#402 ‒ NMR blood analysis: how heart disease risk, insulin resistance, inflammation, and mortality risk can be assessed from a single blood sample | Jim Otvos, Ph.D.

Full analysis

Jim Otvos, the biochemist who invented LDL particle number (LDL-P) testing, walks Peter Attia through what standard cholesterol tests miss, why the "large fluffy LDL is harmless" idea is wrong, and how the same NMR technology now produces markers for insulin resistance, systemic inflammation, and a new composite mortality-risk score called MVX. The episode is dense with actionable science but carries an important caveat: most of these tests are available only through LabCorp, and the analyzers that run them are aging with no clear replacement plan.

What was covered

• How standard lipid panels actually work — LDL cholesterol is not directly measured in most labs; it is calculated using the Friedewald equation (total cholesterol minus HDL minus triglycerides ÷ 5). Jim Otvos helped develop a newer NIH equation that LabCorp uses, but most labs still use the older, less accurate formula.
• Why LDL particle number (LDL-P) beats LDL cholesterol for risk management — A 2007 Framingham study and a 2011 MESA study showed that when LDL-P and LDL cholesterol disagree ("discordance"), cardiovascular risk tracks with particle number, not cholesterol mass. People with metabolic syndrome are especially likely to be discordant.
• The "large fluffy LDL is safe" myth debunked — Otvos argues this is flatly wrong. People with familial hypercholesterolemia (FH) have large LDL particles and die in their 30s when untreated. What looks like extra danger from small particles is actually explained by higher particle count, not particle size itself. Once you control for LDL-P, particle size adds nothing to risk prediction.
• LP-IR score: detecting insulin resistance before glucose rises — A six-variable NMR score (combining VLDL, LDL, and HDL subclass sizes and concentrations) predicts transition to type 2 diabetes before fasting glucose becomes elevated. In the Diabetes Prevention Program it tracked closely with response to lifestyle intervention and metformin. A related score, the Diabetes Risk Index (DRI), adds branched-chain amino acids and is commercially available through LabCorp.
• GlycA: an NMR inflammation marker more stable than hs-CRP — GlycA measures the glycan coating on acute-phase proteins and is correlated with IL-6 and other interleukins. Unlike hs-CRP, which can spike a thousand-fold with acute infection, GlycA is biologically stable — guidelines recommend averaging two or three hs-CRP readings, but nobody does; GlycA doesn't require that workaround.
• MVX (Metabolic Vulnerability Index): a mortality-risk score from the same blood draw — Composed of small HDL particle count, GlycA, citrate, and three branched-chain amino acids (leucine, isoleucine, valine), MVX ranges 0–100. In a Duke cardiac catheterization cohort of ~7,000 patients with mean age ~60, the highest-MVX quartile had roughly two-to-threefold greater five-year mortality than the lowest. Crucially, in healthy 30-year-olds followed for 35 years, the distribution of MVX scores was nearly identical to that of 60-year-olds, and the score predicted premature death across the full follow-up.
• The technology's uncertain future — The Vantera NMR analyzers that run all these tests (the only ones in existence) are first-generation machines now roughly 15 years old. No second-generation device is in development. Otvos says LabCorp is open to licensing the patents, but no IVD company has yet stepped in to build a successor.

Notable claims & predictions

• Jim Otvos: "The notion that large, fluffy LDL particles are benign is completely fallacious." People with familial hypercholesterolemia have large LDL particles and die at 30–35 when homozygous and untreated — particle size confers no protection once particle number is accounted for.
• Jim Otvos: When LDL-P and LDL cholesterol disagree, "cardiovascular risk tracks with the particle number, not the cholesterol" — shown in both Framingham and MESA data. At high cardiovascular risk, patients who hit low LDL cholesterol targets often have not reached equivalently low LDL-P, making them candidates for additional therapy (e.g., a PCSK9 inhibitor).
• Jim Otvos: GlycA "is a very stable and useful clinical marker of systemic inflammation" that adds predictive power above hs-CRP, largely because it doesn't swing wildly with transient illness — relevant for anyone trying to track chronic low-grade inflammation longitudinally.
• Jim Otvos: In healthy 30-year-olds, MVX scores were distributed identically to those in 60-year-olds, and predicted premature mortality over 35 years of follow-up — a finding Otvos calls more about susceptibility to dying than about acquiring specific diseases: "MVX seems to influence dying sooner versus later, not getting the diseases that cause the death."
• Jim Otvos on obicetrapib (CETP inhibitor in phase 3 trials): The drug reduces small HDL particles by 10–20%. Because small HDL-P is among the strongest predictors of all-cause mortality in the MVX framework, he raises the possibility that benefit on LDL could be offset by harm via small HDL reduction — though he calls this "pure speculation" pending hard outcome data.
• Jim Otvos on the Vantera analyzers: "At some point in the future, maybe sooner rather than later, these first-generation analyzers are going to cease to function" — putting the entire NMR testing infrastructure at risk unless an IVD company acquires the technology.

Fact check

"LDL cholesterol is estimated by dividing triglycerides by five (Friedewald equation), and most labs still use this." Accurate and verifiable. The Friedewald equation is standard in most clinical labs. The caveat Otvos names — that it becomes increasingly inaccurate at high triglycerides or very low LDL levels — is well established. The NIH-derived extended equation (Martin-Hopkins equation) exists and LabCorp uses it, but adoption is uneven. No issue here.

"People with familial hypercholesterolemia have large LDL particles but die at 30–35 when homozygous." Accurate for homozygous FH, one of the most lethal genetic lipid disorders. The claim supports his larger point that large LDL particles are not inherently safe. No issue.

"Small HDL-P was the most powerful predictor of two-year mortality in the EPIS study, beating 186 other variables." This is a specific, strong claim. It is unverified here — the transcript references an unpublished or obscurely published analysis and provides no citation the listener can check. The finding is internally consistent with other data Otvos describes, but readers should treat it as a preliminary result, not established clinical fact, until the paper is publicly available and peer-reviewed.

"MVX in healthy 30-year-olds predicted mortality over 35 years of follow-up, and this paper is about to be submitted." The core finding is intriguing and plausible given the published 2023 cardiac cath study and the published MESA paper Otvos cites, but the 30-year-old cohort data is explicitly unpublished at recording time. This is worth noting: the claim is based on work that has not yet cleared peer review.

"The Diabetes Prevention Program data show LP-IR change predicted incident diabetes — but hasn't been published because the primary author died." Unverified and unpublished. The DPP itself is a well-documented, landmark study, but Otvos's NMR sub-analysis remains in manuscript form. Treat as preliminary.

On obicetrapib and small HDL-P: Otvos explicitly labels his concern "pure speculation," which is the correct framing. Readers should not conclude the drug is harmful — the concern is a hypothesis awaiting the outcome trial.

Incentive note: Otvos co-founded LipoScience, holds patents related to NMR diagnostics, and would benefit professionally and reputationally from broader adoption of these tests. His advocacy for NMR over standard lipid panels, and his framing of LabCorp's acquisition as commercially harmful to patients, should be weighed against that context. The