NMR Blood Analysis: What One Blood Sample Can Reveal About Mortality Risk
Peter Attia sits down with the scientist behind NMR lipoprotein testing to trace how a single blood scan came to reveal far more than cholesterol. The conversation moves from LDL particle number and insulin resistance to the MVX score, a striking marker of mortality risk. It is a tour of how much information one small plasma sample can hold.
Overview
This wide-ranging conversation traces the origin of NMR lipoprotein testing, which began not with a lipid question but with a 1986 claim that NMR could detect cancer. Following that thread, Jim discovered the real signal came from lipoprotein particles, and built a way to read VLDL, LDL and HDL from a single 30-second scan. He explains why results are reported as particle concentrations in nmol/L rather than cholesterol mass, and why particle number often tells a clearer story than cholesterol alone.
The discussion explores small dense LDL, showing that its extra risk reflects a greater number of particles rather than their size, and how discordance analyses made that testable. It turns to the LPIR score, a lipid signature of insulin resistance that can flag metabolic strain before glucose rises. Jim recounts the story of LipoScience, the one-of-a-kind Vanta analyzer, and why broad clinical translation stalled after the LabCorp acquisition.
The centrepiece is the MVX score, which combines small HDL particles, an inflammation signal called glycA, citrate and branched-chain amino acids into a striking predictor of mortality. Most surprising of all, MVX is largely independent of age and predicts even in healthy 30-year-olds, suggesting it reflects a kind of metabolic frailty rather than the onset of any single disease. Throughout, the theme is how much one small plasma sample can reveal when the same scan is mined again and again.
Key quotes
5These were three people who had just given birth. So pregnancy was one false positive given in that New England Journal paper.
You measure the whole and then you decompose it into the parts, so the sum of the parts equals the whole.
You are reporting the concentration of the package, not the lipid molecules in the package.
You always eradicate causal drivers of disease the moment they appear.
Dying sooner versus later is what MVX seems to influence, as opposed to getting the diseases that cause this.
Key ideas
8A discovery that began by accident
Jim did not set out to build a lipid test. A 1986 claim that NMR could detect cancer led him to study blood plasma, where he found the real signal came from lipoprotein particles.
Small dense LDL: correlation, then cause
Early studies linked small, dense LDL to higher cardiovascular risk. Closer analysis showed the extra risk tracked with the greater number of particles, not their smaller size.
Turning one signal into many numbers
A 30-second NMR spectrum is decomposed by a deconvolution model into the contributions of every lipoprotein subclass. The whole is measured, then separated into its parts.
Counting packages, not their contents
NMR reports lipoprotein particle concentration in nmol/L. It counts the containers rather than measuring the cholesterol carried inside them.
When particle number and cholesterol disagree
In discordance, risk follows the particle number rather than the cholesterol level. Discordance becomes more common as metabolic syndrome features accumulate.
A lipid signature for insulin resistance
The LPIR score blends six lipoprotein-size and subclass measures into a 0-100 scale. It reflects insulin resistance and the likelihood of moving toward diabetes.
MVX: a window on mortality risk
The metabolic vulnerability index combines small HDL particles, the glycA inflammation signal, citrate and three branched-chain amino acids. It has shown a strong relationship with mortality across many populations.
Metabolic frailty, not disease onset
MVX appears to relate to dying sooner rather than to developing any one disease. Strikingly, it is largely unassociated with age and predicts even in healthy 30-year-olds.
Practical takeaways
6- 1
Look beyond cholesterol 1:03:00
Particle-based measures such as LDL particle number or ApoB can add clarity that cholesterol alone may miss.
- 2
Discordance flags metabolic strain 1:24:00
A gap between particle number and cholesterol is more likely when metabolic syndrome features are present, a cue to look at the wider metabolic picture.
- 3
Insulin resistance shows up early 1:40:00
Lipoprotein patterns can reflect insulin resistance well before fasting glucose rises, offering a much earlier window.
- 4
Address causes, do not wait for disease 1:58:00
The smoking-and-lung-cancer analogy: causal drivers are best acted on as soon as they appear, not once disease is measurable.
- 5
One draw, many insights 2:16:00
A single 150-microlitre plasma sample can yield lipids, glucose, LPIR, the glycA inflammation marker and MVX from the same scan.
- 6
Think in terms of resilience 2:22:00
MVX reflects metabolic resilience, susceptibility to dying from whatever comes, rather than the presence of a specific disease today.
Topics & chapters
14Introduction and Jim's NMR background
Peter Attia welcomes the guest whose work underpins LDL-particle testing, and Jim recounts two decades using NMR as a structural tool in academia.
The 1986 cancer-signal claim
A New England Journal paper claimed NMR could detect cancer. Testing leftover plasma, Jim found narrow signals in women who had just given birth, a clue the signal was about lipids, not cancer.
Plasma signals reveal lipoproteins
With Siemens funding, Jim separated VLDL, LDL and HDL and showed their overlapping signals shaped the composite plasma spectrum. A 1990-91 paper established the approach.
Small dense LDL and commercialization
Working with Ron Krauss, NMR could distinguish large from small LDL. The clinical promise of small dense LDL pushed Jim toward commercializing the technology.
How a lipid panel is measured chemically
Jim walks through reagent-based assays for triglycerides and cholesterol, the separation steps, and the Friedewald estimate that still underlies many LDL-cholesterol numbers.
Reading an NMR spectrum
Frequency on the x-axis, amplitude proportional to how many protons are present. A low-tech 30-second scan is decoded by a deconvolution model into its parts.
Particle count vs cholesterol mass
Results appear in nmol/L because NMR counts lipoprotein particles rather than the cholesterol inside them: the concentration of the package, not its contents.
Size vs number: the discordance debate
Using simple thought experiments, Jim explains why the extra risk of small dense LDL reflects higher particle number, and how discordance analyses settle the question.
Discordance, metabolic syndrome and risk management
MESA Kaplan-Meier curves show risk tracks particle number. Because standard risk equations already include HDL cholesterol, LDLP shines most for managing, not just assessing, risk.
The LPIR score and insulin resistance
Jim's first composite score blends six lipoprotein measures into a 0-100 insulin-resistance signal, building on the triglyceride-to-HDL ratio championed by Gerald Reaven.
Diabetes as a causal, time-integrated process
Insulin resistance is the cause; rising glucose is the late, easily measured effect. Attia draws the smoking-and-lung-cancer parallel for acting on causes early.
LipoScience, the Vanta analyzer and translation hurdles
Jim recounts building the only NMR clinical analyzer, FDA clearance in 2011, the 2014 LabCorp acquisition, and why broad clinical translation stalled.
The MVX metabolic vulnerability index
Combining small HDL particles, the glycA inflammation signal, citrate and branched-chain amino acids, MVX predicted mortality strongly in the CATHGEN cohort and beyond.
Metabolic frailty, CETP edge cases and closing
MVX predicts even in healthy 30-year-olds and is largely age-independent. Jim closes on CETP-inhibitor analytics, obicetrapib and the vision of near-free comprehensive testing.