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

Main ideas, concepts, and lessons

1) How an “NMR cancer signal” became a lipid and risk platform

• The host (Peter Aia) introduces the guest (Jim) as a long-time figure in lipidology/NMR research, noting that many people have lipid panels (LDL/HDL tests) without realizing the test lineage traces back to the guest’s work.
• Jim describes his academic background using NMR spectroscopy as a chemistry tool.
• A turning point:
  • In the 1980s, a New England Journal paper claimed a simple NMR measurement from blood could distinguish cancer vs. no cancer by measuring signal sharpness/width.
  • Jim tried to replicate it using leftover plasma samples and found:
    • It did not reflect cancer per se (false positives occurred, e.g., after childbirth/pregnancy).
    • The “cancer diagnostic” signal was actually driven by lipids/lipoprotein particles (not cancer-related structural changes).

2) Key discovery: NMR signal shape reflects lipoprotein composition (particle size distributions)

• Jim explains that lipoproteins (VLDL, LDL, HDL) produce measurable signals at slightly different frequencies.
• The observed composite NMR peak shape in plasma results from the superposition of:
  • VLDL contributions
  • LDL contributions
  • HDL contributions
  • Plus their size subspecies (small/medium/large particles).
• Mechanism-level interpretation:
  • Cancer-associated signal narrowness was ultimately attributable to typical lipid patterns in cancer patients:
    • Higher triglycerides
    • Lower HDL cholesterol
  • Importantly: later experiments showed isolated VLDL/LDL/HDL signals didn’t inherently differ by cancer status—differences were due to lipid-level mixtures.

3) From “lipid amounts” to “lipid particles” (and why units matter)

• Standard clinical lipid panels report mass concentrations (e.g., mg/dL), but NMR-based results report molar particle concentrations.
• Jim emphasizes:
  • The NMR signal quantity relates to how many lipoprotein particles are present, not necessarily how many grams of cholesterol/triglyceride are inside them.
  • LDL particle measures (LDLP) are therefore conceptually different from LDL cholesterol (LDLC).

4) Particle size vs particle number: clinical risk implications

• Jim recounts work connecting NMR-derived subclass particle size (small vs large LDL) with cardiovascular risk.
• A central debate:
  • Epidemiology suggested small dense LDL (“pattern B”) is more atherogenic.
  • Jim and others tested whether particle size adds risk beyond particle count.
• Core conclusion from the “discordance” logic and stratified analyses:
  • When cardiovascular outcomes are examined while accounting for particle number, particle size often adds little incremental risk.
  • Put differently:
    • The apparent “small LDL is bad” effect may largely reflect more particles, not the size per se.
• Analogy:
  • Two people can have equal total cholesterol but carried in different numbers of spheres; risk correlates better with number (for their data) than with sphere size alone.

5) Discordance concept: LDLC and LDLP can disagree—and outcomes track LDLP in that disagreement

• Jim describes MESA/Framingham-type analyses where participants fall into categories:
  • Concordant: LDL cholesterol percentiles roughly match LDL particle percentiles
  • Discordant: one is high/low relative to the other
• Using Kaplan–Meier curves for cumulative events:
  • When LDLP > LDLC (more particles than cholesterol suggests), event risk is higher.
  • When LDLP < LDLC, risk is lower.
• Practical significance:
  • Clinically, this supports using particle-based metrics when cholesterol metrics are misleading.

6) But: adding LDLP to standard risk models often didn’t improve prediction

• Jim explains a limitation of translating biomarkers into incremental clinical usefulness:
  • Standard cardiovascular risk equations already include HDL and total cholesterol (and diabetes, smoking, hypertension).
  • Because of correlations (e.g., low HDL often accompanies higher particle burden), LDLP may not improve model performance after those variables are included.
• Translation pivot:
  • Liposcience’s original goal (LDLP/LDLC for risk assessment) shifted toward risk management:
    • If you treat to lower LDL-associated risk, particle metrics (LDLP/APOB) can identify residual risk even when LDL cholesterol looks “low enough.”

7) Lipoprotein measurements extended to insulin resistance/diabetes prediction (LPIR)

• Jim introduces LPIR (a 0–100 composite score) as an early composite score derived from NMR-measured lipoprotein subclass distributions.
• Purpose:
  • Move beyond cardiovascular risk into insulin resistance and future type 2 diabetes risk.
• Diabetes logic framework:
  • Diabetes is a time-integrated progression driven by insulin resistance.
  • Glucose becomes abnormal only later (after beta-cell dysfunction begins).
  • Therefore:
    • Primary prevention should target insulin resistance earlier, not wait for elevated glucose/pre-diabetes to become overt diabetes.
• Validation claim:
  • LPIR has been compared to fasting insulin and shows better performance.
  • It has been shown to track with and respond to interventions (lifestyle, metformin) in major prevention research, including mention of DPP baseline and post-treatment comparisons.

8) MVX (Metabolic Vulnerability Index): a multi-parameter mortality risk score

• Jim presents MVX as a composite score (0–100; higher = worse) built from NMR-derived components.
• MVX’s predictive scope (as described):
  • Mortality risk (cardiovascular death, liver disease, all-cause mortality)
  • Demonstrated in both high-risk (catheterized patients) and healthier populations.
• How MVX was developed (top-down / data mining approach):
  • Rather than inventing biomarkers from a mechanistic hypothesis first, the team mined stored baseline NMR spectra from large cohort studies.
  • They extracted new signals and tested them against future outcomes.
• MVX components (as stated in the subtitles):
  • Small HDL particles (small HDLP)
  • Glyc (glycan-related inflammatory marker from the NMR spectrum; linked to systemic inflammation)
  • Citrate and three BCAAs:
    • leucine, isoleucine, valine
  • These were assembled into:
    • An inflammation subscore IVX
    • A metabolic/malnutrition subscore MMX
• Central claim about what MVX reflects:
  • MVX appears to predict susceptibility to dying (metabolic frailty/vulnerability) more than it predicts specific diseases being diagnosed first.
  • It relates to “wasting/malnutrition inflammation syndrome” patterns seen across conditions that increase mortality.

9) Glyc and inflammation vs CRP (why glyc might be useful)

• Glyc is described as:
  • Less volatile than CRP (CRP changes rapidly day-to-day)
  • More reflective of chronic or steady-state systemic inflammation
• Clinical nuance:
  • Acute infections can elevate glyc, but the marker is designed to reflect systemic inflammatory background rather than short-term spikes.

10) Why commercial translation stalled for earlier NMR-based metrics

• Jim explains barriers to broad adoption:
  • NMR testing required a specialized instrument and regulatory/validation effort (FDA clearance).
  • The business model shifted:
    • Liposcience aimed to be an IVD manufacturer (sell/enable NMR machines to clinical labs).
    • LabCorp acquisition changed incentives to a more proprietary, lab-sendout/analysis model.
  • Major commercial limitation:
    • Even if information is “cheap/free” analytically, payers typically resist reimbursement for incremental complexity unless it is tied to guideline changes or specific billing structures.

11) Instrumentation and cost-efficiency argument

• The Vanta NMR analyzer was highlighted:
  • FDA cleared (LDLP and Vanta analyzer in 2011 per the subtitles)
• Key efficiency claim:
  • No consumables are needed per test, so costs scale well with volume.
  • One plasma sample yields many results (lipid profile + LPIR + gly signals + MVX-related components).

12) Edge-case discussion: CETP inhibitors and potential NMR artifact risk

• Jim discusses how some drugs (CETP inhibitors) can create unusual HDL particle sizes/compositions, which may challenge older NMR deconvolution algorithms.
• Proposed explanation:
  • If HDL particles become closer in size to LDL particles, deconvolution can confuse signals.
  • This can make NMR appear to reduce LDLP differently from APOB.

13) MVX practical proposal: a low-cost “panel” at routine checkups

• The guest argues for a future where a standard blood draw can produce:
  • Lipid panel + glucose + LPIR + glyc + MVX
• Emphasis:
  • Analytical measurements come from the same specimen; the “incremental cost” is claimed to be minimal in high-volume use.

---

Methodologies / procedures / approaches

A) Using NMR spectra to quantify lipoprotein particle distributions

• Obtain blood plasma sample.
• Run a low-tech/simple NMR spectrum quickly (described as ~30 seconds).
• Detect a composite signal arising from terminal methyl groups on fatty acid chains present in lipoprotein particles.
• Interpret:
  • X-axis = signal frequency
  • Y-axis = signal amplitude/intensity
• Use a deconvolution model that includes reference templates for:
  • Different sizes/subspecies of VLDL, LDL, HDL particles
• Compute:
  • The contributions of each component such that:
    • sum of deduced parts ≈ measured composite
• Output:
  • Particle concentration measures (reported in molar concentration units such as nmol/L for LDLP)

B) Constructing composite risk scores (LPIR, IVX, MMX, MVX)

• Start from NMR-derived signals that map to biological categories (examples given):
  • Lipoprotein size/class distributions (LPIR; small/large subclass patterns)
  • Small HDL particle measures (part of IVX/MVX)
  • Specific NMR signals interpreted as:
    • glyc = inflammatory glycan decoration signal
    • citrate and BCAAs = metabolic components
• Create subscores when useful:
  • IVX (inflammation vulnerability index) from inflammation-related components
  • MMX (metabolic malnutrition index) from metabolic/malnutrition-related components
• Combine subscores into MVX (0–100 scale; higher = worse mortality/metabolic vulnerability).
• Use longitudinal cohort baseline NMR spectra:
  • Extract baseline biomarker values
  • Link them to future outcomes (mortality and disease endpoints)
  • Validate predictive gradients (e.g., hazard ratios across quartiles/top vs bottom score ranges).

C) Discordance framework (LDLP vs LDLC)

• Convert both LDL cholesterol and LDL particle measurements into percentile ranks.
• Define “concordance” vs “discordance” using percentile differences (example described):
  • Within ±12 percentile units = “concordant”
  • Others categorized as discordant (one high relative to the other)
• Use cumulative incidence plots (Kaplan–Meier):
  • Compare event curves across the concordant/discordant groups.
• Interpret:
  • Risk tracks the metric that is “high” relative to the other (as described: often the particle metric).

D) Diabetes prevention logic using insulin resistance scores

• Identify insulin resistance before glucose crosses diagnostic thresholds.
• Use NMR-derived insulin resistance composites (e.g., LPIR) as surrogate measures for “causal drivers” rather than late-stage effects.
• Validate using:
  • Intervention studies (lifestyle, metformin)
  • Changes in the biomarker and whether those changes predict reduced diabetes incidence.

---

Speakers / sources featured (as mentioned in subtitles)

Speakers

• Peter Aia (Host, Drive Podcast)
• Jim (Guest; creator/developer of NMR lipid particle testing and related scores; name not explicitly provided in the subtitles)

Video/Project context / sources referenced

• New England Journal of Medicine / New England Journal paper (1986) (claimed NMR cancer diagnostic)
• New England Journal of Medicine / cancer NMR paper authors (referenced indirectly; not named)
• Ron Krauss (UC Berkeley / Donner Laboratory) (gradient gel electrophoresis LDL size work)
• NIH (National Institutes of Health) (grants and study framework referenced)
• Seaman’s Medical Systems (funding mentioned)
• MESA = Multi-Ethnic Study of Atherosclerosis
• Framingham / Framingham Offspring Study
• Women’s Health Study
• Diabetes Prevention Program (DPP)
• JUPITER (mentioned as part of risk/evidence context)
• Cath lab cohort at Duke University (Cardiovascular outcomes cohort; unnamed by institution but described)
• Salt Lake City cohort replication (UT cohort referenced)
• EpiS study (older people cohort with many variables and NMR data)
• CETP inhibitors / obetropib (drug referenced)
• Kinect inhibitor “CPT inhibitors” (subtitles appear to mean CETP inhibitors)
• PCSK9 inhibitors, statins, niacin, “CTP inhibitors”/HDL drugs (therapies referenced)
• Nafld/NASH/MASLD (liver disease paper context; “formerly known as NAFD” mentioned—subtitles contain obvious auto-text errors)
• Kaplan–Meier curves (statistical method referenced)

(No other distinct named individuals besides Ron Krauss are clearly specified in the subtitles.)