🧬 Aging Wiki

Human whole-blood NAD+ levels do not vary with age or lifestyle interventions

14 min read

Trętowicz et al. 2026 — Human whole-blood NAD+ levels do not vary with age or lifestyle interventions

A 7-cohort, 303-participant cross-sectional + interventional study from the Houtkooper laboratory (Amsterdam UMC) using a rigorously pre-validated UHPLC-HRMS NAD+ assay to test whether whole-blood NAD+ declines with age in humans. All six independent age comparisons across cohorts were null (P-values 0.24–0.62; R² 0.012–0.051), as were intervention comparisons against exercise and protein supplementation in older adults. The same assay readily detected a positive response to nicotinamide riboside (NR) supplementation, demonstrating that the null age-effect is not a sensitivity artifact. The authors trace much of the prior conflicting blood-NAD+/age literature to pre-analytical handling artifacts — particularly frozen-then-thawed handling that destroys ~30–80% of NAD+ depending on conditions — and conclude that whole-blood NAD+ is not a useful biomarker of human aging or lifestyle.

Citation

Trętowicz MM, Scantlebery AML, Schomakers BV, et al. Human whole-blood NAD+ levels do not vary with age or lifestyle interventions. Nature Metabolism (2026). DOI: 10.1038/s42255-026-01537-5. Received 23 Oct 2025; accepted 24 Apr 2026; published online 14 May 2026.

Senior author: Riekelt H. Houtkooper, Laboratory Genetic Metabolic Diseases, Amsterdam UMC. Co-corresponding: Georges E. Janssens (same institution; was first author on the 2022 Nat Aging paper 1 that originally argued for an NAD+-aging link in human muscle, making this work a partial self-correction at the blood-tissue level).

Funding: EU Horizon Europe NADIS project (101073251); Academy of Finland (286359, 314455, 335445, Profi6 336449, 335443, 314383, 272376, 266286); Dutch NWO and SIA grants; Spanish PID2022-142470OB-I00 + PROMETEO + PID2023-147560OA-I00; Netherlands Genomics Initiative; Nestlé Health Science (MEJNES2019 cohort only).

Competing interests: The authors declare no competing interests.

Peer reviewers (per published peer-review information): Anthony Covarrubias, Marie Migaud, Lindsay Wu — all serious NAD+ field researchers; Lindsay Wu in particular has historically been a methodological skeptic of the precursor-supplementation field.

Cohorts (n=303 total across 7 independent studies)

CohortnAge rangeCountryTrial registryDesign
Aging cohort4020 younger <30 y + 20 older >60 yNetherlands (Amsterdam UMC LAKC residual samples)n/a (residual clinical samples)Cross-sectional, age-stratified
Twin-pair NR supplementation2433–41 yFinland (Helsinki + Oulu)NCT03951285Randomized within twin pairs; 5-month NR; included primarily as positive-control sensitivity check
CardioHT2628–73 yNetherlandsNCT06319417Pre-treatment samples only used here (post-intervention excluded)
ELITE47younger controls <40 y (n=12); older controls >40 y (n=12); athletes 19–31 y (n=23)Netherlands (Amsterdam UMC Movement Sciences)NL71682.018.19 (METc) / NL9328 (Dutch Trial Register)Cross-sectional age + athletic-status comparison
LLS (Leiden Longevity Study follow-up)7063–87 yNetherlands (Leiden UMC)NL-OMON52307, P21.055 (2021)Older-adult cohort within long-lived families
TEAMS65>65 y (31 exercise + 34 exercise+protein, per Results-text; Extended Data Table 1 lists 35 — minor discrepancy noted but Results-text value used)Netherlands (Amsterdam UAS)OMON NL8785 / NCT registered as TEAMS RCTRandomized intervention; pre/post paired samples
MEJNES201931>65 y, frail (control n=7 / supplement-only n=9 / supplement+exercise n=15)Spain (Valencia + Murcia, Universidad Católica de Murcia)NCT06975540 (METc H150037557317)Single-blind randomized; frail community-dwelling older adults

The aging cohort is the head-on age comparison; the other cohorts deliver independent replication across diverse populations (frail, athletic, post-cardiac-event, long-lived-family-enriched) and across geographies (Netherlands, Spain, Finland) 2.

Methods (verified scope: pre-analytical + assay design)

Analytical platform. Ultra-high-performance liquid chromatography coupled to high-resolution mass spectrometry (UHPLC-HRMS) on a Waters Acquity UHPLC system coupled to a Bruker Impact II Ultra-High-Resolution Qq-Time-Of-Flight mass spectrometer (with a parallel Waters Atlantis Premier BEH Z-HILIC column run when the primary Merck Millipore SeQuant ZIC-cHILIC column was unavailable; both produced equivalent results). The assay was specifically designed to handle real-world analytical variability that the authors argue has confounded prior blood-NAD+/aging studies. Bruker TASQ 2024.0.1 (build 11348) for instrument data acquisition; R 4.4.3 with ggplot2 4.0.1, dplyr 1.1.4, tidyr 1.3.1, ggpubr 0.6.1, gridExtra 2.3, ggrepel 0.9.6, mixOmics 6.30.0, and ggforce 0.5.0 for analysis. Sample measurement order was randomized to minimize analytical or batch-related effects. Sex was included as a covariate via linear models in all cohort metabolomics analyses 2.

Method performance (Methods § Validation of quantitative NAD+ measurement, full PDF). Using a stable-isotope-labelled internal standard (¹³C₅-NAD+; Chiralix) spiked into 10 µL of whole blood:

  • Limit of detection (LOD): 0.21 nmol mL⁻¹
  • Limit of quantification (LOQ): 0.69 nmol mL⁻¹
  • Intra-assay coefficient of variation: CV 12.5% (10 replicates from a pooled control)
  • Inter-assay coefficient of variation: CV 16.1% (10 independent runs over 10 days)
  • Linearity confirmed across 10–100 µL whole-blood input (Fig. 1c, R² = 0.9996)
  • Spike-recovery 108–112% across 500/1000/2000 pmol NAD+ additions to 10 µL blood (Fig. 1d)

These figures together establish the assay is sufficiently sensitive that a true age-related blood NAD+ difference of even ~3 nmol/mL would be detectable above its noise floor — the null age result is not a sensitivity artifact.

Pre-analytical handling validation (Fig. 1 + Extended Data Fig. 1). From source data MOESM2:

  • Storage temperature, 4 days: Fresh whole blood ~19.5 nmol/mL → −20°C drops to ~9.3 nmol/mL (~52% loss; P=0.0085) 3. Per the full Results text and Extended Data Fig. 1b: −80°C also degrades NAD+ (fresh ~40 nmol/mL → −80°C ~31 nmol/mL, P=0.0015; −20°C ~24 nmol/mL, P=0.023 vs fresh; −80°C-vs-−20°C P=0.38) — i.e., both freezer temperatures lose NAD+ relative to fresh, with −20°C losing slightly more than −80°C but not significantly differing from it. No freezer condition without methanol pre-preservation maintains NAD+ fidelity.
  • Freeze–thaw cycles (representative donors from MOESM2 source-data sheet “Fig. 1f”; full n=6 donors in Extended Data Fig. 1c): donor “Female_1985_A” fresh 41.5 → cycle 1: 30.1 → cycle 2: 23.6 → cycle 3: 22.6 nmol/mL (45% cumulative loss); donor “Male_1986_C” fresh 33.2 → 19.3 → 16.8 → 11.4 (66% cumulative loss) 4. Per-donor freeze-thaw behavior is highly variable — Fig. 1f shows the two main-figure representative individuals with cumulative losses of just −4% and −32% respectively, with mid-cycle “rebounds” in one donor. Across the broader Extended Data Fig. 1c panel (4 additional donors), losses range from ~14% to ~69% over 3 cycles. The mechanistic implication: freeze-thaw NAD+ loss is not deterministic per cycle — it depends on cell-membrane integrity, intracellular vs extracellular NADase exposure, and ice-crystal-formation rate during cycle transitions. For analytical reliability, no freeze-thaw cycles should be permitted; methanol pre-freeze preservation is the recommended protocol.
  • Frozen-then-extracted vs methanol-pre-frozen: Across 15 paired technical replicates, frozen-and-then-extracted samples averaged ~10–15 nmol/mL lower than fresh; methanol-preserved samples retained NAD+ levels within 1 SD (~3.5 nmol/mL) of fresh values 5.
  • Plasma NAD+ (Extended Data Fig. 1a, n=10 donors, 5M/5F): mostly at or below detection limit (0–0.58 nmol/mL); per the full Results text, plasma NAD+ is 50–100× lower than whole-blood NAD+ and falls below the LOQ (0.69 nmol/mL) for most samples. NAD+ is overwhelmingly intracellular; plasma NAD+ is not a viable analyte for human aging studies 6.

These pre-analytical findings — that prior literature reporting age-related NAD+ decline used frozen-without-methanol-preservation samples that lost >50% of their NAD+ — are central to the authors’ explanation for the conflicting prior literature.

Sample type. Whole-blood NAD+ throughout (not plasma; not isolated PBMCs).

Linearity and spike-recovery validation (Fig. 1c–d, source data MOESM2): NAD+ signal linear from 10–100 µL blood input (10 µL ~29, 100 µL ~330 nmol/mL — slope = 3.3); spike-recovery preserved across 500–2000 pmol NAD+ additions to 10 µL blood, recovery rate ≈ 1:1 within technical-replicate variability 7.

Mouse method validation (Extended Data Fig. 2). C57BL/6J females (24–25 weeks) + C57BL/6N males (11 weeks); 100 mg/kg NMN IP, 1 h post measurement: NAD+ rose from ~7 nmol/mL (T0) to ~15 nmol/mL (T1) across 4 paired animals (~2-fold increase, P<0.05) 8. Confirms the assay detects acute pharmacological NAD+ elevation as expected.

Headline finding: whole-blood NAD+ does not vary with age (Fig. 2 / source data MOESM3)

Each of the six independent age-NAD+ comparisons was null at conventional significance thresholds. Effect sizes (R² where regression-based; or P-values where group-comparison-based) are author-reported in the source-data sheets and reproduced here:

CohortComparisonStatisticValueInterpretation
Aging cohortOlder vs Younger (n=40)P (Older_vs_Younger)0.242Numeric trend toward lower NAD+ in older females (mean ~18 vs ~27 nmol/mL) but not significant; opposite or absent in males
CardioHT (n=26)Age vs whole-blood NAD+ (regression)R² = 0.012NSAge explains <2% of NAD+ variance
ELITE (n=47)Older controls vs AthletesP (Old_vs_Athlete)0.498No NAD+ difference between athletes (19–31 y) and >40 y controls
LLS (n=70)Age (63–87 y) vs NAD+ (regression)R² = 0.051NSWithin older adults of long-lived families, age explains ~5% of NAD+ variance
TEAMS (n=65)Exercise+Protein after vs beforeP0.61912-week exercise+protein intervention did not change whole-blood NAD+
MEJNES2019 (n=31, frail older adults)Exercise group (GE) before/afterP0.621Multimodal intervention did not change whole-blood NAD+ in frail older adults

(All P-values reported as the author’s own computed statistic from the bottom row of each source-data sheet; comparisons are two-sided.)

Reproducibility comment (verbatim from Reporting Summary): “the major finding — namely, the association between NAD+ and age — was consistently replicated across all cohorts. Despite these differences [in setup, participant characteristics, group structures], the major finding… was consistently replicated across all cohorts” 2. In context this means the null age-NAD+ association is what replicated, not a positive one.

Sensitivity check: NR supplementation does elevate blood NAD+ (Extended Data Fig. 3)

The twin-pair NR supplementation cohort (n=24, NCT03951285) was included specifically as a positive-control assay-sensitivity test. Per the full Results text (visible in PDF page 3) and Extended Data Fig. 3:

  • Dose: NR escalated from 250 mg/day → 1 g/day over 5 months
  • Population: 24 individuals from Finnish BMI-discordant monozygotic twin pairs (FinnTwin12 / FinnTwin16; ages 33–41 y; n=12 F, n=12 M); recruited from a parent twin study of n=~5,000 with 1,200 monozygotic twins
  • Whole-blood NAD+ before: ~32 nmol/mL (Before group median, Extended Data Fig. 3)
  • Whole-blood NAD+ after 5 months: ~62 nmol/mL (After group median; ~2-fold elevation)
  • P-value: 1.59 × 10⁻⁹ (linear-regression-adjusted-for-sex, paired pre/post)

The contrast with the null age + lifestyle results (P-values 0.24–0.62) of magnitude ~10⁹ establishes that the assay is sensitive enough to detect pharmacological NAD+ elevation; it is the age and lifestyle effects that are absent, not the assay’s ability to see them. Individual-level data are GDPR-restricted; group-level numbers visible in the public Extended Data Fig. 3.

NADP+ shows a different pattern (Extended Data Figs. 5–6, source data MOESM7–MOESM8)

While whole-blood NAD+ is flat with age, the related cofactor NADP+ shows weak but detectable age-related changes in some cohorts:

  • Aging cohort NADP+ Older vs Younger: P = 0.031 (significant decline with age) 9.
  • CardioHT NADP+ regression: R² = 0.237 (age explains ~24% of variance) 10.
  • ELITE NADP+ Old vs Athlete: P = 0.003 (athletes higher) 11.
  • LLS NADP+ regression (within 63–87 y): R² = 0.028 (NS within older-only range) 12.

The NAD+/NADP+ ratio and NAD+/(NAD+ + NADH + NADP+ + NADPH) summary ratios were also examined, with mixed results — most age comparisons NS, exercise and protein interventions in TEAMS showed shifts in some ratios. (The main paper’s Discussion likely contextualizes these — this section is paywalled and not read by the verifier yet.)

This nuance matters for the wiki: the negative result is specific to NAD+; related pyridine-nucleotide cofactors are not necessarily age-invariant. Future verifier pass should resolve whether the authors interpret the NADP+ signal as biological or as an artifact of NAD+→NADP+ enzymatic interconversion during sample handling.

Statistical power (Extended Data Fig. 4, source data MOESM6)

The authors provide explicit power analyses incorporating the empirically-measured biological + analytical variability of the assay. From source data MOESM6:

  • To detect a Δ = 1 nmol/mL shift with 80% power: N ≈ 347 per group with biological variability alone; N ≈ 786 per group once analytical variability is added 13.
  • To detect Δ = 2 nmol/mL: N ≈ 87 (bio only) → N ≈ 197 (bio + analytical) per group.
  • To detect Δ = 5 nmol/mL: N ≈ 14–31 per group depending on analytical noise assumptions.

This is consequential: most prior individual-cohort blood-NAD+/aging studies (n ≈ 20–100 per group) had power to detect only very large (>3 nmol/mL) effects. The current paper’s 7-cohort design (303 total) is the largest pooled assessment to date.

What this study does not address

  • Tissue NAD+ is NOT measured (muscle, liver, brain) — the authors restrict their claim to whole-blood NAD+. Prior reports of NAD+ decline in human muscle (e.g., Janssens 2022 1) are not directly tested and could remain valid at the tissue level even if blood is invariant.
  • PBMC-isolated NAD+ is not the analyte; whole blood includes erythrocytes (the dominant volume fraction) which have their own NAD+ pool tied to glycolysis and methemoglobin reduction. The paper’s claim is about the integrated whole-blood pool, not cell-type-specific NAD+.
  • Sub-clinical aging conditions that might selectively deplete NAD+ (e.g., CHIP, sarcopenia, frailty trajectory) are not stratified — though MEJNES2019 frail older adults are included and still show no NAD+ deficit at baseline.
  • NAD+ flux / turnover (vs static pool size) is not measured. If NAD+ consumption (e.g., via CD38, PARP1) rises with age but salvage compensates, pool size could be stable while functional NAD+-dependent signaling deteriorates.
  • The “blood NAD+ rises after NR” finding does not establish that NR supplementation has aging benefit — just that it raises blood NAD+. Whether this translates to functional outcomes is a separate question handled by trial-level evidence on nr and nmn.

Wiki-relevant implications

This study materially shifts how the wiki should frame three things:

  1. nad-precursors biomarker rationale. The class-page premise — “supplement to restore an age-related decline in NAD+” — needs hedging at the whole-blood level. The hypothesis is not falsified, but its strongest version (“blood NAD+ drops with age and supplementation corrects it”) is now directly contradicted by the most rigorous human assessment available. Tissue-level decline (muscle, liver) remains a live possibility.

  2. Blood NAD+ as an aging biomarker. Any future biomarkers/nad-blood.md page should open with this paper’s negative result rather than treating blood NAD+ as a useful aging readout.

  3. Pre-analytical handling discipline. Prior blood-NAD+ literature that used frozen-without-methanol samples is now methodologically suspect; the wiki should not cite NAD+/age correlations from such studies without caveat. (Specific superseded works: re-evaluate references 7–11 in this paper — Chaleckis 2016, Wang 2023, Yang 2022, Euro 2025 preprint, Breton 2020.)

Gaps and follow-ups

  • #gap/needs-tissue-level-replication — Is muscle/liver/brain NAD+ also stable with age in humans? Janssens 2022 found muscle NAD+ correlates with healthy aging, but used a different assay. Re-run the validated UHPLC-HRMS method on muscle biopsy banks.
  • #gap/no-mechanism — If functional NAD+-dependent signaling deteriorates with age but pool size is stable, what is the functional readout that captures the deterioration? (NAD+/NADH ratio, sirtuin activity reporters, PARylation flux.)
  • #gap/needs-replication — Independent replication of the negative blood result by labs without ties to the NADIS consortium would strengthen the claim.

Footnotes

  1. Janssens GE, Grevendonk L, Schomakers BV, et al. Healthy aging and muscle function are positively associated with NAD+ abundance in humans. Nature Aging 2, 254–263 (2022). DOI: 10.1038/s43587-022-00174-3. Cited in the present paper as reference 5; the same Houtkooper / Janssens group, but human muscle (not blood). The 2026 paper is a partial methodological self-correction at the blood-tissue level — not a refutation of the muscle finding. ↩ ↩2

  2. Nature Portfolio Reporting Summary PDF accompanying the article, 42255_2026_1537_MOESM1_ESM.pdf (4 pages). Last updated by authors 04 Mar 2026; available at https://static-content.springer.com/esm/art%3A10.1038%2Fs42255-026-01537-5/MediaObjects/42255_2026_1537_MOESM1_ESM.pdf. n’s, trial registries, ethics statements, statistical-design checklist, software stack all verbatim from this artifact. ↩ ↩2 ↩3

  3. Source Data Fig. 1, sheet “Fig. 1e” (MOESM2). Fresh blood (n=3 replicates): 22.86, 18.10, 17.64 nmol/mL; −20°C 5 d (n=3): 10.74, 9.03, 8.10 nmol/mL. ↩

  4. Source Data Fig. 1, sheet “Fig. 1f” (MOESM2). Two-donor representative freeze-thaw series; full n=6 donors shown in Extended Data Fig. 1c. ↩

  5. Source Data Fig. 1, sheets “Fig. 1d” and “Fig. 1h” + Extended Data Fig. 1d (MOESM4). Methanol-preserved before thawing recovers NAD+ within ~1 SD of fresh values across n=15 paired technical replicates. ↩

  6. Source Data Extended Data Fig. 1a (MOESM4). 10 plasma donors (5M, 5F); NAD+ values: 0, 0.43, 0.40, 0.26, 0.34, 0.29, 0.58, 0.34, 0.48, 0.26 nmol/mL. ↩

  7. Source Data Fig. 1, sheets “Fig. 1c” (linearity) and “Fig. 1d” (spike-recovery) — MOESM2. ↩

  8. Source Data Extended Data Fig. 2c (MOESM5). n=4 paired animals; T0 values 7.10, 7.44, 6.97, 6.88 nmol/mL; T1 values 16.16, 17.64, 16.15, 11.61 nmol/mL; paired two-sided t-test. ↩

  9. Source Data Extended Data Fig. 5a (MOESM7), aging cohort NADP+ Older vs Younger, P = 0.031 (last row of sheet). ↩

  10. Source Data Extended Data Fig. 5b (MOESM7), CardioHT NADP+ vs age regression, R² = 0.237 (last row of sheet). ↩

  11. Source Data Extended Data Fig. 5c (MOESM7), ELITE Older vs Athlete NADP+, P = 0.003 (last row of sheet). ↩

  12. Source Data Extended Data Fig. 5d (MOESM7), LLS NADP+ regression, R² = 0.028 (last row of sheet). ↩

  13. Source Data Extended Data Fig. 4a (MOESM6). Sample-size table for power analysis across Δ = 0.5 → 10 nmol/mL. ↩

Graph View

Backlinks