NMN versus NR: bioavailability, conversion to NAD+, and comparative clinical evidence in women's health and longevity

Stage: REPORT_EN
Cutoff date: 2026-09-14
Classification: NAD+ and ovarian function
Line: L9
Global maturity: H1 that both precursors elevate NAD in human blood; H0–H1 for their comparative pharmacology; H0 for female tissue mechanism, healthspan, and longevity
Scope: scientific research report; it does not prescribe or recommend a precursor, product, or dose, and it does not establish individual safety.

1. Executive scientific abstract

The available evidence does not permit declaring NMN superior to NR, NR superior to NMN, or the two clinically equivalent. It supports a narrower and more useful conclusion: both are orally bioactive precursors capable of increasing NAD in human blood, but precursor identity is partially lost during absorption and first pass, the comparative magnitude differs between the two direct human studies, and no pharmacological difference has yet been connected to a clinical function or to female longevity.

The two head-to-head trials published in 2026 do not measure exactly the same object. Christen et al. randomized 65 participants in a 14-day open-label study. NR and NMN increased whole-blood NAD versus placebo; the direct contrast was +6.32 μM in favor of NR, with 95% CI −3.92 to +16.56 and p=0.221. This demonstrates no superiority detected in that window, not equivalence. Berven et al. conducted an open-label crossover in only six people over eight days and observed an approximately 2.3-fold greater blood amplitude with NR. The within-person result warrants replication, but the small n, different products, equal mass rather than equal moles, absence of placebo in the comparative phase, possible carryover, and subsequent extension with NR only preclude turning it into a general hierarchy.

Biochemistry explains why counting steps is insufficient. NR can enter through nucleoside transporters and be phosphorylated by NMRK; NMN can be dephosphorylated to NR, hydrolyzed to nicotinamide (NAM), undergo CD38-catalyzed base exchange, or possibly enter intact as a minor and context-dependent fraction. Both can lose their identity as NAM; the microbiota can deamidate NAM to nicotinic acid (NA), and the host can rebuild NAD+ through NAPRT, NAMN, NAAD, and NADSYN1. Causality for this enterohepatic route has been demonstrated with tracers, germ-free animals, and loss of Naprt in mice, not in vivo in women. BST1 and CD38 can route NR and NMN differently in preclinical systems, but their material contribution after oral administration in humans remains unknown.

The main evidence directly in women also does not rank the precursors. In 25 postmenopausal women with overweight/obesity and prediabetes, NMN improved a measure of insulin-stimulated muscle glucose disposal, but it did not increase total muscle NAD, improve mitochondrial respiration or strength, and had no NR comparator. NR trials in older people or people with obesity increase NAD or turnover metabolites in blood and muscle, while metabolic, muscle, and vascular outcomes are mostly null or exploratory. Indirectly comparing these studies would violate transitivity because population, stage, dose, duration, matrix, and endpoint differ.

Donor-aware computational analysis of two human ovarian atlases likewise did not support the narrative that female aging creates an ovarian immune state progressively dominated by CD38 over BST1. In GSE202601, the young–older change in immune CD38−BST1 was +0.128 log2CPM, 95% CI −3.550 to +3.805; in GSE329417, the change in the immune compartment per decade was +0.034, 95% CI −0.827 to +0.895. Absence of a precise transcriptomic signal does not prove absence of enzymatic activity, but it prevents using age or menopause as a proxy for this routing state.

Relevance to women's health must be formulated by organ. NAD+ can support redox, respiration, PARP, and sirtuins; the effect depends on pool, flux, compartment, demand, and consumer. Favorable ovarian evidence comes mainly from aged female mice and does not compare NR. A potentially adverse osteoclast signal under low estrogen remains a preprint and competes with evidence that preserving NAD may support osteoprogenitors. Therefore, there is not yet a net sign across muscle–bone–ovary–vasculature–brain or a bridge to healthspan.

The report prioritizes three falsifiable hypotheses:

1. Primary, L9-3-AR-H5 v1: the published NR–NMN discrepancy is mainly explained by product, moles, time, matrix, and metrology; a standardized equimolar comparison will reduce the difference.
2. Competitor, L9-3-HG-H1 v2: NR retains a true early molecular advantage in new NAD through day 8, concordant between PBMC and whole blood and reproducible across lots and centers.
3. Translational, L9-3-HG-HT1 v2: if a causal route survives, tissue new NAD and function must be retained as a reproducible bivariate vector that adds external prediction; the prior flux/function ratio is refuted by instability.

The smallest discriminant is not an “anti-aging” trial. It is a Q0 → D0-PK → D0-MICROBE sequence: product and metrology qualification; an equimolar, double-blind female crossover with two lots and positional tracers; and quantification/perturbation of the NAM/NA route. Tissue and function open only after independent replication. Longevity remains outside this phase.

2. Scientific question and relevance

The operational question is:

With comparable oral molar input, formulation, food, and time, do NR and NMN differ reproducibly in the amount and route by which their atoms reach intracellular NAD+; does that difference persist in a target tissue and compartment; does it mediate a relevant function in women; and does that chain change with reproductive or endocrine state?

This formulation separates five estimands that are often conflated:

1. availability of the administered product;
2. systemic appearance of intact precursor and its derivatives;
3. incorporation of precursor-derived matter into intracellular NAD+;
4. functional availability in nucleus, cytosol, or mitochondrial matrix;
5. function, organ event, healthspan, and longevity.

“Bioavailability” is informative only when the measured object is specified. An increase in whole-blood NAD is a systemic pharmacodynamic response; it does not identify how much intact NR or NMN was absorbed, how much reached muscle or ovary, or how much new NAD was used functionally.

Female relevance arises from three facts, not from a pathway exclusive to women. First, reproductive transition changes estrogen signaling, body composition, insulin sensitivity, bone, vasculature, and inflammation. Second, these states can change the demand for and sign of an NAD perturbation even when absorption is identical. Third, clinical evidence has poorly characterized menstrual cycle, STRAW+10 stage, hormone therapy, and years since final menstrual period. The question is not whether “women respond differently” in the abstract, but whether a measured endocrine and tissue state modifies a specific gate.

3. Scope, population, and life stage

The long-term population of interest is adult women across natural reproductive aging: late reproductive stage, menopausal transition, and early postmenopause by STRAW+10. Chronological age should not be used as a substitute. Menstrual pattern, final menstrual period, E2/FSH in context, hormone use, adiposity, glycemia, renal and hepatic function, vitamin B3 diet, activity, sleep, microbiome, inflammation, and prior precursor exposure should be described.

For the first discriminating study, a single stratum is proposed: STRAW+10 +1c, approximately 3–6 years after the final menstrual period and age 50–60 years. This restriction reduces cyclic variation and permits deciding pharmacology before opening a stage interaction. Only after replication at a second center is late transition STRAW −1 added and precursor×stage formally estimated.

Primary ovarian insufficiency, surgical or induced menopause, unstable hormone therapy, pregnancy, lactation, and immediate attempts to conceive are etiologically distinct states and remain outside the initial estimand. This exclusion is not a safety or efficacy conclusion. Mexican and Latin American populations are a later transportability priority, not a basis for extrapolating effect sizes obtained in other contexts.

4. Background knowledge and mechanism map

4.1 NAD+ as a compartmental system

NAD+ participates in redox transfer and serves as substrate for PARPs, sirtuins, CD38/BST1, and SARM1. Its concentration is the result of synthesis, import, consumption, export, and dilution. In compartment c:

d[NAD+]c/dt = synthesisc + importc − consumptionc − exportc − dilutionc.

Therefore:

• a stable pool can conceal high synthesis and consumption;
• a high pool can reflect lower utilization;
• total NAD(H) is not equivalent to NAD+;
• NAD(P)H autofluorescence does not separate NADH from NADPH;
• NAAD, MeNAM, and pyridones are route or turnover footprints, not function;
• blood, PBMC, muscle, ovary, and brain are not interchangeable.

Nucleus and cytosol exchange intermediates, while SLC25A51 transports NAD+ into the mitochondrial matrix. A total cellular difference may disappear, amplify, or reverse after this gate. No active human NR–NMN study has traced both molecules to matrix NAD+.

4.2 NR and NMN conversion network

The relevant network is:

• NR → entry through ENT in some systems → NMRK1/2 → NMN → NMNAT1/2/3 → NAD+.
• Extracellular NMN → dephosphorylation to NR → NMRK route → NAD+.
• NR or NMN → NAM → NAMPT + PRPP → NMN → NAD+.
• NAM → microbial deamidation → NA → NAPRT → NAMN → NAAD → NADSYN1 → NAD+.
• NR + NA ↔ NAR + NAM through BST1 activity in preclinical systems.
• NMN + NA → NAMN + NAM through CD38-catalyzed base exchange.
• Intact NMN → putative context-dependent direct entry → NMNAT → NAD+.

Intact NMN entry cannot be treated as a rule. Ratajczak et al. observed extracellular conversion of NMN to NR and dependence on NRK1 in hepatocytes/cells. Triple-isotope tracing by Sauve et al., including phosphate, found small and tissue-dependent intact whole-body incorporation in male mice at high doses. Grozio et al. proposed SLC12A8 as a murine NMN transporter, but that interpretation was challenged and does not establish a general human gate.

4.3 Enterohepatic convergence

Yaku et al. used WT, germ-free, antibiotic-treated, or Naprt-deficient mice and showed that both precursors can feed a late phase of hepatic NAD through degradation to NAM, bacterial deamidation to NA, and Preiss–Handler. This is the strongest available causal evidence for a common route, but it comes from mice, mainly males, with doses and microbiomes that are not directly transportable.

Christen et al. observed metabolites compatible with that route and human fecal fermentation ex vivo. Without an oral tracer or in vivo perturbation, it cannot be claimed that the microbiota produces most human precursor-derived NAD. Ex vivo feces also do not represent small intestine, transit, epithelium, liver, or the full diet.

4.4 Ectoenzymatic routing

BST1/CD157 and CD38 are paralogs, not equivalents. BST1 can hydrolyze NR and catalyze base exchange; CD38 consumes NAD/NMN and can form NAMN from NMN and NA. The inferred consequence is that activity, location, co-substrate, and shedding could retain information about the precursor even when blood NAD converges.

The ovarian-immune hypothesis of an age-increasing CD38:BST1 ratio was weakened. Analysis of GSE202601 and GSE329417 did not show a positive, precise, and reproducible increase within immunity. A rare subpopulation with activity decoupled from RNA remains possible, but it is scientific only if its prevalence is predefined and activity→flux is demonstrated with loss/rescue; “rarity” cannot indefinitely rescue a negative result.

4.5 From NAD to function

Consumer state can dominate response:

• PARP requires damage and substrate; high protein or NAD does not prove faithful repair.
• Sirtuins require NAD, substrate, and context; SIRT abundance is not equivalent to catalysis.
• CD38/BST1 process precursors and also consume nucleotides/signal.
• Muscle, osteoclast, osteoblast, oocyte, endothelium, and neuron have different demands.

The central error this report avoids is assuming a monotonic chain of “more blood NAD → more tissue NAD → better function → greater longevity.”

5. Evidence method

The Scientist charter, cumulative memory packet, project.json, and the SCOPING, EVIDENCE_MAP, EVIDENCE_VERIFICATION, MECHANISTIC_SYNTHESIS, COMPUTE_DECISION, COMPUTE_OPTIONAL, HYPOTHESIS_GENERATION, ADVERSARIAL_REVIEW, and EXPERIMENT_DESIGN artifacts were reread in full. The synthesis begins from their verified claims and preserves contradictions rather than smoothing them away.

The search was updated through 2026-09-14 using primary sources, PubMed, articles/supplements, and official registries. The two direct human studies from 2026, the status of the osteoclast preprint, and NCT06991712 were reverified. The glaucoma trial remains RECRUITING, with estimated n 138, estimated completion 2026-12-31, and no posted results. No third published human NR–NMN comparison with clinical function was identified.

For each decisive source, design, population, sex/stage, n, intervention, comparator, matrix, endpoint, estimate, uncertainty, attrition, multiplicity, and conflicts were evaluated. Levels are labeled as direct human, indirect human/ex vivo, animal, cellular/biochemical, computational, or inferred. A nonsignificant p value was never converted into equivalence.

6. Evidence map

7. Contradictory evidence and null results

7.1 The head-to-head studies do not converge on one inference

Christen reduces the plausibility of a huge blood difference at day 14; Berven provides an NR>NMN signal through day 8. They cannot be informally averaged because design, population, schedule, products, moles, and window differ. Explanations rank as follows:

1. product, molar dose, time, matrix, and n=6 error;
2. a true early kinetic difference;
3. a majority common route with a specific local fraction;
4. blood composition or preanalytics;
5. modification by female state, still unsupported by data.

7.2 Blood versus tissue

NR consistently elevates the blood NAD metabolome, but in human muscle the NAD+ pool and function are often null. NMN produced a clamp signal in Yoshino without a detectable change in muscle NAD. These discordances do not prove hidden flux; they open rivals: uncaptured flux, extramuscular/non-NAD signaling, a tissue ceiling, an unsuitable endpoint, or a false positive.

7.3 Conversion versus function

Biochemical conversion is not equivalent to function. NAAD/MeNAM can rise without respiration, strength, FMD, or clamp. A common NAM/NA route also does not imply functional equivalence: a minority fraction could reach a critical compartment or produce another signal.

7.4 Ovary and dose

In mice, the oocyte can show a deficit diluted by bulk ovary; low-dose NMN improved a reproductive outcome that high-dose NMN did not improve. This prevents assuming dose–NAD–function linearity. Ovarian CD38 studies also differ between developmental programming and acquired aging.

7.5 Bone

The osteoclast preprint suggests that more NAD/SIRT3 can oppose the E2 brake on resorption. In contrast, peer-reviewed work links NAD loss to osteoprogenitor depletion and bone loss. Net balance could be positive, negative, or null. This tension must not be converted into a clinical warning or benefit; formation and resorption must be measured separately.

8. Multiscale mechanistic synthesis

Molecular

NR has a nucleoside entry and phosphorylation route; NMN faces a phosphate barrier that may be resolved by dephosphorylation, hydrolysis/base exchange, or context-dependent intact entry. This difference could affect early speed. First pass and microbiota erase part of identity toward NAM/NA. BST1/CD38 activity can preserve specific residual routes.

Cellular

New NAD is distributed among cytosol, nucleus, and mitochondria. Response depends on PRPP, ATP, glutamine, NMRK, NAMPT, NAPRT, NMNAT, NADSYN1, and SLC25A51, as well as consumption. A cell can maintain a stable pool at the cost of greater turnover. Only a tracer and mass balance separate these possibilities.

Tissue

Blood/PBMC are accessible pharmacology matrices, not substitutes for muscle, bone, ovary, or brain. In muscle, an advantage is functional only if new NAD in the same compartment precedes an independent readout such as glucose uptake and if loss/bypass orders the chain. In ovary, immune, stromal, granulosa, cumulus, and oocyte compartments must be separated. In bone, osteoclast and osteoblast must not be merged.

Systemic and female

Stage may modify absorption, routing, or demand, but current evidence favors first studying downstream susceptibility: E2, adiposity, insulin resistance, inflammation, renal/hepatic function, and activity can alter the NAD→function transfer function. The negative ovarian analysis weakens the hypothesis of using menopause as a CD38-high state.

Healthspan and longevity

The required chain is:

precursor → new NAD in tissue → persistent causal function → organ-specific event → years free of disease/disability → survival with competing risks.

No NR–NMN comparison has closed beyond the first blood link. Gait, clamp, ovarian reserve, BMD, NAAD, sirtuin expression, or “biological age” are not substitutes for longevity.

9. Computational layer

Public bioinformatics was used to test a necessary condition of the ovarian ectoenzyme hypothesis. BioNeMo was not used because structure or embeddings cannot resolve enzymatic activity, absorption, or flux. Raw counts were aggregated by donor and cell type in two atlases:

• GSE202601: 42,568 nuclei, eight donors, four young and four older.
• GSE329417: immune compartment of a postmenopausal atlas of 28 donors, ages 50–84 years.

Decisive results:

• GSE202601 immunity, older−young CD38−BST1: +0.128 log2CPM; 95% CI −3.550 to +3.805; exact p 0.971; direction unstable leave-one-donor-out.
• GSE202601 stroma: +0.947; 95% CI −1.380 to +3.273; p 0.286; exploratory direction without replication.
• GSE329417 broad immunity: +0.034 per decade; 95% CI −0.827 to +0.895; p 0.938.
• GSE329417 macrophages: −0.256 per decade; 95% CI −0.934 to +0.423; p 0.460.

No association in the frozen panel survived FDR ≤0.10. Immune composition did not explain a large aggregate contrast. The result is negative and informative: it deprioritizes age-based ovarian-immune localization, but does not refute CD38/BST1 biochemistry or demonstrate a dominant microbiome route.

10. Primary hypothesis

L9-3-AR-H5 v1 — design, product, and metrology null

Falsifiable statement: with content, purity, dissolution, moles, schedule, food, quench, matrix, cellular composition, period, order, and lot harmonized, the paired NR–NMN difference in AUC and absolute change of new NAD will be materially smaller than that observed by Berven and will fall within a prespecified mechanistic margin.

Role: primary.
Status: active.
Maturity: H1 for observed heterogeneity; H0 for causal attribution.
Confidence: 0.62.

Parents: L9-3-AR-C1/C2/C3/C10, L9-3-K9 v1, and discrepancies E1–E17.

Mechanism: equal mass delivered approximately 13–15% more NR molecules; products, salts, dissolution, schedule, and windows differed. Whole blood integrates cellular composition and preanalytics. An n=6 magnifies uncertainty. After controlling these layers, precursor name will contribute little independent information.

Predictions:

1. the contrast decreases with equimolar dosing and absolute change;
2. lot/product explains part of the variation;
3. direction will not be stable between centers if product dominates;
4. a signal confined to whole blood will not replicate in PBMC;
5. NR and NMN will continue to increase NAD versus placebo even if the active contrast falls within operational equivalence.

Evidence for: marked heterogeneity between the two direct studies, non-equimolar doses, different products and times, large individual variation, and lack of tissue concordance.

Evidence against: Berven observed a within-person difference that persisted after mathematical correction for molecular weight; it was also present in absolute change.

Kill criteria: two independent centers, two lots, double-dummy, and equimolar dosing replicate NR>NMN in PBMC and blood new NAD in the same window, without lot/center/order interaction and with subsequent tissue concordance.

Smallest discriminant: Q0 followed by D0-PK.

11. Competing hypothesis

L9-3-HG-H1 v2 — true early kinetic advantage of NR

Falsifiable statement: in women under harmonized equimolar conditions, NR produces a greater AUC of new NAD+ than NMN through day 8 in PBMC and whole blood; the difference must replicate across lots and centers before function is studied.

Role: biological competitor.
Status: weakened relative to v1; it no longer presupposes convergence at plateau or functional advantage.
Maturity: H1 for one human signal; H0 for replication and tissue.
Confidence: 0.24.

Parents: L9-3-SCOPE-H1 v2, L9-3-MECH-H1 v1, L9-3-HG-H1 v1, and Berven 2026.

Mechanism: nucleoside entry and NMRK could give NR a faster early amidated phase. NMN would require dephosphorylation or other routing before convergence. No claim is made about day 14–21 because NMN was not followed longitudinally in the same study.

Predictions:

1. NR/NMN ratio >1.25 for AUC of new NAD through day 8;
2. concordance between PBMC and blood after count normalization;
3. effect independent of baseline, period, order, and lot;
4. replication at a second center;
5. follow-up of both at days 14 and 21 decides convergence, persistence, or reversal.

Evidence for: Berven's within-person signal; ENT/NMRK routes; possibility that phosphate delays NMN.

Evidence against: n=6, open-label design, different products, possible carryover, mass rather than molar equality, lack of replication, and imprecise/nonsignificant contrast in Christen at day 14.

Kill criteria: equivalence CI within 0.80–1.25 in two matrices; percentage-only signal caused by lower baseline; dependence on lot/order; PBMC–blood discordance; or replication failure.

Smallest discriminant: D0-PK, without a functional challenge at this stage.

Subordinate mechanistic hypothesis: L9-3-HG-H2 v2

The common NAM/NA route remains alive but is no longer equated with function. It predicts that >50% of late new NAD from both precursors loses identity through NAM/NA and that pncA/NAPRT control a causal fraction; it dies as the dominant route if the fraction is <50% or perturbation does not change flux despite valid engagement. This subhypothesis explains how AR-H5 could be true, but it must be adjudicated separately.

Parked female hypotheses

• L9-3-HG-H3 v2: endocrine state can modify tissue-specific NAD→function without a prespecified muscle–bone sign. H0, confidence 0.12.
• L9-3-HG-H4 v2: a rare ovarian niche with CD38/BST1 activity could route locally despite null mean RNA. H0, confidence 0.05 and strict no-go.

12. Translational hypothesis

L9-3-HG-HT1 v2 — bivariate vector, not ratio

Falsifiable statement: only after validating a causal route, tissue new-NAD AUC and an independent function, retained as a bivariate vector with joint uncertainty, will be reproducible and add external prediction beyond nominal precursor, baseline/blood NAD, stage, and covariates.

Role: translational/metrological.
Status: the ratio of new-NAD AUC to functional change is refuted and replaced; the bivariate version remains parked.
Maturity: H0.
Confidence: 0.03.
Requirement: HUMAN_QA_REQUIRED before partnering or translational evaluation.

Mechanism: a baseline pool mixes synthesis and consumption; a dynamic readout could contain different information. Dividing by a functional response near zero amplifies error, can reverse sign, and conditions on the outcome. Jointly modeling both components avoids that defect.

Predictions:

1. each component achieves ICC ≥0.75 and error smaller than the target difference;
2. the association retains direction when leaving out woman, period, and center;
3. it improves external error and calibration over the comparator model;
4. the gain disappears if the causal route dies or if blood does not agree with tissue;
5. it does not require post hoc selection of time, tissue, subgroup, or threshold.

Evidence for: repeated discordance among blood NAD, metabolites, muscle pool, and function suggests uncaptured dynamic information.

Evidence against: no reproducible multicenter tissue assay, integrated causality, or external incrementality exists.

Kill criteria: ICC/CV/drift fail; no external gain; dependence on post hoc selection; blood–tissue discordance; or death of the causal hypothesis.

Smallest discriminant: blinded test–retest pilot only after positive D1.

13. Integrated falsifiable predictions and elimination criteria

14. Discriminating experiment

Gate Q0 — product, assay, and identifiability

Two lots of each active, 3.0 mmol/day corrected for salt, potency, water, and purity. Qualify identity, degradants, moisture, and dissolution. Validate isotope-resolved LC–MS/MS for NR, NMN, NAM, NA, NAR, NAMN, NAAD, NAD+, NADH, MeNAM, and pyridones where the matrix allows.

Go criteria:

• potency 95–105% and between-lot difference ≤5%;
• recovery 85–115%;
• intra/interday CV ≤10% for NAD+/NADH and ≤15% for fractions;
• isotope mass balance 80–120%;
• error ≤10 percentage points in blinded route mixtures;
• fewer than 20% below LLOQ at decisive times;
• MDC95 test–retest <20%.

If error does not fit within the operational margin 0.80–1.25, equivalence is not declared. If the NMN phosphate label cannot be identified, the number of estimable routes is reduced before dosing. Microtracer identity requires CMC, toxicological, and ethical review.

D0-PK — equimolar female crossover

Population: women at STRAW+10 +1c, age 50–60 years, 3–6 years since FMP.
Design: three periods of NR, NMN, and placebo; Williams sequences; double-blind, double-dummy; two lots balanced within woman; 21 days per period; minimum 28-day washout and demonstrated individual return to baseline.
Exposure: 3.0 mmol/day and stable microtracer ≤1 mol%, conditional on Q0 and ethics.
Sampling: day 1 at 0, 0.5, 1, 2, 4, 8, 12, and 24 h; day 8 at 0, 2, 4, 8, and 24 h; days 14 and 21 predose and 4 h; washout follow-up.
Matrices: PBMC primary; whole blood as bridge; plasma, urine, and feces. Hematocrit, complete blood count, PBMC number/viability, and time to quench explicit.

Coprimary outcomes:

1. paired geometric NR/NMN ratio of AUC0–24 h of absolute new NAD+ in PBMC on day 8;
2. paired difference in absolute baseline→day-8 predose change in total PBMC NAD+.

They are modeled jointly, not as a ratio. The model includes participant, treatment, period, sequence, lot, baseline, and carryover. Operational equivalence requires the 90% CI of both within 0.80–1.25, valid Q0, and sensitivity versus placebo. A material difference requires the 95% CI entirely outside the margin, PBMC–blood concordance, and small treatment×lot/center interaction.

Size: n=54 completers, initial recruitment 62, and blinded variance re-estimation after 24 completers, maximum 72. This size is based on a crossover with plausible paired log SD 0.40–0.55 and 90% power for the provisional margin; it is not sized for function or rare safety events.

D0-MICROBE — NAM/NA dominance

The in vivo component estimates the late fraction of new NAD by route; “dominance” is frozen as >50% for both precursors with a 95% CI excluding 50%.

In 24 autologous fecal communities, selected before response is known:

• live community, inactivated community, and sterile control;
• pncA-competent consortium and isogenic pncA-inactive background;
• intestinal epithelium–hepatocyte system with intact NAPRT or NAPRT reduced by two tools;
• bypass with NA;
• lumen–effluent–cell balance, barrier, and viability.

The decisive readout is the fraction of new hepatocyte NAD that disappears with pncA/NAPRT and reappears with NA. Ex vivo capacity without in vivo concordance demonstrates possibility, not human predominance.

Conditional scaling

D1-MUSCLE opens only after a replicated difference/route. Myotubes from female donors receive exposures achieved in D0; mitochondrial new NAD and insulin-stimulated 2-deoxyglucose uptake are measured. NMRK1, NAMPT, or NAPRT is perturbed according to the winning route, with rescue/bypass. If equalizing new NAD equalizes function, the difference is pharmacokinetic; if function persists, downstream pharmacodynamics or a non-NAD mechanism is sought.

Bone remains parked until replication/review of the osteoclast node and must separate resorption from formation. Ovary advances only with tissue not intended for clinical use, preserved architecture, a predefined subpopulation, and activity→flux with perturbation. No module opens a longevity endpoint.

15. Biomarkers and stratification

No validated biomarker exists for selecting NR or NMN. Current measures are:

• pharmacodynamic: NAD in blood/PBMC;
• exposure/turnover: NR, NMN, NAM, MeNAM, and pyridones;
• unvalidated route measures: NAR, NAMN, and NAAD;
• context: hematocrit, cell counts, renal/hepatic function, B3 diet, functional microbiome, E2/FSH, and stage;
• experimental tissue measures: absolute and compartmental new NAD.

A parameter advances to candidate only after identity, stability, ICC/CV, tissue concordance, causality, and external incrementality. NAAD alone demonstrates neither Preiss–Handler nor benefit. STRAW+10 is biological stratification, not a routing biomarker.

16. Individual variability

Probable sources of heterogeneity are:

1. product, salt, purity, dissolution, and food;
2. moles and schedule;
3. absorption, transit, and microbiome;
4. hematocrit and blood/PBMC composition;
5. NMRK, NAMPT, NAPRT, NMNAT, CD38/BST1, and SLC25A51;
6. vitamin B3 diet and previous exposure;
7. renal/hepatic function and excretion;
8. adiposity, insulin resistance, activity, sleep, and inflammation;
9. measured endocrine state;
10. quench, matrix, LLOQ, lot, and carryover.

“Responders” will not be created through post hoc thresholds. Variation must be modeled continuously and replicated; an exploratory subgroup does not rescue a null primary outcome.

17. Pharma relevance and maturity

Pharmacological routing platform

The most mature opportunity is analytical: an isotope-resolved system capable of distinguishing product, route, and cellular delivery before efficacy trials. Maturity H0–H1. Gates are Q0, reproducible D0, and external replication. Risks: identifiability, tracer CMC, cost, sampling burden, and blood–tissue concordance. HUMAN_QA_REQUIRED before partnering.

Precursor or formulation selection

This would make sense only if a specific route produces functional tissue delivery. If NAM/NA dominates, optimization should focus on common exposure and tissue fate, not the NMN/NR name. Maturity H0. The difference may belong to the product, and a majority route may not be functional.

NMRK/NAPRT/CD38/BST1 nodes

No node should be prioritized from RNA or association. Perturbation, bypass, function in a specific female tissue, therapeutic window, selectivity, and toxicology are required. CD38/BST1 have pleiotropic immune and metabolic functions. Preclinical maturity H0.

Pharma judgment: there is an opportunity for a translational pharmacology platform; no precursor, formulation, or target is ready for candidate selection, clinical claims, or partnering evaluation without human review.

18. Limitations

1. There are only two direct human comparisons; one has n=6 and both match mass, not moles.
2. Whole blood is the main comparative matrix; there is no active–active tissue tracing.
3. The causal microbial route comes from mice and ex vivo systems.
4. The most direct female functional evidence studies NMN without NR in a narrow phenotype.
5. Separate NR and NMN trials are not transitive.
6. No head-to-head study is powered for sex or STRAW+10.
7. Most entry/routing mechanisms come from male mice, cell lines, or purified enzyme.
8. The ovarian analysis has few donors, surgical selection, nuclear RNA, and incomplete metadata; it does not declare activity equivalence.
9. The bone finding remains a preprint; it proves neither human risk nor net balance.
10. Industrial conflicts are common and increase the need for independent replication.
11. The 0.80–1.25 margin is operational/metrological, not clinical.
12. The positional microtracer may not pass identifiability or ethical review.
13. PBMC does not represent muscle, ovary, bone, or brain.
14. No hard events, healthspan, or comparative longevity outcomes exist.
15. The dated search may omit regional registries or still-unpublished results.

19. Conclusions

NMN and NR are bioactive precursors that share much of a metabolic network, but they are not automatically pharmacologically identical. NR has a nucleoside route; NMN can be dephosphorylated, hydrolyzed/base-exchanged, or perhaps enter intact in specific contexts. Both can converge through NAM/NA and Preiss–Handler. The human magnitude of each fraction remains unmeasured comparatively.

Comparative clinical evidence stops at blood NAD. One study at day 14 did not detect superiority; another with n=6 at day 8 favored NR. Neither establishes equivalence, tissue superiority, function, chronic safety, or longevity. Evidence in women is fragmentary and does not permit a hierarchy.

The strongest inference for women's health is negative but productive: reproductive stage must not be used as a proxy for absorption or routing, blood NAD must not be used as a proxy for organ, and an organ effect must not be used as a proxy for healthspan. The donor-aware ovarian analysis lowers the priority of an age-dependent CD38:BST1 explanation. A possible tissue-specific differential sign remains H0.

The decisive test is logically small, though operationally demanding: qualify product and assay, compare equimolarly within women using tracers and two lots, adjudicate the NAM/NA route, and replicate before opening tissue or function. The result may be a molecular difference, operational equivalence, or metrological impossibility; all three would be advances. What the evidence does not permit is choosing a winner today for women's health or longevity.

20. References

1. Christen S, Redeuil K, Goulet L, et al. The differential impact of three different NAD+ boosters on circulatory NAD and microbial metabolism in humans. Nature Metabolism. 2026;8:62–73. DOI: 10.1038/s42255-025-01421-8. PMID: 41540253. NCT05517122.
2. Berven H, et al. The NAD-brain pharmacokinetic study of NAD augmentation in blood and brain using oral precursor supplementation. iScience. 2026;29:114764. DOI: 10.1016/j.isci.2026.114764. PMID: 41858901. NCT05698771.
3. Yoshino M, Yoshino J, Kayser BD, et al. Nicotinamide mononucleotide increases muscle insulin sensitivity in prediabetic women. Science. 2021. DOI: 10.1126/science.abe9985. PMID: 33888596.
4. Okabe K, Yaku K, Tobe K, Nakagawa T. Effect of oral administration of nicotinamide mononucleotide on clinical parameters and nicotinamide metabolite levels in healthy Japanese men and women. Frontiers in Nutrition. 2022. DOI: 10.3389/fnut.2022.868640. PMID: 35479740.
5. Katayoshi T, et al. Nicotinamide adenine dinucleotide metabolism and arterial stiffness after long-term nicotinamide mononucleotide supplementation. Scientific Reports. 2023. DOI: 10.1038/s41598-023-29787-3. PMID: 36797393.
6. Yi L, et al. The efficacy and safety of β-nicotinamide mononucleotide supplementation in healthy middle-aged adults. GeroScience. 2023. DOI: 10.1007/s11357-022-00705-1. PMID: 36482258.
7. Trammell SAJ, Schmidt MS, Weidemann BJ, et al. Nicotinamide riboside is uniquely and orally bioavailable in mice and humans. Nature Communications. 2016. DOI: 10.1038/ncomms12948. PMID: 27721479.
8. Martens CR, Denman BA, Mazzo MR, et al. Chronic nicotinamide riboside supplementation is well-tolerated and elevates NAD+ in healthy middle-aged and older adults. Nature Communications. 2018. DOI: 10.1038/s41467-018-03421-7. PMID: 29599478.
9. Elhassan YS, Kluckova K, Fletcher RS, et al. Nicotinamide riboside augments the aged human skeletal muscle NAD+ metabolome. Cell Reports. 2019. DOI: 10.1016/j.celrep.2019.07.043. PMID: 31412242.
10. Remie CME, Roumans KHM, Moonen MPB, et al. Nicotinamide riboside supplementation alters body composition and skeletal muscle acetylcarnitine concentrations in healthy obese humans. American Journal of Clinical Nutrition. 2020. DOI: 10.1093/ajcn/nqaa072. PMID: 32320006.
11. McDermott MM, Martens CR, Domanchuk KJ, et al. Nicotinamide riboside for peripheral artery disease: the NICE randomized clinical trial. Nature Communications. 2024;15:5046. DOI: 10.1038/s41467-024-49092-5. PMID: 38871717.
12. Ratajczak J, Joffraud M, Trammell SAJ, et al. NRK1 controls nicotinamide mononucleotide and nicotinamide riboside metabolism in mammalian cells. Nature Communications. 2016. DOI: 10.1038/ncomms13103. PMID: 27725675.
13. Sauve AA, et al. Triple-Isotope Tracing for Pathway Discernment of NMN-Induced NAD+ Biosynthesis in Whole Mice. International Journal of Molecular Sciences. 2023. DOI: 10.3390/ijms241311114. PMID: 37446292.
14. Yaku K, et al. Nicotinamide riboside and nicotinamide mononucleotide facilitate NAD+ synthesis via enterohepatic circulation. Science Advances. 2025. DOI: 10.1126/sciadv.adr1538. PMID: 40117359.
15. Yaku K, Palikhe S, Izumi H, et al. BST1 regulates nicotinamide riboside metabolism via its glycohydrolase and base-exchange activities. Nature Communications. 2021;12:6767. DOI: 10.1038/s41467-021-27080-3. PMID: 34799586.
16. Madawala RE, et al. CD38 mediates nicotinamide mononucleotide base exchange to yield nicotinic acid mononucleotide. Journal of Biological Chemistry. 2025;301:108248. DOI: 10.1016/j.jbc.2025.108248. PMID: 39894219.
17. Grozio A, Mills KF, Yoshino J, et al. Slc12a8 is a nicotinamide mononucleotide transporter. Nature Metabolism. 2019. DOI: 10.1038/s42255-018-0009-4. PMID: 31131364.
18. Schmidt MS, Brenner C. Absence of evidence that Slc12a8 encodes a nicotinamide mononucleotide transporter. Nature Metabolism. 2019. DOI: 10.1038/s42255-019-0085-0. PMID: 32694648.
19. Kropotov A, et al. Equilibrative Nucleoside Transporters Mediate the Import of Nicotinamide Riboside and Nicotinic Acid Riboside into Human Cells. International Journal of Molecular Sciences. 2021. PMID: 33573263.
20. Luongo TS, Eller JM, Lu MJ, et al. SLC25A51 is a mammalian mitochondrial NAD+ transporter. Nature. 2020;588:174–179. DOI: 10.1038/s41586-020-2741-7. PMID: 32906142.
21. Bertoldo MJ, Listijono DR, Ho W-HJ, et al. NAD+ repletion rescues female fertility during reproductive aging. Cell Reports. 2020. DOI: 10.1016/j.celrep.2020.01.058. PMID: 32049001.
22. Arslan A, et al. Nicotinamide Mononucleotide and Nicotinamide Riboside Reverse Ovarian Aging in Rats Via Rebalancing Mitochondrial Fission and Fusion Mechanisms. Pharmaceutical Research. 2024. DOI: 10.1007/s11095-024-03704-3. PMID: 38684562.
23. Perrone R, et al. CD38 regulates ovarian function and fecundity via NAD+ metabolism. iScience. 2023;26:107949. DOI: 10.1016/j.isci.2023.107949.
24. Yang Q, et al. NADase CD38 is a key determinant of ovarian aging. Nature Aging. 2024;4:110–128. DOI: 10.1038/s43587-023-00532-9. PMID: 38129670.
25. Høyland LE, et al. Subcellular NAD+ pools are interconnected and buffered by mitochondrial NAD+. 2024. PMID: 39702414.
26. Marques-Carvalho A, et al. Estrogens protect bone mass by inhibiting NAD+ metabolism in osteoclasts. bioRxiv preprint. 2025. DOI: 10.1101/2025.07.11.664289. PMID: 40791440. Not peer reviewed at cutoff.
27. Kim H-N, et al. A decrease in NAD+ contributes to the loss of osteoprogenitors and bone mass with aging. npj Aging and Mechanisms of Disease. 2021;7:8. DOI: 10.1038/s41514-021-00058-7.
28. Jin C, Wang X, Yang J, et al. Molecular and genetic insights into human ovarian aging from single-nuclei multi-omics analyses. Nature Aging. 2025;5:142–158. DOI: 10.1038/s43587-024-00762-5. Data: GSE202601.
29. Byrne J, Martin N, Soygur B, et al. Single-nucleus profiling reveals age-associated remodeling opposed by parity in the postmenopausal human ovary. bioRxiv preprint. 2026. DOI: 10.64898/2026.05.11.724381. Data: GSE329417.
30. Shats I, Williams JG, Liu J, et al. Bacteria Boost Mammalian Host NAD Metabolism by Engaging the Deamidated Biosynthesis Pathway. Cell Metabolism. 2020. DOI: 10.1016/j.cmet.2020.02.001. PMID: 32130883.
31. ClinicalTrials.gov. Comparisons of NAD Precursors for Neuroenhancement in Glaucoma Patients. NCT06991712. Status at 2026-09-14: recruiting; estimated n 138; no results.

Concrete scientific delta of the report

Adversarial integration changes the conclusion from a comparison of “boosters” to a test of causal identity: the most likely explanation of the current blood discrepancy is product–moles–time–metrology, but it competes with a true early kinetic advantage of NR and with a common NAM/NA route not yet quantified in women. The negative transcriptomic result prevents attributing that difference to an age-driven, CD38-dominant ovarian immune state, while the Q0→D0-PK→D0-MICROBE design establishes how to falsify all three explanations before opening tissue, function, or longevity.