uh, in, in, I think 1991 saying that, yes, it was feasible that one could get VLDL, LDL, and HDL simultaneously from this simple NMR spectrum. So that seemed to have some advantage over the usual way of measuring triglycerides and LDL and HDL cholesterol, uh, via normal chemical methods. So, um, so that's as far as we thought we could go, and if it wasn't for the fact that Siemens people were advising me that I never would have filed for a patent on [chuckles] how to do this, um, what did I know about patenting and what did I care about commercialization? Um, but I did file a patent, and the patent was issued. And, um, so, so, you know, if, if this seemed to be something useful and clinically useful, then it would have to be clinically translated, and that would have to be via some sort of commercial entity because there were no NMR machines in clinical laboratories. In fact, there are no [chuckles] NMR machines to this day in clinical laboratories. So, um, you know, we needed to figure out how to transition NMR spectroscopy into clinical laboratory medicine, and we needed a commercial vehicle to do that. So that, that was an idea that evolved, uh, over the early 1990s to about 1995, '96. Had a couple NIH grants to support the analytic development. And what we didn't expect to be able to do but found that we could was not only to differentiate VLDL, LDL, and HDL, but the different size subspecies that make up what we call total VLDL, total LDL, total HDL, smaller, medium-sized, larger particles. And because these signal positions are so close to each other, it just didn't seem feasible that you would be able to work backwards from the composite signal and accurately get the concentrations of the subspecies. But we, uh, kn- by that time, I had been reading the literature a bit and, uh, Ron Krauss at, at UC Berkeley at Donner Laboratory was showing that via a quite laborious separation method, gradient gel electrophoresis and others, that you could differentiate LDL, uh, on the basis of size and found that people with a prevalence of small, dense LDL, um, had greater cardiovascular risk at a given LDL cholesterol level than somebody with, with large LDL. So this was something that was very interesting and had been replicated in literature and people were talking about and, and, um, and, and yet it took a couple days from start to finish to do this electrophoresis and get the result. So it wasn't really clinically translatable a-a-and efficient. And, and so, um, just to see if NMR could do this, I hooked up with Ron. He sent me about 45 samples along with the gradient gel electrophoresis tracing so I could see who had pattern A, the large LDL, pattern B, the small LDL. And, uh, sure enough, when we applied our analysis for decomposing the composite signal into its parts, we could definitely tell the difference between large and small LDL. So aha, okay, now we spent a couple years seeing if we could refine the methodology for quantifying small LDL, large LDL, small HDL, large HDL. And, uh, and, and that was quite successful. Um, so it, it, it really was with this idea that there was something really clinically useful about being able to differentiate the size of LDL particles that drove us to take the [chuckles] step that I was very unqualified to take, which is commercialization of NMR testing technology. Um- It, it really did seem that yes, you not only could generate the same information as a lipid panel by NMR, but really what would drive the utility of NMR testing was if it could measure something better and different. And so if we could measure small dense LDL pattern A and B, threefold greater risk associated with that at a given level of LDL cholesterol, well, that would be a pretty useful thing clinically.

   15. PAPeter Attia12:49

       Tell us

3. 12:50 – 18:16

   How standard lipid panels measure cholesterol and triglycerides, and why LDL cholesterol is estimated rather than directly measured

   1. PAPeter Attia12:50

      how, we don't have to go into great detail, but tell us how a, a plasma... So you go to the doctor, they draw your blood. The last thing the patient sees is that tube of dark blood that's leaving them. Tell me what has to happen from there until they get a basic lipid panel back, which says total cholesterol is this many milligrams per deciliter, LDL cholesterol, HDL cholesterol, triglycerides, all in milligrams per deciliter. How do they get that out of that tube? What is the, what is the basic chemistry?

   2. JOJim Otvos13:16

      How do they do that chemically?

   3. PAPeter Attia13:17

      How do they do that, yes, chemically-

   4. JOJim Otvos13:19

      Chemically

   5. PAPeter Attia13:19

      ... and, and in a, in a, in a clinical laboratory.

   6. JOJim Otvos13:22

      These are standard chemistry-based, um, assays that are like all such assays. You, um, add a reagent that reacts with what you're trying to measure, like triglycerides. Actually, triglycerides is interesting because, you know, triglycerides are, are a, a fatty acid esterified to glycerol. Uh, so what actually happens in that assay is the, the blood is, um, um, exposed to, uh, a lipase that, that, that hydrolyzes, that, that separates the fatty acid from the glycerol. It leaves the glycerol, and then the glycerol is what something else is added to, to, to make, make a color change in proportion to the amount of glycerol. So this is how they all-

   7. PAPeter Attia14:10

      You're actually counting the glycerol-

   8. JOJim Otvos14:11

      You're counting the glycerol

   9. PAPeter Attia14:11

      ... and rever- imputing how much triglyceride you had-

   10. JOJim Otvos14:14

       Imputing how much triglyceride

   11. PAPeter Attia14:14

       ... because you know it was a three-to-one ratio.

   12. JOJim Otvos14:16

       Yes. Um, the clinical issue, it's not, uh, uh, common, but there are situations where somebody has a lot of glycerol not esterified [chuckles] . And so the assays for standard assays for triglyceride would say that this person has very high triglycerides-

   13. PAPeter Attia14:33

       It's overestimating

   14. JOJim Otvos14:34

       ... and they actually don't have high triglyceride. Anyway, so, so the same thing with cholesterol. So you're adding chemicals, adding reagents that cause a color change or a change in the UV spectrum that is monitored, and you have a standard curve that relates, uh, known amounts of, of LDL cholesterol to the signal, to the color that's created, and you can work backwards from that measurement.

   15. PAPeter Attia14:58

       So these can be completely automated. You're using-

   16. JOJim Otvos15:00

       It's totally-

   17. PAPeter Attia15:01

       ... potentially optical density or something like that.

   18. JOJim Otvos15:03

       That's right.

   19. PAPeter Attia15:03

       Shine a light through it, and then, you know.

   20. JOJim Otvos15:04

       That's right. UV detection or visible, visible light detection. So these are, these are really standard, very, very efficient auto analyzers do this. The challenge with, uh, cholesterol, though, so HDL cholesterol, so the problem is you're measuring the cholesterol inside VLDL, LDL, and HDL.

   21. PAPeter Attia15:23

       So when does that get broken? When do the lipoproteins break open, in what part of the assay, so that you are just looking at the total amount of cholesterol contained?

   22. JOJim Otvos15:32

       Right. So, so the original interest in cholesterol and its relationship to cardiovascular disease risk was just total cholesterol. So basically, the reagents find all the cholesterol inside all these particles, don't differentiate where it's coming from. If you, if you want LDL cholesterol, you have to separate the LDL particles, these are spherical containers that, that, that, that contain the cholesterol, and separate it from the HDL containers and the VLDL containers, and then do a regular cholesterol assay on what you've separated. So there's a separation step. Same thing for HDL. And for many years, until fairly recently, uh, HDL cholesterol was always, um, measured by first getting rid of the VLDL and LDL by precipitation, and then it left HDL that you then did a cholesterol assay. LDL cholesterol is more tricky because you would have to separate VLDL from LDL, and that takes an ultracentrifuge step, which is laborious, and even clinical laboratories today don't even have ultracentrifuges. Um, so, um, you know, people devised a way of calculating LDL cholesterol by, by measuring, um, you know, total cholesterol minus HDL cholesterol, which is VLDL plus LDL cholesterol, and then estimating VLDL cholesterol by dividing triglyceride by five. Triglycerides are mostly in VLDL. So it was a, uh, an easy but not terribly accurate way of, of quantifying or estimating VLDL cholesterol from which you could subtract the HDL cholesterol from total cholesterol and get LDL cholesterol.

   23. PAPeter Attia17:11

       Is it safe to say, Jim, that when a patient gets a lipid panel today, unless it says LDL direct, which we'll talk about, and it just says LDL cholesterol equals 127 milligrams per deciliter, is it a safe assumption that it may have been done using that exact same methodology you described?

   24. JOJim Otvos17:28

       Yes, absolutely. And the only thing that's changed fairly recently is recognition that dividing triglyceride by five doesn't give a very accurate VLDL cholesterol estimate, and that impacts the accuracy of the LDL cholesterol estimate. And so now there are equations that interrogate other things, non-HDL cholesterol, triglycerides. There's an NIH equation that I helped people, um, you know, develop that is used by LabCorp and other major laboratories. Um, and, and so, so you can, you can do better than estimating it by a Friedewald formula. But most laboratories, I think, to this day still use the Friedewald divide by five to get the, uh, the LDL cholesterol.

4. 18:16 – 26:52

   How NMR spectroscopy measures lipoprotein particle size and concentration

   1. PAPeter Attia18:16

      So let's talk a little bit now about NMR. So Um, again, uh, maybe someone remembers back in an organic chemistry, in organic chemistry class that one of the problems you would receive on an exam or something was you would be shown a picture, and the picture was, uh, very much like how you were just describing your experience in the '80s and '90s, where you, you had along an X-axis a line, and then it would have these spikes, and they would sort of correspond. So, so tell us, what, what did the X-axis correspond to, and what did the amplitude or Y-axis correspond to? And, and, and of course, I want you to get to the point of these are protons, but, but, you know, get there in, in your own way, of course.

   2. JOJim Otvos18:52

      Yeah, so you're, you're exactly right. The-- That's what the output of a normal NMR spectrum looks like, irrespective of whether you're detecting protons, hydrogen, nuclei, or carbon-13 or nitrogen-15, et cetera. Um, so on the X-axis is frequency. It's just frequency. These are signals that have different frequency, and their amplitude is proportional to how much of the molecules carrying this, um, the hydrogens. Let's just talk about proton hydrogen NMR. Um, they show up in different places. They have different frequencies depending on their chemical environment, and that's why this is a useful structural tool for organic chemists. So a, a CH2 group next to another CH2 group as opposed to a CH2 group and a, and a double bond or whatever it might be, they show up in very different places, and there's other differences that are, are structure-dependent. So you can work back from an NMR spectrum and deduce the structure of the compounds that are giving rise to the, to that. Um, so the, so, so this was never used, though, as a quantitative analysis tool. It was used so its relative signal intensities would tell you how many protons are here on a molecule, how many protons are there on the molecule. Does that fit the structure that you were attempting to synthesize, for example?

   3. PAPeter Attia20:19

      And so just I just wanna make sure people understand what you're getting at there. So you're trying to synthesize something. You know the structure of the thing you're trying to synthesize.

   4. JOJim Otvos20:27

      You think you know it.

   5. PAPeter Attia20:28

      You think you know it, and therefore, you should know what its spectra looks like, and now you're basically trying to match the-

   6. JOJim Otvos20:36

      Yes

   7. PAPeter Attia20:36

      ... NMR of what you've synthesized to say, "Look, this should have a carbon, a double bond, a carbon, a single bond. Over here's gonna be an oxygen that's gonna produce..." And, and it's, it's, I mean, I-- it's, it's a really fun game. It's, it's, I mean, I-- not to be too nerdy about it, but it's a super fun puzzle to solve effectively.

   8. JOJim Otvos20:55

      Yeah, and it's, it's, it's pretty complicated, and I haven't done that sort of thing in 50 years, [chuckles] so, uh, so I moved to a, to different application, which is using NMR to detect a s- a single signal. So I, I didn't care about what was in the rest of the spectrum. I cared about the signals that came, that were these putative cancer signals, so the, from the terminal methyl groups of the fatty acid chains that are carried on different types of lipids that are carried in these lipoprotein particles.

   9. PAPeter Attia21:25

      By the way, did the NMGA, uh, uh, uh, New England Journal of Medicine authors know that they were looking at lipoproteins?

   10. JOJim Otvos21:32

       I think they said lipids. I can't remember.

   11. PAPeter Attia21:34

       Okay.

   12. JOJim Otvos21:34

       You know?

   13. PAPeter Attia21:35

       Yeah.

   14. JOJim Otvos21:35

       And, and again, and, and I think, yes, th-they did, and, and, and they, they suggested that, that if you had cancer, there was some structural alteration in the lipoproteins-

   15. PAPeter Attia21:45

       Got it

   16. JOJim Otvos21:45

       ... that gave rise to a different signal, and that's what we disproved by-

   17. PAPeter Attia21:49

       Yep

   18. JOJim Otvos21:49

       ... showing that that wasn't true.

   19. PAPeter Attia21:51

       Yep. Okay.

   20. JOJim Otvos21:52

       So anyway, back to this issue of whether we could use that so-called cancer signal as a source of quantitative information about the lipids and lipoproteins or the lipoproteins themselves. So what's, what, what I told you is, is true. It was empirical observation that there was this very consistent relationship between where the frequency of the signal from the very same methyl groups on the very same molecule, so the, the molecules d-don't differ at all in terms of what's carried in VLDL, LDL, and HDL. It's the same lipids, so you'd expect they would all show up in exactly the same place. But for some magical physical chemical reason that is explained by complex equations that I don't even understand very well, um, it's been shown that a larger particle will always give rise to a signal that has a slightly lower frequency, and a, and a smaller signal will have a slightly higher frequency. So the same lipids in different size packages show up in slightly different places, but very reproducibly different places. And so the idea is that if you understand exactly where the signal shows up from a particular diameter lipoprotein particle, and also measure that because the shape of the signal will differ, that's another complexity, but adds to the accuracy of what we're able to come up with. Um, so with a complete understanding of what the different parts are that make up the mixture, make up the whole, the whole idea is that you measure the whole, and then you decompose it into the parts, so the sum of the parts equals the whole. Um, what's efficient about the methodology is that you're measuring something with a really low-tech, simple NMR spectrum that you can obtain in 30 seconds. A computer then with a deconvolution model that has in it what the signals look like from all the different size VLDL, LDL, and HDL subclasses, and then take that measured, 30-second measured composite signal and spit out how big the signals must be from all the different constituent parts to when they, when they s-superimpose, they will cr-recreate that shape of the composite signal. Okay? So, so that's the idea. The, the concentration information comes from how big the deduced NMR signal intensities are in this mixture, blood

   21. PAPeter Attia24:28

       Sorry, one point to just add to that, Jim, that a person who's looking at their own NMR result in their blood test will notice that the units are reported-

   22. JOJim Otvos24:37

       Yes

   23. PAPeter Attia24:37

       ... as nanomoles per liter as opposed to milligrams per deciliter. So it's not a mass concentration, it's a molar concentration. Maybe explain to people-

   24. JOJim Otvos24:45

       Yeah, that's just where-

   25. PAPeter Attia24:45

       ... what that distinction is

   26. JOJim Otvos24:46

       ... I was gonna go next.

   27. PAPeter Attia24:46

       Okay.

   28. JOJim Otvos24:46

       Because, um, uh, as I told you, the, the chemical constituents in these particles, and they're, they're quite heterogeneous. There's different size leg-legs of fatty acid chains. Some are saturated, some are monounsaturated, some are polyunsaturated. All of these things, um, give rise to the, the lipid complexity of these particles. But the way, the NMR detection, at least of this signal, knows nothing about any of that. It just, it, it's basically a lipoprotein particle signal. And, and so what should be the case, if that's true, is that how big that NMR signal is from the particle should relate to the number of particles, irrespective of what the lipid concentrations are, and there are variable amounts of cholesterol and triglyceride in most lipoproteins. So our-- what we realized at the beginning was if this was a lipoprotein particle signal we were interrogating, we could get lipoprotein particle concentration information, but we could not and should not report lipoprotein cholesterol levels or lipoprotein triglyceride levels 'cause we actually weren't able to differentiate the signals from those different chemical species. So that [chuckles] is what we started to produce when we did this commercialization thing, which I may come back to a little bit, but, but as you said, what's reported, LDL-P, is the particle concentration in nanomoles per liter. So there's 6.02 times 10 to the 23rd particles in a mole, uh, uh, uh, of, of LDL, and so this is 10 to the minus ninth. Anyway, so it's a, it's, it's a big number. It's, it's still 10 to the 16th particles. So there's a lot of these particles in your blood. Um, but y-you're reporting the concentration of the package, not the lipid molecules in the package, okay?

5. 26:52 – 35:02

   Why LDL particle number matters more than particle size, and why large, “fluffy” LDL is not benign

   1. JOJim Otvos26:52

      So then the question is, or one question is, um, is there any advantage to, to that? I mean, because [chuckles] by then... So I'll, I'll go back a little bit to the commercialization because as I said, the commercialization was driven by the idea that small, dense LDL is much more atherogenic, much more to worry about than large LDL. And, and in our own studies, when we started measuring small and large LDL by NMR and did it in large population studies, we found exactly the same thing that Ron Krauss and others did.

   2. PAPeter Attia27:28

      And which population did you look at? Did first MESA or Framingham?

   3. JOJim Otvos27:32

      Framingham for sure back in the day. MESA came a little bit later. But it, I can't remember what, what we initially looked at. But the, the point is that when looked at through the lens of at a given quantity of LDL, if the LDL is small versus large, does it make a difference in your cardiovascular risk? When I said the concentration or the quantity of LDL, the way everybody thinks of the concentration of LDL is LDL cholesterol, and that's what Ron Krauss and that's what we did. We said, "Okay, at a given level of LDL cholesterol, you take people and you stratify them according to low, medium, and high LDL cholesterol, or you do a multilinear regression, uh, and, and put, um, LDL cholesterol in the model, and now you're asking, does the size of the LDL add anything to LDL cholesterol in cardiovascular risk prediction?" And sure enough, it does, and it's quite an impressive increment of risk. Um, so we reproduced what Ron Krauss did. However, we realized because we were, uh, also in the business of measuring LDL particles, that by definition, if an LDL particle is smaller versus larger, it's always full of lipid. It's, it's, you don't get a, a, a partially filled container of LDL. It's always full. So a smaller LDL particle means by definition that it's carrying less cholesterol, less lipid per particle than a large LDL particle. So the phenomenology is at a given level of LDL cholesterol, people with small, dense LDL have higher cardiovascular risk. The trouble is that people with small, dense LDL at a given level of LDL cholesterol have more LDL particles. Their LDL-P is higher than would've been imagined from the LDL cholesterol measurement. So an alternate explanation f-for the extra risk that small, dense LDL seem to confer is that there are simply more particles rather than the size of the particles being the determining, uh, you know, characteristic of, of the atherogenicity of these particles.

   4. PAPeter Attia29:39

      Yeah. Let, let me just u-use a silly childlike example to make this point. So imagine you had four lipoproteins that each contain three units of cholesterol. They're fully saturated at three units of cholesterol. So you have four of them, so you have 12 units of cholesterol. Now imagine you have three spherical lipoproteins that each have four units of cholesterol. They're fully saturated, so they're obviously bigger, but they also collectively have 12 units of cholesterol. So here you have two people. One has 12 units of cholesterol, but it's being carried with four lipoproteins. The other has the same 12 units of cholesterol carried by three. You're telling me all the data say the first person is at higher risk. The question is, are they at higher risk because their spheres are smaller, or are they at higher risk because they have more spheres which happen to be smaller?

   5. JOJim Otvos30:32

      Correct. So you said, you explained that very well. And so this is basically how science works. So you have a different explanation for the phenomenology of small, dense LDL having this seemingly extra atherogenicity. It's then testable. So [chuckles] it's, it's, it's quite straightforward at that point to ask the question at a given level of LDL particles, if you stratify people according to LDL-P, and then say some people have small LDL, some people have large, is there any difference in their risk? And the answer is none, not a bit. And we did this in many-

   6. PAPeter Attia31:08

      So now again, the, now the example is you have two patients that each have 20 particles.

   7. JOJim Otvos31:13

      Yep.

   8. PAPeter Attia31:15

      One of them is 10 big, 10 small. The other one is 15 big, five small. If the particle size matters, the first one should be at higher risk. If the particle size is irrelevant, once you've corrected for total number, they should be at the same risk.

   9. JOJim Otvos31:31

      Right.

   10. PAPeter Attia31:31

       You're saying they're at the same risk.

   11. JOJim Otvos31:32

       That's right. And it... And why does this matter? I mean, if you're only interested in assessing the risk of a person, you're equally well off with the cholesterol information and the size information as the particle information and the size of the particle information, the size doesn't add to that. Um, the, the reason that's important is that if you believe that small LDL is bad and you can make it less bad by making the particles bigger therapeutically somehow, then you will be telling patients that at a given, let's say they get treated with statins and they lower their LDL cholesterol or their LDL particle concentration to an acceptably low level, but the particles are still small, somebody who believes small, dense LDL is particularly bad will then try to do something to make them bigger and imagine that there's clinical benefit. And a lot of drugs actually have that effect. Uh, CETP inhibitors are one of them, uh, niacin, um, you know, uh, uh, HDL drugs of a different type. So triglyceride-lowering automatically will, will, will have this effect. So, um, there really are a lot of clinical trials, uh, that, that imagined when they set out to do the clinical trial that they, there would be a lot of efficacy, like niacin, for example, because not only did that modestly raise HDL cholesterol and lower triglyceride, which were two good things seemingly, but it made the LDL particles bigger, and it made the HDL particles bigger 'cause there's also a similar argument about the size being important in HDL. Um, and, and yet there was no efficacy, uh, when they did the outcome studies. And, and this has been reproduced in w- with, you know, the fibrates and, and, and, and other drugs.

   12. PAPeter Attia33:29

       So Jim, what would you say to the person listening who, 'cause I hear this all the time, who says, "Um, hey guys, I do have a high LDL cholesterol, uh, and, and even my LDL particle number is very high on my NMR test, but I'm very pattern A. You see, all of my LDL are very large." And they even use words like fluffy-

   13. JOJim Otvos33:51

       Yes, they do

   14. PAPeter Attia33:51

       ... and buoyant. "I have large, fluffy, buoyant-

   15. JOJim Otvos33:55

       Benign

   16. PAPeter Attia33:55

       ... yes, LDL. Um, so my LDL particle number is over 2,000 nanomole per liter, um, which probably places me at the 80th percentile or so, but I don't need to worry about it because they're all big. I don't have that pattern B. I don't have those small." What, what would you say to that person?

   17. JOJim Otvos34:11

       I would say that's, uh, a fallacious idea. Um, and, and the data s- that, that I've mentioned, uh, supports it completely. And then you also have to think about the fact that one of the best-known or best-accepted genetic reasons for cardiovascular risk is FH, familiar hypercholesterolemia. People with FH have very high LDL cholesterol levels. They also have very high LDL particle levels, but those particles are large. They're not small. And there's, you know, these are people that die when they're 30 or 35 years old when they're homozygous FH. [chuckles] So this idea that somehow fluffy, large LDL particles are not to worry about or benign i- is completely fallacious.

6. 35:02 – 44:27

   Discordance between LDL cholesterol and LDL particle number: which measure better reflects cardiovascular risk?

   1. PAPeter Attia35:02

      Let's talk about one more thing on this before we pivot away from this, which y- um, you alluded to very briefly, but we, we, we, we went off on a different path, which is the potential discordance that exists between LDL cholesterol and LDL particle number. And, um, in fact, when I first came to learn about your work, Jim, this was actually one of the first papers I read of yours, uh, literally 15 years ago, almost to the, to the month. Um, and it was in the MESA population, and it was showing four, uh, Kaplan-Meier curves. Um, so I'll let you explain what a Kaplan-Meier curve is, but it was basically the four scenarios, which was discordance between LDL-P and LDL-C when one is higher than the other, uh, each, each time, and then concordance between them. So maybe there were three curves then in that-

   2. JOJim Otvos35:58

      That's exactly right.

   3. PAPeter Attia35:58

      Yeah, there were three curves. Okay, so concordance between LDL-C and LDL-P, and discordance in favor of LDL-P, discordance in favor of L- okay. So that was, that was a very, very eye-opening paper to me, um, and I'd, I'd love you to just kind of explain that finding and, and what the significance are, 'cause to this day, it's still a very important-

   4. JOJim Otvos36:18

      Right

   5. PAPeter Attia36:18

      ... thing for clinicians and patients to understand.

   6. JOJim Otvos36:20

      No, it is, and, and prior to that MESA paper in 2011 was a 2007 paper, so it was the first one that we sort of introduced this idea of assessing the situation by Whether the LDL-P and the LDL-C agreed or not, and in that case was a Framingham study, and in that case one question that always comes up is, well, what defines discordance? And, and [chuckles] basically, it doesn't matter, and everybody defines it differently, and it, it doesn't matter because we're not saying discordance is a risk factor, right? Um, we're simply trying to disentangle a situation where most people don't have a discrepancy between LDL-C and LDL-P. So if you're trying to understand whether LDL-P is a better measure of LDL-associated cardiovascular risk, and you do a whole study population, and 80% of those people have concordant or in agreement LDL-C and LDL-P-

   7. PAPeter Attia37:23

      And just explain what that means 'cause the nu- people will say, "Well, the numbers are different. How can they be in agreement?"

   8. JOJim Otvos37:27

      Well, the... Okay.

   9. PAPeter Attia37:27

      You mean-

   10. JOJim Otvos37:28

       So what one way to do it, which we did in the, in the Mesa paper, is, uh, transform the milligram per deciliter cholesterol number into a percentile. So that defines people according to their rank, low, high, intermediate, and if you do the same thing to the LDL-P, now you're comparing apples to apples, percentile of one, percentile of the other, and you plot that, and then you, you see the, the data points all over the place. And the ones that track on the diagonal, the ones which are close to the diagonal, are the ones for which, uh, a 50 percentile LDL-P is close to a 50 percentile LDL-C. So that's what we call concordance. And in that case, we picked [chuckles] this weird number of, of 12 percentile difference was our cut point, plus or minus. So everybody within 12 percentile units of each other we called concordant. Why 12? Why not, why not a round number? [chuckles] And it was because we were trying to make the concordant population about equal in number to the discordant in one direction and the discordant in other direction population.

   11. PAPeter Attia38:32

       Yeah. So you had-

   12. JOJim Otvos38:32

       Roughly

   13. PAPeter Attia38:32

       ... 50% in one, 25... and 50% in the two discordant groups.

   14. JOJim Otvos38:36

       No. It's actually we're trying to make it sort of 30/30, you know-

   15. PAPeter Attia38:39

       Ah, I see

   16. JOJim Otvos38:40

       ... 30/30/30.

   17. PAPeter Attia38:40

       30/30/30. Got it. Okay.

   18. JOJim Otvos38:42

       Uh, so, and so that's what, w- what we did, and then we simply took these three groups of people, and you talk about Kaplan-Meier. All this is basically cumulative incidents of cardiovascular events, so on the X-axis you can see the number increasing. If it increases a lot, this is a higher risk subpopulation than somebody who, uh, whose risk is, is, is much shallower and doesn't increase very much as a function of concentration. So that very clearly showed that for people who- whose LDL-P and LDL-C agree, you can make no argument about why LDL-P [chuckles] is a better thing to measure than LDL-C. Then it's a matter of, well, from first principles, why would we expect the cholesterol to be a better measure of risk or the particle number? And really, if you're honest about it, there is no mechanistic explanation that you should really be too comfortable with for one versus the other, and we explain that in that paper that what, what has happened with, with when Ron Krauss showed that small, dense LDL was more atherogenic and that this was based on epidemiologic s- you know, population studies, the question was, "Well, why is that?" W- and then basically the speculation was s- s- ultimately supported by various lines of evidence that, yes, smaller LDL particles are likely to get into the arterial wall easier than a bigger particle, and then the difference in the shape of the, uh, stuff on the surface of the particles can bind more avidly to molecules that are in the arterial wall and retain it more, and then it's readil- more readily oxidized. So all these things, I can't tell you how many papers I've read where in the discussion section is this obligatory paragraph about why small, dense LDL is so much more atherogenic. The flip side argument is that, okay, large LDL particles also get into the artery wall, and when they get oxidized-

   19. PAPeter Attia40:46

       They're more retained.

   20. JOJim Otvos40:46

       [chuckles] They're... W- they're... No, no, they're not more retained.

   21. PAPeter Attia40:49

       Well, I mean, but I'm just saying you could make that argument, right? You could always-

   22. JOJim Otvos40:51

       Well, let's just say that they, they get th- they get to the end of the trail, and they get taken up by macrophages and deposit their cholesterol contents into the arterial wall. Well, bigger particles have more cholesterol-

   23. PAPeter Attia41:04

       More cholesterol

   24. JOJim Otvos41:04

       ... to deliver.

   25. PAPeter Attia41:04

       More oxidative damage. Yeah.

   26. JOJim Otvos41:05

       So you'd think the more cholesterol-rich particle should be more atherogenic.

   27. PAPeter Attia41:08

       Yep.

   28. JOJim Otvos41:09

       But if that's counterbalanced by not getting to the end of the party, uh, as, as often as the small particles, maybe it's a wash. How do you answer the question? You do the study. [chuckles] And so that's what the discordance analysis allowed us to do. Whenever there's a discrepancy in one direction or another, the cardiovascular ricks- risk tracks with the particle number, not the cholesterol.

   29. PAPeter Attia41:31

       Yeah, so the curve really had a beautiful separation of those three figures. You had the concordance in the middle, and then above that line, which means these are people that are dying quicker, that's when LDL-P was above LDL-C. And below the line, these people died much slower than you would expect. It was the flip. It-

   30. JOJim Otvos41:50

       Their cholesterol was high, but their particles were low.
7. 44:27 – 50:32

   How metabolic syndrome and lipid-lowering treatment contribute to the discordance between LDL-C and LDL-P, and the value of particle number for managing risk

   1. PAPeter Attia44:27

      Now, Jim, I don't know if this was one of your papers, um, but I think it was. Um, and it was around that time, probably 2012, 2013. And if I recall the paper, the figure specifically, it was, uh, uh, it was a histogram, right? So on the X-axis, you had zero, one, two, three, four, five, and you were looking at number of criteria met for metabolic syndrome. So for the listener, just to remind everybody, metabolic syndrome, uh, has these five criteria, and the more of them you have, the more likely you are to be insulin resistant. So it's, there's one about having high blood pressure, uh, obesity, so the truncal diameter, truncal girth, uh, blood pressure, fasting glucose, and HDL cholesterol. I think those are the five, right?

   2. JOJim Otvos45:14

      Yes.

   3. PAPeter Attia45:14

      Doesn't include... Yeah. Okay. So and what this figure showed was if you had zero of them, the probability that you were discordant between your LDL-C and LDL-P was very low. If you had one of them, the probability went up a little bit, two of them considerably, three a lot, and it was a monotonic increasing relationship, Jim, such that by the time you had five out of five, um, uh, MetSyn criteria, you were virtually guaranteed to be discordant.

   4. JOJim Otvos45:46

      Right.

   5. PAPeter Attia45:46

      And so I bring that up to say earlier you mentioned there's nothing wrong with being discordant per se, although it's... You could also argue that the m- the more li- the more discordant you are, the more, the higher the probability that you're discordant means the higher the probability that you probably have some underlying metabolic insulin resistance.

   6. JOJim Otvos46:05

      Extra risk.

   7. PAPeter Attia46:06

      Yeah.

   8. JOJim Otvos46:06

      Yeah.

   9. PAPeter Attia46:06

      And so the question then becomes, is the reason that very elaborate multiparametric risk models can erase the LDL-P is that they are simply capturing all of the cardiometabolic risk indirectly and directly.

   10. JOJim Otvos46:21

       Yeah, you, you basically just explained what I was getting at, which is because HDL cholesterol is in the standard risk assessment models, when you have low HDL cholesterol, you're likely to have more LDL-P than the LDL cholesterol indicates or suggests. And so your risk is higher because you have higher LDL-P, not because HDL cholesterol is low. But you're not doing better in risk assessment because the HDL cholesterol, for the wrong reason, if you will, not a, not a causal reason, but an association reason, um, is telling you the same thing. So th- this was really important to Liposcience back in the day because, and, and this is something that [chuckles] I still argue with, but I, Alan Sniderman and I, we'll get to the question of ApoB.

   11. PAPeter Attia47:13

       Yes, yes.

   12. JOJim Otvos47:13

       But, um, he insists, he keeps wanting to convince people that ApoB is important for risk assessment. And what we pivoted to in light of the evidence that we found ourselves, that above and beyond the standard risk equations, you couldn't add LDL-P and see a, a significant increase in risk. Liposcience pivoted from trying to convince people that LDL-P was a better thing to measure for risk assessment to a better thing to measure for risk management. Because what you do when you have high risk is you manage it by lowering your LDL cholesterol. This is the s- the best tool we have in the toolbox. There are not actually very many others. We have statins, and then we have things that are doing the same job as statins, but better, so we can lower the heck out of LDL. And the question for a patient with high cardiovascular risk is, how much LDL, LDL lowering do I need? And so what you really would like is a biomarker that tells you, "My LDL-associated risk is adequately controlled. I've gotten my LDL low enough." If it's not, I should add, if, if, if I'm on high-dose statin and I'm not there, then I want to add a PCSK9 inhibitor or something else to lower it even more. So you want the best objective measure of LDL-related risk, not total risk, but LDL-related risk. In that case, it's very clear that LDL-P or ApoB is a better biomarker because a lot of people who achieve very low LDL cholesterol have not achieved equally very low LDL particles or ApoB, and they therefore, with visibility to that, would be candidates for more aggressive LDL lowering.

   13. PAPeter Attia49:11

       So we absolutely notice that as clear as day in our practice, Jim, because we are very aggressive at managing these things. And is, is there a biologic reason for why the discordance really happens at low levels in that way, or is it a chemical assay? Is it a, is it a property of the assays that's causing that?

   14. JOJim Otvos49:29

       No, it's [chuckles] like everything that you ask. It's, it's, it's complicated. You can get into the weeds, but what's, what's true is that the lower your LDL level is, the more likely that the cholesterol in the LDL particle is replaced to some extent by triglyceride, 'cause there's always this interchange, this swapping-

   15. PAPeter Attia49:50

       Mm-hmm

   16. JOJim Otvos49:50

       ... of cholesterol ester and triglyceride in the core of the particle, and it's driven by the relative amounts of the triglyceride-rich particles and the LDL particles, which are cholesterol rich. The triglyceride-rich particles, the bigger that gap is, put triglyceride into LDL in exchange for cholesterol ester. And so as you lower LDL with statins or whatever, the LDL particles are lower, but the LDL cholesterol is even lower because not only has the particle number gone down, but the cholesterol in the particles independently has gone down. [chuckles]

   17. PAPeter Attia50:24

       And proportionately is going down.

   18. JOJim Otvos50:25

       Okay.

   19. PAPeter Attia50:25

       That, that's, that's actually a great explanation.

   20. JOJim Otvos50:27

       Good.

   21. PAPeter Attia50:27

       I, I did not know that, and that makes sense because we see that as clear as day.

   22. JOJim Otvos50:31

       Right.

8. 50:32 – 1:05:04

   Using the NMR-derived LP-IR score to detect insulin resistance and predict type 2 diabetes before glucose rises

   1. PAPeter Attia50:32

      Okay. One other thing I wanna talk about on this front before we pivot is, at least to my knowledge, the first composite score that you then developed out of that, which was the LP-IR score. Was that indeed sort of the, your first foray into pool composite scores?

   2. JOJim Otvos50:46

      Yes. Um, and it was because-

   3. PAPeter Attia50:49

      Still used to this day, so many people-

   4. JOJim Otvos50:50

      Yeah, no, it's, it's-

   5. PAPeter Attia50:50

      ... again, listening to us will have an LP-IR score-

   6. JOJim Otvos50:52

      Yeah

   7. PAPeter Attia50:52

      ... every time they get their blood test.

   8. JOJim Otvos50:54

      No, it's, it's, it's, it's interesting because, as I said, we first got into this game because we could measure small LDL, and, and so the, the, the ability to differentiate different-sized lipoprotein particles seemed like it was clinically useful. But at the end of the day, as I've gone through, it turns out that it's really the particle number that matters, and if you can conin- convince people to not pay so much attention to LDL cholesterol and pay more attention to ApoB or LDL-P, uh, then you're better off. But that means that measuring the size of LDL [chuckles] doesn't matter. Well, that's a bummer because we have a great efficient way of doing that. Um, so then the question was, well, is, are the lipoprotein subclass distributions useful for something else besides cardiovascular risk assessment and management? And the answer is absolutely yes. There is a well-known association between higher triglycerides and lower HDL cholesterol and insulin resistance and diabetes, and insulin resistance leading to diabetes risk or leading to diabetes. So it's already known that there's a lipid signature for insulin resistance. And the idea was if we could measure the different sizes of VLDL, LDL, and HDL, could that do a better job than just the triglyceride over HDL cholesterol ratio? So the poor man's insulin resistance measure from a lipid panel is triglyceride over HDL cholesterol ratio. And Jerry Reaven, who was the-- [chuckles] really discovered and promoted the idea of insulin resistance being, being extremely important, um, he advocated that it be used because people were getting lipid panels, and this information was not being used for anything. Um, so w- w- with this LP-IR score, which brews together six VLDL, L- LDL, and HDL size and, and subclass concentration parameters, it brews them into a score from zero to 100, higher scores being more insulin-resistant. And then the question was, does this LP-IR score do a better job in assessing whether somebody is likely to become diabetic? So the pathophysiology has to be talked about a little bit here, um, because there's some interesting parallels to where we are today with respect to primary prevention of cardiovascular disease versus primary prevention of diabetes. So in the cardiovascular situation, as everybody pretty well understands, the initiating causal factor there is elevated cholesterol or elevated LDL or elevated ApoB, but it's acting over time, so it requires an integration of exposure over a long period of time, and that gradually leads to cholesterol deposition in the artery wall. That's atherosclerosis. Atherosclerosis over time starts little, more, more, but that doesn't trigger any clinical concern. This is a subclinical manifestation of cholesterol doing its dirty deed over a long period of time, and then at some point, you might have a myocardial infarction or a stroke, and that's when the atherosclerosis has transformed into a clinical event. The good news from a measurement biomarker standpoint is that the initiator, the causal factor, is cholesterol, which is easily measured. So now what about diabetes? Diabetes, very similar. It's a time-integrated process where if you are insulin-resistant- Over time, your beta cells have to s- spit out more insulin to keep your glucose under control. And so you're making the beta cells work harder, if you wanna think about it that way, if you're insulin resistant versus insulin sensitive. And so insulin resistance times time leads gradually to hyperglycemia, elevated glucose. So if it's n- 90 or below, you're A-okay. But over time, if you're gonna convert to diabetes, you go through a transition of the glucose going higher and higher until it crosses this magic 126 [chuckles] milligram per deciliter line that defines diabetes.

   9. PAPeter Attia55:27

      Just for folks to know, we're talking average.

   10. JOJim Otvos55:30

       Absolutely.

   11. PAPeter Attia55:30

       Yeah, yep, yep.

   12. JOJim Otvos55:32

       So the causal factor is insulin resistance. The thing that's sort of equivalent to atherosclerosis on the CVD side is hyperglycemia, sub- sub- le- less than 126, so not diabetes, but pre-diabetes. So when glucose gets over 100, before it goes from 100 to 126, you're pre-diabetic. Guess what's easy to measure? Glucose. So in that case, the effect of the cause is measurable. The cause itself is not. And so what that means is that from a prevention standpoint, what you really wanna do is to keep insulin resistance from transforming over time into hyperglycemia and ultimately diabetes. If you don't know that you're insulin resistant, you just... you're waiting for the easily measured glucose to become elevated. And now, once that happens, you've lost about 50% of your beta cell function. So the opportunity for real effective prevention, primary prevention, primordial prevention, is to act on people whose glucose is okay and hasn't gone to this transition yet because the beta cells have started dysfunctioning.

   13. PAPeter Attia56:50

       So Jim, I wanna pause you there-

   14. JOJim Otvos56:52

       Yeah

   15. PAPeter Attia56:52

       ... because the way you've laid that out is very elegant, and I, I like the-- I've, I've never thought of it the way you just explained it, so I'm repeating it just as much for me as for others. Um, people who listen to this podcast know we constantly use the d- the way you describe the CVD prevention thing, which is you have a causal marker. You don't need to wait until disease is measurable to treat it, and the example I always give is smoking and lung cancer. We have a causal marker, for lack of a better word, smoking. Um, we always have to specify causality doesn't mean one-to-one mapping. There are some smokers that never get lung cancer. There are some never smokers who still get lung cancer. None of those facts erase the causality of tobacco and lung cancer. Do we need to wait for a smoker to develop a small cancer to tell them to stop smoking? Absolutely not. That would be malpractice. You always eradicate causal drivers of disease the moment they appear, and that's why when you have elevated LDL-C or ApoB or LDL-P, you treat it immediately, not once they have disease. Same with hypertension, same with smoking as it pertains to cardiovascular disease. Now let's pivot to what you said, which is, look, we know that insulin resistance is the cigarette to diabetes as cancer.

   16. JOJim Otvos58:19

       Yes.

   17. PAPeter Attia58:20

       Do we wanna wait until we actually see the glucose rise, which by the way, is the biomarker that defines diabetes, when in reality, by the time that's happening, there's potentially already cellular damage at the level of the beta cell in the pancreas, and it's basically running out of steam. And what else can we measure? Now, I wanna come back to the idea of there are things that we can do, but they're very laborious. So an oral glucose tolerance test is a fantastic way to find out that canary in the coal mine years before it shows up, but it's so fallen out of favor as a clinical test because it takes two hours. It's, I mean, it's just so cumbersome-

   18. JOJim Otvos59:01

       Nobody does it

   19. PAPeter Attia59:01

       ... to do that outside of our practice and a few others, I just don't imagine many people wanna do it. So Jim comes along and says, "What if somewhere in this NMR spectrum is a whole series of things that turn out to be a fantastic marker for insulin resistance that we can use as causal proof that you're on the wrong path before your glucose goes up?"

   20. JOJim Otvos59:27

       Yes, and it seemed to us, uh, a very compelling case that [chuckles] you would want to act on the causal factor and not the effect or the c- the, you know, the, the downstream effect of, of, of the action of insulin resistance. Um, but it's, it's interesting. I mean, [chuckles] back to convincing people, clinical translation, I mean, this is-- I, I use that word. I don't think I was familiar with that word when I started this company called Liposcience, uh, to d- to try to get NMR testing introduced into clinical laboratories. But I thought the argument of LDL-P versus LDL-C was quite compelling and that people really should be using it as the biomarker to determine, uh, management of LDL. Tremendous resistance by the establishment. I could never understand why it was so, why they were so resistant. Uh, the messaging people learn about cholesterol in medical school, "Oh, we'd have to tell a different story, and people wouldn't get it," and all sorts of, you know, reasons that seem pretty weak to me. Um, so on the, on the diabetes side, it's the same thing. People imagine that hyperglycemia, pre-diabetes, is a risk factor for diabetes. It's not a risk factor. It is the disease. It's just in a less manifest form. So why not address the cause? A lot of resistance to that. I, I, it sort of blew my mind, and we did NMR analysis in a number of studies to show the efficacy of, or at least the relationship between insulin resistance score and likelihood of future diabetes. And by the way, transitioning to pre-diabetes by no means means that you're gonna get diabetes. I mean, there's [chuckles] you, you really... My wife has been pre-diabetic 105 or so milligram per deciliter for 25 years. Doesn't change. Um, so the insulin resistance score can tell you whether you're more or less likely, once you're pre-diabetic, to transition. And then, then at that point, even though you'd like to have intervened earlier, it's not too late to still do something about it.

   21. PAPeter Attia1:01:49

       And so Jim, was the LP-IR score validated on longitudinal data to predict the development of type 2 diabetes?

   22. JOJim Otvos1:01:57

       Yes.

   23. PAPeter Attia1:01:58

       Or... Okay. Um, so in that sense, we'll talk about MVX and how similar that was in that regard. Has there ever been a, um, a comparison that says how well does it do versus-

   24. JOJim Otvos1:02:10

       Yes

   25. PAPeter Attia1:02:10

       ... an oral glucose tolerance test? Yep.

   26. JOJim Otvos1:02:12

       Yes. Uh, and, and so, well, not oral glucose tolerance test because in the real world it's not being used.

   27. PAPeter Attia1:02:17

       Too complicated.

   28. JOJim Otvos1:02:18

       So you're really... Fasting insulin is sort of a, a, a, an easier way to assess insulin resistance. Um, it has some analytic issues and, and practical issues, uh, so nobody has really, uh, been interested in, in, in using insulin for that purpose generally in, in, in clinical practice. Um, but the LP-IR score has been compared to fasting insulin and is better. Um, the study that shows it the best is one that unfortunately hasn't been published yet. It's, it got very close to having the manuscript be written. It's in the diabetes prevention program, so you couldn't ask for a better study because this was a study that put, uh, intervention, lifestyle intervention, metformin on the map in terms of being able to do something about transitioning to diabetes once you had hyperglycemia, once you were pre-diabetic. Um, so this study was done about 20 years ago. We have baseline samples and then one-year samples post-treatment. All the... Everybody in the study had an NMR analysis done. LP-IR and other things that we can measure by NMR that make LP-IR better if they're, if you want it to be better, um, were shown to be independently predictive, better than insulin. But most importantly, because that was an intervention study, you know, lifestyle change and weight loss was shown to significantly reduce the incidence of diabetes in these people. Metformin, less effective, but significant compared to placebo. So what we're able to show really nicely is that the LP-IR score was reduced significantly by lifestyle, less significantly by metformin. Branched-chain amino acids, which we haven't talked about yet, but higher branched-chain amino acids are also related to insulin resistance and can improve the LP-IR score, and we actually have developed another score called the Diabetes Risk Index that integrates or adds branched-chain amino acids to LP-IR.

   29. PAPeter Attia1:04:30

       Okay. I was gonna ask you that. So just to confirm, the, the DRI, the Diabetes Risk Index, is the LP-IR score inclusive of the three branched-chain amino acids, and do you recommend... When, when do you recommend using one of those versus the other?

   30. JOJim Otvos1:04:47

       Well, you know, it really never has gotten to be far enough along or far to, to have the amount of acceptance, um, that that discussion has even occurred.
9. 1:05:04 – 1:17:00

   The development, commercialization, and uncertain future of the Vantera NMR Analyzer and NMR-based diagnostics

   1. PAPeter Attia1:05:04

      We should also point out for folks who are wondering, your company, Liposcience, that you founded in mid-'90s, was acquired by LabCorp 10 years ago.

   2. JOJim Otvos1:05:13

      Yes. Yes.

   3. PAPeter Attia1:05:13

      A little more than 10 years ago.

   4. JOJim Otvos1:05:14

      Yes.

   5. PAPeter Attia1:05:14

      And so, um, for tho- those of us like me, the dinosaurs, like we used to still order a Liposcience test. Now it just all happens through LabCorp. Um, so that, presumably that has increased, I assume, some, some uptake of the, of the test?

   6. JOJim Otvos1:05:29

      So I, I wanted at some point to get into, uh, the question of why hasn't broad clinical translation occurred, because it hasn't. Right now, you can only go to LabCorp for this information. Um, so I'll, I'll go there now briefly, if you don't mind.

   7. PAPeter Attia1:05:48

      Please.

   8. JOJim Otvos1:05:49

      Um, so Liposcience began, as I told you, as a spinoff of the university. I left the university, and we, we, we tried to convince people that size didn't matter and LDL-P did matter, and we did that for quite some time. And we were a laboratory testing company, and so samples were sent to Liposcience, and you'd get the results back.

   9. PAPeter Attia1:06:11

      You had beautiful results. I, I-

   10. JOJim Otvos1:06:12

       And that color report.

   11. PAPeter Attia1:06:13

       I lo- I loved it so much.

   12. JOJim Otvos1:06:14

       Why not use color printers? [chuckles]

   13. PAPeter Attia1:06:16

       Yeah. No, it was, it was beautiful.

   14. JOJim Otvos1:06:17

       But the am- the ambition was never, or the, uh, the business objective was never to be a lab that got bigger and bigger and did more testing. It was to make the ability to do NMR testing available to any laboratory in the world. So, you know, we started by what was available. I talked about NMR machines being in every chemistry department. These are research NMR machines. They're engineered for, um, multifunctionality. You can do any weird NMR experiment. You c- there's lots of variations on the theme. We wanted NMR to do one thing very efficiently and as rapidly as possible, and so f- you know, ideally 30 seconds or less, pop one sample in, automatically another sample comes in, boom, boom, boom So we realized that we couldn't, uh, use a research NMR spectrometer for that purpose. We used them for the initial years at Liposcience because that's all there was. But we wanted to transition from a lab testing company into an IVD company, in vitro diagnostics company. All laboratories rely on IVD companies to supply them the machinery and the reagents to do all these assays. So LabCorp does 3,000 assays. They rely on other people, Roche, other people, to provide them, um, with the wherewithal to do the testing. Those are IVD companies. We wanted to be an IVD company at Liposcience, make an NMR analyzer that looked just like a regular chemistry analyzer to a med tech that had no experience, no knowledge of NMR, walk up to it with 200 samples, present the tray, push the green gr- go button, and walk away. And so that's what we actually did, and a lot of investment and a lot of time and effort was put into that, and that is the Vantera NMR Analyzer. It's the only existing NMR analyzer in the world. And, uh, we went to the FDA because we, you know, these analyzers have to be FDA cleared in order to go into different laboratories. That was an adventure because, uh, what did FDA know about NMR spectroscopy? And it was a new, uh, platform, a new, a new way of testing, uh, that, that had to be understood. So anyway, it took-

   15. PAPeter Attia1:08:34

       When did you get the CLIA approval?

   16. JOJim Otvos1:08:36

       We got the-- No, not CLIA. We got FDA clearance for LDL-P and the Vantera Analyzer in 2011.

   17. PAPeter Attia1:08:43

       Oh, wow. Okay.

   18. JOJim Otvos1:08:44

       Yeah, 2011. And so these, these Vantera Analyzers, uh, a number of them were, were manufactured. Uh, and, and then, uh, you know, we-- About the time that just before, a year or so before Liposcience was sold to LabCorp, these analyzers were started to be distributed to major laboratories. LabCorp was one of the recipients of these analyzers, so they could do the testing in-house rather than having to send the samples to Liposcience. Clinical laboratories hate sending send-out samples. Um, and, and when that's for a rare cancer or something, when it's j- a fairly rare event, no big deal. If you're doing hundreds of these a day, you, you wanna-- you don't want that hassle. So they were very happy to receive the Vantera Analyzer so they could do the testing in-house, and every time they did, they would pay Liposcience for each analysis, okay? That was the business model. Um, we also made Vantera Analyzers available to the Mayo Clinic, Cleveland Clinic, Scripps Clinic, ARUP, a, a big reference laboratory in Utah. Um, so we were on the way to making it available broadly because we didn't wanna be the only people that could do NMR testing.

   19. PAPeter Attia1:10:02

       For sure.

   20. JOJim Otvos1:10:02

       And also, we wanted to convince people that this wasn't, you know, magic. You-- It, it, it was, it was real, and it was analytically, in many respects, much better than reagent-based chemistry testing. Unfortunately, by that time, Liposcience had gone public. Um, I was no longer on the board. I, I was basically the investment that we needed to start Liposcience. We basically took too long to get to the payout of the initial investors. So, uh, people, um, the, the investors were not patient. They were very patient up until then, but you, you almost couldn't blame them because they wanted to get their money back, venture capital in particular. Um, so the company went public, and then LabCorp came along and said, "We'd like to buy you because we can make more money if we don't have to pay whatever number of dollars to Liposcience every time we do this test." So it was a purely financial decision. LabCorp really didn't care about measuring things other than the NMR LipoProfile, the LDL-P, et cetera. Um, it was financially based. That was too bad for the vision of having NMR analyzers in every laboratory because LabCorp is a lab testing company. It's not an IVD company. So too bad the IVD business model, the IVD vision was ended [laughs] in 2014. Um, so yes-

   21. PAPeter Attia1:11:34

       Did you join? Did you go and become a scientist there?

   22. JOJim Otvos1:11:38

       So the, the, the research group largely was retained by LabCorp, so-- And we had a pipeline of things like LP-IR and things, DR-I, things that were coming down the pike, uh, that we'll talk about later. Um, so, you know, really the, the, the best was yet to come in terms of the clinical value and the things that could convince people that having Vantera Analyzers in their lab was a, a, a, a commercially useful thing, but also a clinical useful thing. So the bad news is that [laughs] LabCorp, understandably not sharing the IVD vision and being a laboratory that wanted to have, uh, NMR be proprietary to themselves, took back these analyzers that had been placed in these other reference laboratories. And then unlike Liposcience that knew that when it developed a new test, it needed to go to the trouble of creating awareness and interest and do- doing the studies to prove the clinical efficacy of these tests. That's a, that's a very important activity. LabCorp doesn't do that because they're not an IVD company. They are reactive. They're not proactive. They're very good at being reactive. And in COVID, they ramped up the COVID testing and, you know-- So I'm, I'm not, you know, saying that, that LabCorp doesn't do a really good job at what they do. But they really didn't know what to do with new knowledge that they generated in-house by acquisition of Liposcience. And so there was no marketing, no awareness creation. And basically, LipoScience went invisible and is still largely invisible. Um, most of the testing, uh, is still done by the people that were, uh, interested in the testing, thanks to LipoScience's efforts. So these X number of analyzers that were produced 15 years ago are what Lip- LabCorp is using to produce this information and this more exciting-

   23. PAPeter Attia1:13:40

       And what's the, what's the life of these analyzers?

   24. JOJim Otvos1:13:42

       Uh, good question, because, uh, nobody's ever... So the magnets themselves, these superconducting magnets, uh, last a long time and, and don't degrade, but they're, they're using PCs, you know, tw- 10-year-old PCs. And anyway, they're all the moving parts, so, so they're not gonna last much longer. And this is what maybe I'm most concerned about and most interested in people hearing about, because what needs to happen for this to continue, this clinical translation to continue at LabCorp, but also ideally broaden to the rest of the world, is for an IVD company to come in and basically acquire the LipoScience technology, and there's a lot of patents, intellectual property associated with this. A lot of expertise that comes from, in terms of service and keeping these machines functioning well, and y- you know, learn. I mean, we've done so many millions of tests, and you, you learn by doing. So what people don't know is that at some point in the future, maybe sooner rather than later, these Vantara analyzers, first generation, only generation, uh, are gonna cease to function. And it would be a real shame, especially with the things that we're gonna talk about that are more exciting than LP-IR and, in my mind, LDL-P, et cetera. So there's an issue here. There's a problem, and hopefully it'll be addressed by an IVD company taking over. I've tried to make the case to big, uh, multinational IVD companies, but NMR is too exotic. I mean, I've-- I basically show them that this is a completely de-risked populat- uh, proposition because we've already gone through the regulatory hurdles. We've already made the analyzers. We already have the experience. Uh, so it's not like starting from scratch with something that you're unfamiliar with, technology you're unfamiliar with. But, you know, these big companies have a lot of inertia and, you know, talking to the right people or having these big companies be entrepreneurial. So I think this is maybe, if it happens, it'll be possibly a smaller IVD company or a new startup IVD company basically taking the torch that, that LipoScience-

   25. PAPeter Attia1:16:02

       And has LipoScience said that they would be willing to sell those assets, the IP?

   26. JOJim Otvos1:16:06

       It's, it's LabCorp that owns all those assets.

   27. PAPeter Attia1:16:07

       Sorry, that's what I mean.

   28. JOJim Otvos1:16:08

       Yeah.

   29. PAPeter Attia1:16:08

       Has LabCorp said that they're willing to sell that IP?

   30. JOJim Otvos1:16:09

       Yes, they're-- And, and I think they realize, and they... Well, yes, they're, they're interested in licensing the, the patents. Most of the patents have to do with the assays.
10. 1:17:00 – 1:26:54

    The analytical efficiency of NMR testing and the data-driven development of the Metabolic Vulnerability Index (MVX)

    1. JOJim Otvos1:17:00

       And what we've talked about, we talked about LDL-P, and the NMR LipoProfile was the, the report that, that, that gave people the LDL-P information. But that comes from this, this simple NMR spectrum that can then use to be-- extract much more information than LDL-P. LP-IR, DRI is just part of the story. There's all these other things that we'll talk about later. Um, so in terms of efficiency, analytic efficiency, um, you, you couldn't ask for anything better because, you know, if you want to use NMR to produce a lipid panel, and actually, after ha- to- telling you that NMR only measured particles and not cholesterol, the basic information is, is encapsulated in all these NMR signals. So you can actually feed NMR data to a machine learning algorithm, AI, and train it to produce accurate lipid panel and ApoB information. So we published this four or five years ago, that you can actually use the NMR spectrum to produce an extended lipid panel, ApoB plus lipids, no incremental cost to a lipid panel. The ApoB doesn't add cost. It more than doubles the cost of a lipid panel if you want to do it by chemistry and use regular reimbursement, uh, in the US. So, um, you know, this-- So it's, it's extremely analytically efficient and cost-effective. The thing that people probably don't appreciate, and I didn't appreciate as a naive professor, I thought, well, surely this will be commercially attractive if it can cut the cost of doing these tests. The cost of these tests is so much smaller than the price that is charged for these, and the insurance pays for these tests, that cutting the price, the cost in half of doing the analysis makes no difference whatsoever. So this is, you know, my-- We'll maybe get into this later because what is true is that NMR, a single scan can tell you a heck of a lot more than just your cardiovascular risk, your inflammation level, your diabetes risk, your-

    2. PAPeter Attia1:19:16

       Your overall mortality risk as well

    3. JOJim Otvos1:19:18

       ... your mortality risk

    4. PAPeter Attia1:19:18

       as we're about, as we're about to discuss.

    5. JOJim Otvos1:19:19

       This all comes in the same assay essentially.

    6. PAPeter Attia1:19:24

       Well, I can think of no better way to introduce the MVX assay. So I'll tee it up for you and then take, take away the story. So by the way, I went to have my first MVX drawn. We've been doing it on our patients for about four months now, and I just haven't got around to doing a blood test on myself. So I, I went out to, to LabCorp, uh, two and a half weeks ago 'cause I wanted to have the results back when we were sitting here. Um, and wouldn't you know it, Jim, they screwed up the assay, so I don't have it. They got everything else. Every other test we ordered, they got, but they, they butchered this one, so I don't have my, my own to talk about. But the MVX is a, is a, is a composite score that measures, if I recall, six things. So small HDL-P, GlycA, which we haven't introduced yet and we'll talk about, citrate, and the three branched-chain amino acids, so, uh, leucine, isoleucine, and valine. Um, so why don't you tell us the story of how you developed this score, and I'll just give the punchline so the listener knows why they should be paying attention. This score seems to have remarkable... When it's normalized to a, you know, zero to 100 number, the higher the score, the worse it is. This score seems to have remarkable predictive value of all sorts of things we want to avoid, starting with death, cardiovascular death, liver disease. Um, while its first-- While the first study that I saw was done in a very, very high-risk group of catheterized patients, and it would be easy to dismiss that it was only valuable in that population, it's been demonstrated in healthy populations as well. So tell us about the score.

    7. JOJim Otvos1:21:08

       Yeah. So it's a really interesting story, and, you know, one question that it gets asked is, "Well, why did you think of measuring those things, or did you have some mechanistic reason for focusing on these things that ended up contributing to this MVX Metabolic Vulnerability Index score?" And the answer is, is no, that's not how it happened, and I just-- I do wanna make this point because MVX and, and, and es- especially MVX raises more questions than, than answers right now. Uh, okay? So you really would like to understand if it predicts mortality so well, why does it do that? Is it causal? Can we intervene? Well, you know, these are the things that really matter clinically. Um, so it wasn't starting out with some idea that branched-chain amino acids are really important mechanistically, even though they are. Um, this was because NMR analysis has given us a very efficient tool to measure baseline samples from studies that follow people over time, 10 years, 20 years, 30 years, longer, to see what develops. And so you, you really would like... The ultimate idea in a biomarker is that it predicts the future. You, you, you wanna know not if it's related to cardiovascular disease, but is it related to getting cardiovascular disease in the future? And so we have availed ourselves of, over the years, of baseline samples from many very large clinical studies. You mentioned MESA, Multi-Ethnic Study of Atherosclerosis. That's a NIH-funded study that started in about 2000. It's had more than 20 years of follow-up now. We were asked to measure the baseline samples 15, 20 years ago for free. Now, I work for a company that has to make some money, [chuckles] but I had enough freedom to be able to offer that test. Initially, I asked for money, but then they finally said, "Well, j- we just can't find the money for that, but we'd really like to have the information." So it was measured for free. Same thing from o- other very large studies. Women's Health Study, 26,000 women at baseline in a study that's been followed up for many, many years. Um-

    8. PAPeter Attia1:23:34

       Framingham Offspring

    9. JOJim Otvos1:23:34

       ... many clinical trials. Framingham Offspring Study. Uh, all sorts of intervention trials, JUPITER, uh, et cetera. Um, um, the Diabetes Prevention Program. I, I would like to quickly go back to that just for two seconds-

    10. PAPeter Attia1:23:49

        Of course

    11. JOJim Otvos1:23:50

        ... because we didn't finish the, the, the thought of what the Diabetes Prevention Program, even though it's not published, tell us. And the important thing that it told us was not just that LP-IR score goes down with lifestyle intervention and branched-chain amino acids go down when, when-- which is all good thing. Um, but that those things going down, the delta between where it started and where you got after the intervention very powerfully predicts incident diabetes or, or the lessened diabetes risk. So that was sort of the missing link. If you really wanna argue causality, um, you really wanna show that lowering LP-IR translates to lower diabetes risk, and it does. It's a, it's quite powerful. [chuckles] It, it hasn't been published yet, in part because the, the, the primary author, uh, died recently, unfortunately.

    12. PAPeter Attia1:24:43

        Mm.

    13. JOJim Otvos1:24:43

        And so, uh, it will be, uh, eventually. All right. But that's one study. So we, we've gone out of our way over time to make it possible for people that didn't have funding to get NMR data on baseline samples for observational studies and intervention studies going forward. MESA has been particularly useful to us. Six thous- 7,000 people, baseline NMR, and because we supplied the assay for free, the NIH has given us access to the information about who developed cancer, who developed cardiovascular disease, uh, et cetera, dementia, et cetera, any, any, all sorts of outcomes. And so the beauty is that once we have taken the NMR spectrum and gotten from it what we initially wanted to get from it, let's say LDL-P, these spectra are sitting there in a computer, in a-- stored on a disk, and when we develop the wherewithal to extract new information from the NMR spectrum, we can go back literally in a couple hours to interrogate a very large clinical trial and get prospective information as to whether it's predicting a particular outcome. So back to MVX, that's exactly what we did in the, this cardiac catheterization study at Duke University. Seven thousand people over a period of years recruited into this biorepository. When they came to the cath lab, when they came to the cardiologist with chest pain or some issue that, uh, qualified them to have coronary angiography done, and their blood was taken at baseline, and it's stored at minus eighty, where it's perfectly stable for NMR analysis. So we got, uh, a, a, a relationship with Duke to, to obtain those samples, do NMR analysis, and we found, um, in, in, in published papers that small HDL particles were particularly powerful in pr- in, in, in relating to, uh, the likelihood somebody would die during the roughly five or ten-year follow-up of this study.

11. 1:26:54 – 1:35:42

    GlycA as an NMR-derived marker of systemic inflammation: its discovery, biological basis, and advantages over hs-CRP

    1. JOJim Otvos1:26:54

       And then there was another thing that we could measure by that time called GlycA, and we might as well talk about GlycA now.

    2. SPSpeaker1:27:02

       Let's, yeah.

    3. JOJim Otvos1:27:02

       Um, it's another signal in the NMR spectrum that's not where the signal is that we interrogate for LDL-P. And so for ten or fifteen years, we didn't care a whit about any of those other signals. But, uh, a funny thing happened with the, where the, the, the GlycA signal is. It, it was basically superimposed or in the way of another broad signal that we were trying to interrogate to learn about the fatty acid composition of somebody's plasma, how much mono/polyunsaturated fats were in the blood. And it-- to interrogate that, the signal, the sharp signal was on top of that and was getting in the way. So we basically worked out a way of quantifying how big that signal was so we could subtract it from the other signal. But that was perfectly good quantification information. So that signal, which I wouldn't have, I, I wouldn't have thought to do anything with, the person who's, uh, does all my, um, uh, epidemiologic analyses, Irina Shalyrova is her name, uh, we had the Mesa data sitting there. We could go and very quickly use the software we had developed to measure this g- sharp signal. And she just said, "Well, what it-- does it relate to anything that happens in the future in Mesa?" Oh my gosh, it relates to [chuckles] uh, th- all sorts of things, including mortality very, very strongly. So what's up with that signal? Well, it turns out, going to the literature, this often happens. Twenty years ago, somebody published a paper saying that the signal was an inflammatory marker. Um, it, it actually comes from the carbohydrate, the glycan decoration that's on a lot of proteins in the blood, and most of these proteins are so-called acute phase proteins that, uh, increase in concentration in inflammatory conditions. And so even though this signal was not telling us which acute phase proteins were contributing to it, it, it was a composite, and not only did it essentially quantify the most abundant four or five acute phase proteins that contributed to this signal, but this carbohydrate decoration, this glycan decoration is used for all sorts of purposes, signaling of different types, et cetera. So there's very complex, y- you know, th- there's... People worry about the glycome. It's like the proteome-

    4. SPSpeaker1:29:39

       Yeah

    5. JOJim Otvos1:29:39

       ... and the genome. There's a glycome. I know nothing about [chuckles] any of this. But one thing that, that, where we know where the signal comes from, and it comes from a particular sugars on, on this carbohydrate. And inflammatory conditions, more of this decoration is put on some of these proteins. So actually, this GlycA signal is reflecting not only the levels of these acute phase proteins but how much glycan is on them, which is also connected to inflammation. So it turns out, rather miraculously, that how big this signal is is a very useful measure of your steady state, of your systemic inflammation level. It's a very stable parameter because it's the integration of lots of different things. So rather than CRP that you typically measure clinically to assess inflammation, it's very volatile. It goes up and down day to day, so all clinical recommendations for the use of CRP information say that you should take the average of two or three different measurements. Never-- nobody does that, [chuckles] but that's the recommendation to get around some of this biological variability. GlycA doesn't suffer from that problem. Part-- That's probably in large part the reason that when added to CRP in a prediction model, it typically assesses the outcome more strongly than CRP does. But CRP tends to independently add, so inflammation is a very complex thing. This is a very unspecific marker But it's a very stable and useful clinically marker of systemic inflammation.

    6. PAPeter Attia1:31:21

       Does it include, do you think, or capture what we see in the various interleukins?

    7. JOJim Otvos1:31:26

       Yes. So it's, it's correlated strongly with IL-6 and other interleukins. Um, uh, and, and these correlations-- So, but that's all we know. I mean, so y-yes. Um, and you'd really like to be more specific and, and if there's local inflammation as opposed to systemic inflammation, this is not gonna tell you anything, okay? But what it does do is offer you a simple and very cheap, because it comes along for the ride with the other NMR information, uh, if you quantify this GlycA signal, you have a very powerful marker of the things that systemic inflammation contributes to.

    8. PAPeter Attia1:32:06

       Now, um, using hs-CRP as an example, Jim, which, um, as you pointed out, is gonna rise with inflammation, one of the things that clinically we pay attention to is how high is it. So if, um, if I see a brand new patient and their hs-CRP is two or two and a half, in many ways, that's more disconcerting to me than if it's 40. Because the 40 is so high that I know it's really in response to something acute. They're probably getting over a cold. Maybe they got a vaccine two, you know, four days ago or something like that. Um, now of course, that doesn't obviate the point you made, which is I still wanna see longitudinal data. Like, I can't make a d-- I can't assume the two is bad because I could also be catching something on the way down or on the way up.

    9. JOJim Otvos1:32:58

       Could be three or one depending on the next day. Yeah.

    10. PAPeter Attia1:33:01

        Exactly. But, but it's usually the case that when something is very, very high, it really speaks to acute inflammation, which is less pathologically concerning. These low simmering ones that I see, those are the ones that, that give me pause. Is the GlycA the same as that?

    11. JOJim Otvos1:33:18

        No, it's very different than that. Uh, so CRP levels could go up a thousandfold-

    12. PAPeter Attia1:33:23

        Yeah

    13. JOJim Otvos1:33:23

        ... on an infection. GlycA levels are-- the response is much more muted.

    14. PAPeter Attia1:33:28

        Because it contains so much information in it, you think?

    15. JOJim Otvos1:33:31

        I can't really answer the because [chuckles] as well.

    16. PAPeter Attia1:33:34

        Yeah. It's just the observation.

    17. JOJim Otvos1:33:35

        I'm just-- Just the observation.

    18. PAPeter Attia1:33:36

        Yeah.

    19. JOJim Otvos1:33:36

        And, uh, and so i-it, it's true that if somebody has an active infection and you're trying to relate the GlycA level to mortality in, in people that didn't have an infection, you're gonna be misled by that. It'll be higher by twofold, not a thousandfold. So it's not immune to those, those changes, but you wouldn't want it to be because, you know, it's an inflammation marker.

    20. PAPeter Attia1:33:58

        Yep. Yep.

    21. JOJim Otvos1:33:58

        So in, in people with inf-inflammatory diseases, rheumatoid arthritis, psoriasis, et cetera, GlycA levels are, are, are significantly elevated. And, um, and they are responsive to anti-inflammatory treatment. And so it, it has all the characteristics of a, of a useful biomarker to assess not only things that, that, that you would like to have some visibility to, like systemic inflammation. I think we're coming to b- to understand that that's an awfully important contributor, even though we don't understand the mechanistic-

    22. PAPeter Attia1:34:31

        Yeah

    23. JOJim Otvos1:34:31

        ... fine points, uh, to so many things, including, uh, mortality risk. So anyway, this is just-- I, I just wanted to tell you that, that, that, that the way that we discover these biomarkers is different than the way other people discover a lot of biomarkers. So it's more the top-down, where we have a very efficient way of doing the epidemiology. So we know already that these markers have a strong relationship to human health conditions. A lot of the work that's done starting at from bottom up with a mechanistic idea and a particular enzyme that you might maybe wanna target as a therapy, um, you have to then do animal models, and then you go up to humans, [chuckles] and then you have to do expensive trials, and then you typically measuring these things is not as easy as it is to measure these things by NMR. So it's a completely different approach to discovery, and it's sort of irrational discovery because you're basically using these large databases of population studies and then discovering things that you really didn't go after discovering

12. 1:35:42 – 1:46:00

    Developing the Metabolic Vulnerability Index (MVX): biomarkers of inflammation, malnutrition, muscle wasting, and mortality risk

    1. JOJim Otvos1:35:42

       in the first place.

    2. PAPeter Attia1:35:43

       So there are three components to the MVX. You've already talked about the lipoprotein one, which is the small HDL-P. You've just es-explained the inflammatory one, which is GlycA. The third one is sort of the metabolic one, which has the citrate and the three BCAAs. So how did you come to figure those out?

    3. JOJim Otvos1:36:01

       So you're right. So we actually, with the Duke collaborators, there had been a paper published showing that, that GlycA, there was a GlycA paper predicting mortality. The interesting thing about this cath gen population is they came to the cath lab because they had some presumed cardiac issue. They have, uh, in the study that we did, uh, for five-year mortality, 17% of the people died in five years, and they were about mean age of 60 coming in. So that's a pretty high-

    4. PAPeter Attia1:36:33

       How many? Fifteen percent?

    5. JOJim Otvos1:36:34

       Fif- 17.

    6. PAPeter Attia1:36:35

       Fine. Seventeen percent.

    7. JOJim Otvos1:36:36

       Seventeen percent mortality, five years, in people that were not-

    8. PAPeter Attia1:36:41

       Wow

    9. JOJim Otvos1:36:41

       ... elderly. Um, three or four times higher for sure than, than a regular population. So the presumption was that these are people that died of cardiac causes, but less than half of them died from cardiac issues. Uh, m- m- 60% died of non-cardiac, cardiovascular, uh, causes.

    10. PAPeter Attia1:37:04

        It's very interesting. Seven-- So, so 60% of the 17 people-- 17% of people who were dead by 65 were non-cardiac deaths.

    11. JOJim Otvos1:37:11

        Non-cardiac causes, yeah. So anyway, the mortality was the most prevalent outcome. It wasn't a new myocardial infarction or a, a recurrent myocardial infarction. These people died. [chuckles] And- And so GlycA predicted it, small HDL particles predicted it. We might talk more about that. And then we simply looked at all the other things that we had learned to measure, and we had just kind of started this activity. So branched-chain amino acids, isoleucine, leucine, and valine. Um, ketone bodies. Um, um, plasma protein. I mention those two things because when you look individually at those things, ketone bodies and p-plasma protein, they have significant associations with mortality. So why didn't we use those as part of the MVX composite biomarker? Because we were looking for things that contributed independently and additively to the other things that we've already talked about. So the inflammatory part, GlycA and small HDL-P, we created a subscore called the Inflammatory Vulnerability Index, IVX. The other four seemed to relate-- So we, we, we discovered that these branched-chain amino acids and citrate, independent of GlycA and small HDL-P, added to the prediction of mortality. Then it was, "Okay, [chuckles] why?" And so then we went to the literature. So we, we really approached all-

    12. PAPeter Attia1:38:43

        And, and you're doing all of this inside of LabCorp. So LabCorp, at least still at this point in time, had the appetite for-

    13. JOJim Otvos1:38:50

        Yes. Although-

    14. PAPeter Attia1:38:50

        ... the R&D

    15. JOJim Otvos1:38:51

        ... really, yes, they paid our salary, but that was-- [chuckles] We were sort of left alone on, on, in this little building [chuckles] to continue what we were doing at, at LipoScience. So thanks to LabCorp for, for not getting rid of everybody. So then it was going to the literature, trying to figure out, does this make sense biologically that these things might be related to mortality? And that's where you come to the literature that, uh, very powerfully speaks to the great mortality risk that people with several acute diseases, especially kidney disease and dialysis patients, heart failure patients.

    16. PAPeter Attia1:39:30

        Anything with cachexia, I could see increasing-

    17. JOJim Otvos1:39:32

        A-anything with cachexia.

    18. PAPeter Attia1:39:34

        Yeah.

    19. JOJim Otvos1:39:34

        So does, you know, sarcopenia. Um, old people. [chuckles] Okay? So there's a lot of literature, especially in the kidney disease literature-

    20. PAPeter Attia1:39:45

        Mm-hmm

    21. JOJim Otvos1:39:47

        ... that describe this vulnerability as coming from a so-called malnutrition inflammation syndrome, and it's called many other things. Protein energy wasting syndrome. So the cachexia, the wasting syndrome, is part of this, but inflammation is part of this. Inflammation is probably the, the context that, that allows these, these dysregulated metabolistic, metabolism situations to exist. So the fact that malnutrition inflammation syndrome, we had the inflammatory parts, we speculated, but the, the branched-chain amino acids were related, and they were related in the opposite direction that branched-chain amino acids are related to diabetes risk. So high branched-chain amino acids speak to insulin resistance, obesity, diabetes risk. Low branched-chain amino acid levels speak to mortality risk. We can talk more about-- Y-you know probably much more about why, uh, this makes sense in terms of mechanism, because it's partly related to mTOR signaling and, and the whole, uh, skeletal muscle-

    22. PAPeter Attia1:40:58

        Yeah, turnover of amino acids-

    23. JOJim Otvos1:41:00

        Turnover and catabolism-

    24. PAPeter Attia1:41:00

        ... and the muscle protein synthesis

    25. JOJim Otvos1:41:01

        ... et cetera. Um, but it was, it was satisfying to find this literature and to say, "Oh, maybe these are just better biomarkers of something that's already understood in the acute context, acute cl-clinical context." But in the CATHGEN population, w-nobody had, had described this in a cardiac-- uh, well, in car-cardiovascular cohort. Um, and then, uh, so we, we said, "Well, this exists in spades, apparently, in this, this, this presumed cardiovascular cohort." But then we did subgroup analysis within this 7,000 people in the CATHGEN study, and, and we asked, "Does this MVX association with mortality exist equally strongly in men and women, in people with and without obesity, in people with and without diabetes, with and without heart failure, with and without a previous myocardial infarction, with or without, uh, coronary catheterization, uh, uh, occlusion of the coronary arteries, uh, with and without kidney disease?" [chuckles] And it's, it, it doesn't, it, it's not affected by any of those things. It's equally strong, if not stronger, in people without the chronic disease versus those that are disease-free. Um, so in fact, the strongest relationship of MVX to mortality, you mentioned how exquisitely strong it is.

    26. PAPeter Attia1:42:30

        I thought, I thought that the hazard ratios were bigger in the CATH study than in the MESA study of healthy people.

    27. JOJim Otvos1:42:36

        That's true. That's true. And, and, and the reason is that, for whatever reason, if these people came to the cath lab and they were enriched in people who ultimately suffered from these wasting syndromes, okay? And the fact that you see the same prediction, if not as strong, for sure, in people with absolutely no evidence of any chronic disease. In fact, the studies that were most recent, most recently published, and the most interesting one that I'll mention is one that's not yet published but is about to be submitted for publication. Uh, MVX in young people, 30-year-olds. Um, would we expect to see the relationship? And you do see the relationship, but it's weaker, and the I, I, I didn't sort of finish that. So the two sub- sub-parts of MVX are the IVX inflammation part, and the other four parameters brewed together to form MMX, metabolic malnutrition index. Um, it's kind of an arbitrary thing to, to talk about these two parts of MVX because it's known that there's synergy between these. It's a syndrome. It's interrelated, intertwined. But on the surface at least, it looks like it might be useful to take a h- high MVX score, and it might be due more to inflammatory reasons than, than the, than the metabolic malnutrition wasting reasons. In, in which case, different therapies might be better suited for that person rather than somebody with the same MVX score with a more malnutrition issue. So we thought it might be clinically useful. That's what-- why we did that.

    28. PAPeter Attia1:44:24

        Are they both... I know the aggregate score is reported zero to 100. Do the two sub-scores get reported that way as well?

    29. JOJim Otvos1:44:30

        Yes. Same way.

    30. PAPeter Attia1:44:31

        Okay. So you could say, uh, you know, "Mr. Smith is a 75, which is very high risk. But when I look at his aggregate score, his IVX is only 25. His MMX is 80."
13. 1:46:00 – 1:52:25

    What MVX can tell us about longevity and mortality

    1. PAPeter Attia1:46:00

       Well, let's, let's talk a little bit about that. Uh, the one that just came out, of course, was the MASIL-D paper. Is that the one you wanted to chat about?

    2. JOJim Otvos1:46:07

       No, actually.

    3. PAPeter Attia1:46:08

       Oh.

    4. JOJim Otvos1:46:08

       We could, and it just so ha-- Because these papers are coming out now.

    5. PAPeter Attia1:46:12

       They're just coming out at a geometric rate.

    6. JOJim Otvos1:46:14

       Yeah, and, and, and that-

    7. PAPeter Attia1:46:15

       Which is great. I mean, that's so exciting

    8. JOJim Otvos1:46:15

       ... and the reason, the reason is that we just have to go back to existing NMR data. So we're mining other people's costly studies. We're piggybacking on their work, their, their funding, and, you know, getting very quick gratification about whether MVX is good for this or that or the other thing.

    9. PAPeter Attia1:46:35

       Is there one in an intervention study? 'Cause that would be the next step, is... I mean, maybe you've already done it, where you say, "Look, we've got a MVX at baseline. Ha- we're gonna treat you guys placebo, non-placebo."

    10. JOJim Otvos1:46:46

        Yeah. That's-

    11. PAPeter Attia1:46:47

        You know where I'm going.

    12. JOJim Otvos1:46:48

        Absolutely. [chuckles] That's, that has to happen.

    13. PAPeter Attia1:46:50

        Okay.

    14. JOJim Otvos1:46:51

        It isn't gonna happen if nobody knows about MVX. And the thing that was-- has been a little disappointing to me is how little... I mean, nobody read that 2023 paper. I was very, that proud of that paper. I thought there was a, there was a lot of meat in that paper and a lot of implications clinically and otherwise. Um, but, but people read papers that are called to their attention, and this is part of the problem. If this was a LipoScience , if we were at LipoScience , we would be promoting awareness of this paper in cl- in meetings, et cetera. Um, and this isn't happening at LabCorp, [chuckles] okay? So, uh, but now this will help. Uh. [laughs]

    15. PAPeter Attia1:47:35

        Yeah. I mean, there's gonna be a lot of people-

    16. JOJim Otvos1:47:36

        This will help

    17. PAPeter Attia1:47:36

        ... that are, that are listening to us, so then-

    18. JOJim Otvos1:47:38

        Right.

    19. PAPeter Attia1:47:38

        Yeah.

    20. JOJim Otvos1:47:38

        And so, uh, yes, w- we would... That's the next, next step. What, what will be published this year will be, I think, sufficient for anybody to see the replication of the phenomenology. Uh, fairly quickly after that first paper in heart failure patients, we could show this. The other thing that was kinda nice about that heart failure population is they also had a lot of frailty information. A lot of these were older people with heart failure in, in Minnesota. And so they were able to calculate frailty scores and, um, also biological age by, uh... Well, not in that study, but another study of older people. So some of the other things that are talked about a lot about, uh, relating to longevity and, and, and so on were measured in, in these studies that we've been able to look at MVX in. And in terms of frailty, physical frailty, um, correlation, but fairly weak with MVX, about a 0.2 correlation coefficient. So it isn't like, you know, you need to have physical frailty to see the MVX be high or vice versa. Um, and in terms of mortality risk, frailty, physical frailty on top of MVX score adds considerably. But MVX very powerfully still, in the presence of frailty score, predicts mortality.

    21. PAPeter Attia1:49:13

        And is it always a five-year look forward?

    22. JOJim Otvos1:49:16

        No, and, and in many studies, it's, it's longer than that, in part because you want the statistical power from more people having the outcome.

    23. PAPeter Attia1:49:24

        But there's something powerful about the short window.

    24. JOJim Otvos1:49:27

        Oh, absolutely.

    25. PAPeter Attia1:49:27

        I mean, in many ways, that's actually a feature, not a bug-

    26. JOJim Otvos1:49:29

        Absolutely

    27. PAPeter Attia1:49:30

        ... if you can offer that insight.

    28. JOJim Otvos1:49:31

        Absolutely.

    29. PAPeter Attia1:49:31

        'Cause we don't have many short-term predictive biomarkers.

    30. JOJim Otvos1:49:34

        Absolutely. And actually, uh, the, the one study that I was sort of confusing with the, the heart failure study is, is, is a study called EPIS. Uh, and it's a, it's a a study of older people with lots of things being measured, including these biological age measures like chemistry assaying, uh, not the epigenetic flavors. And a couple years ago, I think it was 2022, they had NMR information, and they had all this frailty information and, you know, 186 different variables. And it was done by some, um, um, epidemiologist, epidemiologists in Minnesota using the most, uh, high-tech ways of trying to de- deduce whether the associations were causal or not. I'm, I'm not really sure that that really, to me, demonstrated that. But they used methodology that purports to assess causal relations. And out of those hundred and 86 things that were looked at, small HDL particles were the most powerful at predicting two-year mortality in these people.
14. 1:52:25 – 2:02:17

    How MVX may reveal metabolic frailty and predict premature mortality decades in advance in young, healthy adults

    1. JOJim Otvos1:52:25

       reduction in mortality risk?

    2. PAPeter Attia1:52:26

       Before you do that, Jim, I wanna go back to the 30-year-olds because we didn't really finish the swing on that. So-

    3. JOJim Otvos1:52:32

       No, I... No, we didn't. But-

    4. PAPeter Attia1:52:34

       Can... Are you able to talk about that? I... Or is that not published?

    5. JOJim Otvos1:52:36

       I'm, I, I'm gonna talk about it and I'm not gonna say the name of the study, but-

    6. PAPeter Attia1:52:40

       Okay

    7. JOJim Otvos1:52:40

       ... um, and the paper, I just, you know, the, the draft of the paper is about to be submitted.

    8. PAPeter Attia1:52:45

       It's just hard for me to wrap my head around the fact-

    9. JOJim Otvos1:52:47

       No

    10. PAPeter Attia1:52:47

        ... that any biomarker in 30-year-olds could predict anything.

    11. JOJim Otvos1:52:50

        This is very true [chuckles] and very... And that's why this is so interesting and novel. So the paper that appeared a couple months ago was from Mesa. So we've been talking about Mesa. 60-year-old people at entry. They weeded out all the people in Mesa that had any self-reported or otherwise diseases, so restricted to healthy. And average age 60 or so. And MVX by quartile had this stepping stone relationship. Not as strong. The hazard ratios weren't as different as in CASGEN, but very significant. So that was the first paper.

    12. PAPeter Attia1:53:32

        Do you recall in that study, Jim, what the, uh, the difference was between the first and the fourth quartile in hazard ratio? Were l- you talking about like a one point six?

    13. JOJim Otvos1:53:42

        So unadjusted, unadjusted, it was about, um, I might be con- they're confusing other studies. But adjusted, it was like one point five to two.

    14. PAPeter Attia1:53:50

        Okay. And we'll link to every one of these studies-

    15. JOJim Otvos1:53:53

        Yeah

    16. PAPeter Attia1:53:53

        ... in the show notes.

    17. JOJim Otvos1:53:54

        Yeah.

    18. PAPeter Attia1:53:54

        But basically, in otherwise healthy 60-year-olds, the difference between being in the worst quartile, 75 to 100 score, versus the bottom quartile, zero to 25, would be about a 50 to-

    19. JOJim Otvos1:54:07

        No, no, that-- I was-- That's per standard deviation. So this, this is-- it's actually greater than that. It's, it's maybe-

    20. PAPeter Attia1:54:12

        Ah

    21. JOJim Otvos1:54:12

        ... two to threefold greater.

    22. PAPeter Attia1:54:14

        By quartile.

    23. JOJim Otvos1:54:15

        By quartile.

    24. PAPeter Attia1:54:15

        Top to bottom quartile.

    25. JOJim Otvos1:54:16

        Yeah. Yeah.

    26. PAPeter Attia1:54:16

        Got it. Wow. Okay. So big, big difference.

    27. JOJim Otvos1:54:18

        Yeah. So, okay, so those are 60-year-olds. So these are people-- So the-- My idea about MVX, I mean, people die when they're older. [chuckles]

    28. PAPeter Attia1:54:28

        Yeah.

    29. JOJim Otvos1:54:28

        And so this thing, MVX, comes into play when you're older, and maybe MVX scores go up with age. MVX scores are virtually unassociated with age.

    30. PAPeter Attia1:54:39

        What?
15. 2:02:17 – 2:10:16

    Potential applications of MVX for predicting treatment response, surgical resilience, and clinical trial outcomes, and the barriers to broader use

    1. JOJim Otvos2:02:17

       the death.

    2. PAPeter Attia2:02:17

       So listening to you say this gives me an idea for a study that I'd love to see you do, Jim. So you're-- I don't know how much time you spend in the oncology world, but I'm sure you're familiar with Keytruda. It's the sing-- I-- To my knowledge, Keytruda is the single best-selling drug of all time. Um, and in many ways, it's been a miracle drug in oncology, the single most exciting development in cancer in the last 25 years. For folks unfamiliar with it, this is a checkpoint inhibitor. So people that have a PD-1 mutation that take this drug, regardless of what kind of cancer they have, this could be pancreatic adenocarcinoma, lethal cancer. If you have this mutation, this drug basically takes the brakes off the immune system, and your immune system eradicates the cancer. But here's the question: Why does someone-- Why do you take-- Why could you take two people that have the exact same PD-1 mutation, the exact same cancer, by all intents and purposes, and you give them both Keytruda, and one of them responds, and one of them doesn't? Like, we don't know. We, we do not understand what's happening at the immune level to understand why that's happening. It would be very interesting for me to understand, and using Keytruda as an example, but you could do this with any therapeutic intervention where mortality is very quick, right? And you could ask the question, does this become a prognostic indicator of not just mortality, but probability of success of an intervention?

    3. JOJim Otvos2:03:54

       Yes. Yeah, ex- ex-- that's precisely what, what possibilities exist. The-- When I first talked about this to people at Duke, uh, the collaborators of the CathGen study, um, the people around the table, the first thing they said was, "Wow, this would be a great test for surgeons who are asked to operate on people who are frail or are less likely to survive the surgery or to benefit from the surgery." You'd like to be able to screen them for resilience somehow, but there are no biomarkers, good objective biomarkers to do that. Um, malnutrition, metabolic malnutrition, there, there are people, I've read papers where people are-- surgeons are suggesting that, that people really should avail themselves of these, uh, you know, interrogating whether somebody i-i-is, is sort of metabolically or, or, you know, physically frail, um, and ha- and has evidence of wasting. But if there's a metabolic component to that, that is accessible via MVX, it could be very useful. Uh, and this is one of probably all sorts of possible, um, applications. But you mentioned this paper that was just published two days ago, uh, on, uh, MASLD. [laughs]

    4. PAPeter Attia2:05:25

       MASLD, yeah. The for-former-- the artist formerly known as NAFLD.

    5. JOJim Otvos2:05:29

       Formerly known as NAFLD. So K-- uh, so liver disease. And, um, the paper speaks very forthrightly about the possibility of using MVX to, uh, for, for entry into clinical trials.

    6. PAPeter Attia2:05:46

       Am I, am I correct in remembering this paper, which I skimmed, so I, I'm ashamed to admit I didn't read the paper, but I could have sworn it said that MVX added predictive value to FibroScan in predicting subsequent fibrosis in the MASLD patients. Is that-- Does that ring at all?

    7. JOJim Otvos2:06:04

       I think that's true. I only skimmed it as well. [chuckles]

    8. PAPeter Attia2:06:06

       Okay. Yeah, it just came out.

    9. JOJim Otvos2:06:06

       I was not a co-author on that paper.

    10. PAPeter Attia2:06:08

        So yeah.

    11. JOJim Otvos2:06:08

        Yeah. But, you know, as you well understand, so many clinical trials are expensive and, and are not done. Many are not done because the events are too rare.

    12. PAPeter Attia2:06:21

        Yep.

    13. JOJim Otvos2:06:21

        And so you need huge populations.

    14. PAPeter Attia2:06:23

        Or too far away.

    15. JOJim Otvos2:06:23

        Or too far away. And so being able to juice up your likelihood of people dying, for example. So having-- And then-- But mortality is, is probably the endpoint that people care most about, right? And so, uh, it, it has sort of in the hierarchy of, of events, it, it's gr- it's, it's people care more about dying than they do about getting an MI or getting diabetes or whatever. So anyway, there's all sorts of possibilities, but we're just at the beginning of the trail of answering the questions that-- and I don't even know all the questions that could be posed, but this really is very fascinating. And, and the fact that it was discovered fairly serendipitously by interrogating these epidemiologic data sets, and then the relationships seem to make sense in terms of what's been published about the detailed, you know, cell biological mechanisms, which I don't understand nearly well.

    16. PAPeter Attia2:07:26

        And it's, and it's an inexpensive test. It's a-

    17. JOJim Otvos2:07:28

        Well, it's, [chuckles] it's basically free if you think about it. So I mean, this is what-

    18. PAPeter Attia2:07:33

        How many tubes of blood do you need to run it?

    19. JOJim Otvos2:07:35

        No, so, uh, so you mentioned that you got your MVX score, and they probably drew an extra tube of blood for that. They don't need to do that. The same specimen, actually 150 microliters of plasma, produces the s- the NMR spectrum that produces the NMR lipoprofile. produces glucose, produces LPIR, produces GlycA, produces MVX. All of that comes from the same analysis. Uh, [chuckles] and when done in high volume settings, these are tests that are, that really literally cost a dollar or less. Okay. But this is the problem commercially, and this is the problem with our healthcare system and the way things are set up, that there's sort of no, [chuckles] uh, there's almost a disincentive to provide analytically free information if you can't charge incrementally for it. And what you'd like to do, if you're a company, is charge a whole lot more for it, and then you have the tension between what you'd like to charge and what the insurance wants to pay for. And then convincing the insurance company that it's worth paying for is what keeps you from being successful commercially in, uh, you know, producing this test globally, you know, broadly. I had this experience with LDL-P. [chuckles] Tremendous resistance to pay. So, you know, I think the way around that, one way around that is to not try to get pay, paid incrementally for it and just do something that... So the analogy is, is the comprehensive metabolic panel that you get done every time you go for a... There's 14 things that are measured there. If you add up what the CMS reimbursement rate is for those 14, it comes to $60 plus and change. CMS pays $12 for that, and it's because these are all done at the same time. They have some, you know, clinical reason to be done at the same time, and economies of scale make it efficient enough that you can make money and people don't, aren't gonna starve producing this test, getting paid $12, as much as they'd like to get paid a lot more. This is a situation just like that, where [chuckles] the information is essentially free. And we made that point in the paper we wrote about, um, the lipid panel, the extended lipid panel that includes ApoB. I mean, I'm-- You uh, the, the... We didn't really get to the ApoB.

    20. PAPeter Attia2:10:09

        It was actually the last thing I wanted to get back to-

    21. JOJim Otvos2:10:11

        Yeah. Let's-

    22. PAPeter Attia2:10:11

        ... which was the discordance between ApoB and LDL-P.

    23. JOJim Otvos2:10:13

        Let's get back to that because the ApoB

16. 2:10:16 – 2:16:01

    ApoB versus LDL-P: their clinical similarities, the additional information provided by NMR, and barriers to broader adoption

    1. JOJim Otvos2:10:16

       story is, is the same as the LDL-P story. And the reason that I have partnered rather than competed against Alan Sniderman, who's the biggest proponent of ApoB, is that it would be disingenuous to, you know, say that one is really better than the other. I could make the case that the NMR analysis tells you LDL-P but also TRLP, triglyceride-rich particles. Subspecies might be differentially related. There are people publishing papers that suggest that's true. So you could definitely be ahead of the game with, with more information than ApoB provides. ApoB is just a single measure of all the ApoB on LDL and VLDL particles. But the challenge is convincing people that you should do something other than measure cholesterol. [chuckles] And, and so you need as, as many, um, people in that fight as possible. So Alan and I are both telling the same story, and that's why, uh, we transitioned. I mean, I have advocated that we use NMR to produce ApoB, which, which-- And actually, we got FDA clearance for the quality of the ApoB information that comes from the NMR spectrum via the machine learning approach. So the idea was the extended lipid panel would have no analytic cost associated with adding ApoB to a lipid panel. Now you have a better lipid panel. The way that, uh, Medicare reimbursement is set up now, ApoB gets paid 20 bucks, lipid panel about 13 bucks. Last I looked, it might have changed a little. Um, so that more than-- If you wanna add ApoB to make the lipid panel better, it's more the, more than double the cost to the, to the payers. The payers aren't gonna wanna do that.

    2. PAPeter Attia2:12:03

       And what's the CMS reimbursement on the NMR of lipids?

    3. JOJim Otvos2:12:07

       It's about $30 and change. And [chuckles] so that's for the NMR LipoProfile. The NMR LipoProfile comes with LP-IR.

    4. PAPeter Attia2:12:19

       Okay. Yep.

    5. JOJim Otvos2:12:20

       Okay. We couldn't get that FDA cleared at the time, but you might know that a lot of laboratories can offer tests that are not FDA cleared through sort of a loophole in the, in, uh, or the FDA ha- has decided to exercise, um, discretion about whether they, uh, will enforce this or not. So, um, laboratory developed test, LDTs, individual laboratories can develop their own test, go through a, a, a, you know, get CLIA certification. So this is more laboratory certification for the, the, the, the, the, the, how well they perform the test. But the actual demonstration of the clinical utility of the test is something that FDA cares about, but CLIA doesn't care about. And so it's an easier path to offering commercially a test that doesn't have to go through FDA clearance. Uh, and so LP-IR was, was added to the NMR LipoProfile as an LDT, not a, not part of what was cleared. LDL-P was cleared, but [chuckles] not without great difficulty. Um, so anyway, um, a lot of, a lot of the reason that NMR wasn't commercially successful has to do with what I just explained about the resistance of payers to pay any increment to what they're paying for now. Uh, and, and if you really need to demonstrate... Uh, but the path to, to, to getting insurance to pay is to get some, um, advisory panels or some clinical guideline group to bless it And, and Alan Sniderman can speak to the difficulty of having ApoB blessed by the cholesterol, uh, guidelines. Uh [chuckles]

    6. PAPeter Attia2:14:14

       Although the European guidelines have.

    7. JOJim Otvos2:14:15

       The European guidelines and now the US guidelines are, you know-

    8. PAPeter Attia2:14:19

       Yeah

    9. JOJim Otvos2:14:19

       ... getting, but it's still, it's just ridiculous. But part of it is because the guideline writers are trying to protect the payers, you know, uh, which that shouldn't be their job. They should be assessing the clinical utility only and let capitalism, uh, worry about... The actual Medicare reimbursement cost for ApoB, which was set many, many years ago, has nothing to do with what it costs to do these immunoassays on these modern analyzers. Um, [chuckles] so again, there's this complete disconnect between what's, what's, what's charged and what's paid for and what it costs to measure. It's the same thing that drug companies, you know, are defending the prices that they pay, uh-

    10. PAPeter Attia2:15:00

        That's fair, yeah

    11. JOJim Otvos2:15:00

        ... to support the research, et cetera. So you can make the argument that, you know, you need to stay in business, you need to, to make more money. But anyway, um, I, I think I, I, I fa- I succeeded more as a, a scientist than as a, uh, an entrepreneur [laughs] in what happened to, um, to Liposcience, because I, I, I really, we, we really had gotten quite far down the road of making NMR testing broadly available to the benefit of so many people internationally, and that just got, you know, tanked when, when it was purchased by a lab testing company instead of an IVD company. So I hope anybody listening who has-

    12. PAPeter Attia2:15:42

        Well, I mean-

    13. JOJim Otvos2:15:43

        [chuckles] a few hopes

    14. PAPeter Attia2:15:43

        ... yeah, I don't, I don't know that that door is closed indefinitely. I, I, I think-

    15. JOJim Otvos2:15:46

        No, not at all

    16. PAPeter Attia2:15:47

        ... I think that the MVX test offers a very compelling reason why another company might want to come along and, and purchase those assets and, and-

    17. JOJim Otvos2:15:57

        I think you're right

    18. PAPeter Attia2:15:57

        ... especially given the prognostic, um, utility of that test.

17. 2:16:01 – 2:23:53

    Interpreting the effects of the CETP inhibitor, obicetrapib, on LDL-P, apoB, and small HDL particles

    1. PAPeter Attia2:16:01

       So let's talk now about the, this edge case of CETP inhibition, and one of the first things that stood out to me looking at the Broadway and Brooklyn trials, which were the phase three trials of obicetrapib, were that the reductions in LDL-P and LDL-C were greater than the reductions in ApoB, if memory serves correctly, uh, correctly. What do you think is happening there?

    2. JOJim Otvos2:16:32

       I know what's happening. [chuckles] Um, so the NMR analysis, first of all, was using an older algorithm than the one that we've been using for the last five years. But even in the older algorithm, the issue of whether NMR can reliably quantify these very abnormal HDL particles that are produced by CETP inhibition. So HDL cholesterol doubles or more than doubles, not because the number of HDL particles doubles. It's i- in fact, uh, the number of HDL particles actually goes down a bit overall. So what happens with CETP inhibition is smaller particles are made into larger particles. So the number of small particles goes down, number of large particles goes up. These contain, the large particles contain five or 10 times more cholesterol per particle than the small ones, so HDL cholesterol goes way up, um, and HDL particle number does not. But the problem in terms of the analysis is that there's a natural... So again, we're, we're taking advantage of NMR signals from the different size lipoproteins being detectable and differentiated from their neighbors, right? And so at the interface of small LDL, LDL gets so small, and then the largest HDL is its nearest neighbor, and there's a decent gap between-

    3. PAPeter Attia2:18:11

       Yeah

    4. JOJim Otvos2:18:11

       ... the diameters of those particles, so they don't get confused normally. But when you've got CETP inhibition creating human beings that don't exist naturally and have HDL cholesterol of 120, 30, 50, um, your HDL particles get perilously close to the size of small LDL particles, and now NMR has the po- the possibility of confusing the two.

    5. PAPeter Attia2:18:34

       That's incredible just given the size difference between these particles normally. Like the ApoB and the ApoA1 particles-

    6. JOJim Otvos2:18:41

       Yeah

    7. PAPeter Attia2:18:41

       ... I, I thought they were like a mile apart-

    8. JOJim Otvos2:18:43

       Right

    9. PAPeter Attia2:18:43

       ... on that spectra.

    10. JOJim Otvos2:18:44

        So the good news, though, is that when you have metabolic situations that cause you to have large HDL, you also have the LDL size distribution skewed to the large LDL, so there's fewer or no small LDL particles. So [chuckles] what, what you can get away with in these extreme cases where somebody has really large, uh, HDL, and the NMR can tell when you encounter that situation, that you basically take away from the deconvolution model the smallest LDL particles, so it doesn't have the opportunity to say this large HDL is partly small LDL. [chuckles]

    11. PAPeter Attia2:19:23

        Yeah.

    12. JOJim Otvos2:19:23

        Okay? So what happened with the algorithm that was used in the, in that study is you didn't have the opportunity to have small LDL at all, even though some small LDL was probably there. And so you saw this big decrease in LDL-P, but not ApoB, because the NMR model was not allowing-

    13. PAPeter Attia2:19:52

        No, it just, it just counted-

    14. JOJim Otvos2:19:53

        So it's really an artifact of the difficulty NMR has with this s- situation.

    15. PAPeter Attia2:19:57

        So is the implication that, in this case, obicetrapib produces a disproportionate reduction in- uh, cholesterol content of particles relative to number of particles

    16. JOJim Otvos2:20:09

        I know you've talked about obicetrapib, and you and most people are very optimistic about the prospects of obicetrapib, despite CETP inhibition not panning out for many other drugs. And of course, these were all initially investigated because of the potential to create higher HDL cholesterol. And now we certainly know that HDL cholesterol is not the HDL biomarker of interest, and people were being fooled into thinking that would have benefit. Part of the reason I think, this is pure speculation, but it comes from somewhere, that even in the face of the CETP inhibitors that came closest to being efficacious, 20-30% LDL reduction, but no benefit. Maybe something bad was happening on the HDL side to counteract what was good happening on the LDL side. What was bad on the HDL side, you know, given the understanding now that just having a lot of large cholesterol-rich HDL particles doesn't put you ahead of the game in terms of cardiovascular risk. What we now know is that small HDL-P is powerfully related to mortality, all-cause mortality. So what I told you is true, especially with the most powerful CETP inhibitors. They reduce small HDL-P by 10 or 20%. If you ignore what's going on in HDL and you only look at what's happening with ApoB and LDL, you think obicetrapib is a no-brainer, it's going to be positive. But what if people are actually being hurt, maybe in terms of mortality risk, by the small HDL-P going down if there is a causal relationship there? Don't know that there is yet. We haven't, you know, proved that. There's biological plausibility because of the proteins that occupy, that hang on to, on small HDL particles, which is partly why we think it makes sense in terms of anti-oxidation, anti-inflammation, that small HDL particles might have this inverse association with mortality risk. So anyway.

    17. PAPeter Attia2:22:20

        But are you saying that you think that it's possible that, because again, we still don't have the heart outcome trial, but your thinking is that if the heart outcome trial is, it demonstrates utility, it might be, you're saying it could be just due to the reduction in small HDL-P more than the reduction in LDL.

    18. JOJim Otvos2:22:38

        I'm suggesting that people's predictions about how much efficacy there's going to be-

    19. PAPeter Attia2:22:45

        Could be wrong if they're based on the LDL reduction

    20. JOJim Otvos2:22:47

        ... could be wrong. And it may still be that the trial overall is positive or positive enough to have the drug go forward. But the most dramatic demonstration of something bad happening while something good is happening and the two counteracting each other is if the trial doesn't succeed like the other ones succeeded. So I'm just-

    21. PAPeter Attia2:23:06

        Yeah, interesting

    22. JOJim Otvos2:23:08

        ... you know, we have to wait for the trial. And then if the trial doesn't succeed, then I'll say, "Yay, I was right." But it's pure speculation.

    23. PAPeter Attia2:23:18

        Well, very interesting. And I'll have to go back and look and see what the magnitude of the ApoB reduction was. But I just remember that it was less than the-

    24. JOJim Otvos2:23:27

        It was less, but it's still-

    25. PAPeter Attia2:23:28

        But it's still reduced

    26. JOJim Otvos2:23:28

        ... still very, yeah, significant. And-

    27. PAPeter Attia2:23:32

        But you're heartened by the fact that small HDL-P was decreased.

    28. JOJim Otvos2:23:37

        Yeah. And we've actually reanalyzed that data set, you know, with the more recent... We do a better job in differentiating large HDL from small LDL. And so those LDL-P results are much more in line with the ApoB reductions than that paper indicated.

18. 2:23:53 – 2:27:10

    The future of NMR diagnostics and MVX: translating scientific potential into broader clinical use

    1. PAPeter Attia2:23:53

       Well, Jim, this is such a fascinating space, and this is a discussion that has been long overdue. Again, I don't think there's many people listening to us that haven't at least heard of, you know, LDL-P, HDL-P. They might not know what Liposcience is. They might not understand in vitro diagnostics and any of the other things that go around to it. Probably a lot of people are not familiar with MVX, but my hope is that that starts to change after this. So regardless of how you think you've fared as an entrepreneur, you've fared remarkably well as a scientist, and I think that's the most important thing because without the scientific foundation, I don't think any of the entrepreneurial stuff matters, but we do typically want them to be aligned. But put it this way, at the risk of insulting an entrepreneur, I would argue that it's easier to find a good entrepreneur than it is to find a good scientist.

    2. JOJim Otvos2:24:53

       Yeah, that's true, and that's sort of the frustration that the science is so solid, so much more solid than many startup companies, you know, are investing in. But at the end of the game, it's commercial. It's financial.

    3. PAPeter Attia2:25:10

       I suspect it's the sector, right? I suspect you wouldn't have this difficulty getting people interested if we were talking about therapeutics. I just think that the diagnostic space and the reimbursement environment in the United States is one that is not especially attractive to investors. That's my suspicion as to the issue, and that's why the MVX, I think, offers more than just a diagnostic. You know, if it could be paired to a therapy, if it has the ability to save enormous cost on the back end with respect to therapeutic selections, you know, there are a few trials that need to be done to demonstrate that, but to me, that's the interesting area.

    4. JOJim Otvos2:25:57

       No, you're right, and that's what will make it successful commercially. My vision, though, that wasn't realized is how cool would it be to go to your yearly physical and get a lipid panel that had glucose, LPIR, GlycA, MVX at no incremental cost. You know, in the rest of the world, not the U.S., where you have national health care, there is a premium put on how efficient a diagnostic is. And if you can get a lot of information for less work and money, that's worth something in the rest of the world. It's just, you know, we tried to skin that cat in the U.S. So it's, you know, I haven't lost hope, but I really left LabCorp because I didn't want to beat my head against that wall any longer and wanted to spend my remaining years doing the science. So that's what I'm continuing to do.

    5. PAPeter Attia2:26:56

       Fantastic. Well, thank you, Jim, and thanks for taking the time to come out here today.

    6. JOJim Otvos2:26:59

       You bet. [outro music]

Episode duration: 2:27:10