Resveratrol and pterostilbene: stilbene chemistry, SIRT1, and the limits of bioavailability

Two closely related plant stilbenes are often compared as though replacing two hydroxyl groups with methoxy groups automatically produced stronger “longevity activation.” The chemical change does affect polarity, conjugation, and distribution, but its pharmacological significance cannot be established by a single bioavailability figure. For resveratrol, the gap between absorption of the total labelled dose and the concentration of free aglycone is especially important; for pterostilbene, the distinction concerns rodent findings, human measurements, and still limited clinical outcomes. At the enzyme level, the conditions under which SIRT1 activation was observed must be examined separately; at the patient level, the question is which changes were prespecified trial endpoints.

How plants synthesize the two stilbenes

Resveratrol and pterostilbene are specialized phenolic plant metabolites. Their production may increase with infection, ultraviolet irradiation, and tissue injury; “phytoalexin” describes an ecological function in the plant, rather than a protective mechanism in humans after eating a berry. Resveratrol occurs in grapes, peanuts, and Japanese knotweed, among other plants, whereas pterostilbene occurs in some Pterocarpus and Vaccinium species. Concentrations depend on cultivar, plant part, and growing conditions, so the presence of a compound in food does not specify a pharmacological exposure.

In the phenylpropanoid pathway, phenylalanine ammonia-lyase (PAL) converts L-phenylalanine into trans-cinnamic acid. Cinnamate 4-hydroxylase (C4H/CYP73A) produces p-coumaric acid, and 4-coumarate–CoA ligase (4CL) activates it to p-coumaroyl-CoA. Stilbene synthase (STS), a type III polyketide synthase, uses this starter unit and three malonyl-CoA molecules for sequential condensations with decarboxylation and closure of the second aromatic ring. The related chalcone synthase directs a similar polyketide intermediate towards a different cyclization and the flavonoid branch. A shared Cys–His–Asn catalytic triad characterizes the family, but specific residue numbers and active-site geometry belong to a named STS structure, rather than all plants. Both enzyme identity and reaction conditions affect the product profile.

Resveratrol O-methyltransferase (ROMT) then transfers methyl groups, rather than “methyl radicals,” from S-adenosylmethionine to the phenolic oxygens at positions 3 and 5. The pathway can be represented as follows:

$$\begin{aligned} \mathrm{L{-}Phe}&\xrightarrow{\mathrm{PAL}}\mathrm{cinnamate}\xrightarrow{\mathrm{C4H}}\mathrm{p{-}coumarate}\xrightarrow{\mathrm{4CL}}\mathrm{p{-}coumaroyl{-}CoA},\\ \mathrm{p{-}coumaroyl{-}CoA}&\xrightarrow[3\,\mathrm{malonyl{-}CoA}]{\mathrm{STS}}\mathrm{resveratrol}\xrightarrow[2\,\mathrm{SAM}]{\mathrm{ROMT}}\mathrm{pterostilbene}. \end{aligned}$$

This is a plant biosynthetic pathway, rather than conversion of resveratrol to pterostilbene in human tissues after administration.

Two aromatic rings are joined by a central ethylene linkage. Trans-resveratrol is $(E)$-3,5,4′-trihydroxystilbene, with formula $\mathrm{C_{14}H_{12}O_3}$; trans-pterostilbene is $(E)$-3,5-dimethoxy-4′-hydroxystilbene, with formula $\mathrm{C_{16}H_{16}O_3}$. The first therefore has three phenolic hydrogen-bond donor groups and the second has one, while methoxy oxygens remain potential acceptors. These descriptors help predict solvation and interactions, but do not by themselves determine absorption rate, free brain concentration, or affinity for SIRT1. Values of $\log P$ and topological polar surface area depend on the calculation or measurement method and should be compared within the same method. Precise van der Waals volumes without a stated calculation algorithm do not support a causal conclusion.

The ethylene bridge permits $(E)$ and $(Z)$ isomers. Light can induce photochemical conversion of the trans form to the cis form, along with other degradation reactions; solution composition depends on wavelength, solvent, oxygen, and time. Exposure studies therefore require isomer identification, storage conditions, and analytical separation. The claim that the cis isomer is always entirely inactive against SIRT1 or PDE and exactly 10–100 times weaker does not follow from a geometric model alone: each protein and substrate requires its own test. Greater pterostilbene lipophilicity may facilitate some steps in membrane partitioning, but neither the enterocyte membrane nor the blood–brain barrier reduces to a TPSA threshold. Protein binding, efflux, metabolism, and the free plasma fraction determine eventual tissue access.

Absorption is not the same as free circulating aglycone

In a human study of labelled resveratrol, Walle and colleagues showed that a substantial portion of an oral dose was absorbed while systemic exposure to the unchanged aglycone remained low. These measures have different denominators: the radioactive label includes the original molecule and its metabolites, whereas a free-resveratrol assay measures one chemical form alone. Thus, “resveratrol is hardly absorbed” is inaccurate; equally inaccurate is the conclusion that high absorption ensures a micromolar free-compound concentration at the target. Absolute percentages, $C_{\max}$, and half-life must be accompanied by dose, formulation, food conditions, sample matrix, and analytical method. The sequence “70–75% absorption → 0.3–0.8% free fraction → 10–50 nM in every person” combines measures from different protocols.

Most presystemic conversion occurs through phenolic groups. UDP-glucuronosyltransferases (UGTs) transfer glucuronic acid from UDP-glucuronate to the aglycone oxygen, producing resveratrol 3-O- and 4′-O-glucuronides, among other products. Sulfotransferases (SULTs), using 3′-phosphoadenosine-5′-phosphosulfate (PAPS), produce the corresponding sulfates, including the 3-O-sulfate. Enterocytes and liver differ in isoenzyme composition and donor-substrate availability; UGT1A1, UGT1A9, and SULT1A1 are possible participants, but a table of $K_m$ values from different enzyme systems does not establish their contribution in vivo. Competition between fluxes depends on affinity as well as $V_{\max}$, free-aglycone concentration, localization, and enzyme expression. Conjugates should not automatically be considered inactive: they have their own distribution, and local deconjugation can restore aglycone under certain conditions, although that does not establish clinical benefit.

After polar metabolites form, ABC-family transporters, including apical enterocyte ABCC2/MRP2 and ABCG2/BCRP, can return some products to the intestinal lumen. Basolateral ABCC3/MRP3 carries certain conjugates into portal blood; hepatic uptake and canalicular secretion depend on substrate composition and transporter expression. Microbial β-glucuronidases can cleave some excreted intestinal conjugates, creating an opportunity for reabsorption, but a secondary plasma peak alone does not demonstrate sufficient free-form exposure at the target. The assumption that resveratrol necessarily uses SGLT1/GLUT1 by “imitating a monosaccharide” through phenolic hydroxyl groups requires direct transport evidence and should not replace established discussion of diffusion and efflux.

Pterostilbene's two methoxy groups block positions 3 and 5 from direct conjugation, but the phenolic 4′-OH remains available. Oxidative O-demethylation can also create new phenolic positions and additional metabolites; calling the 4′-monoglucuronide the “only pathway” is excessive. Pterostilbene and its conjugates did show a different pharmacokinetic profile in preclinical comparisons with resveratrol. However, the frequently repeated “80% bioavailability,” “15–20 times more in the brain,” and “7–10 times longer half-life” arise from individual animal models or incomparable conditions. They cannot be presented as a human comparative result without accounting for species, dose, matrix, and free fraction. Likewise, detection of a molecule in a brain homogenate does not establish the free neuronal concentration required for direct SIRT1 activation.

Oxidative metabolic pathways require chemical precision: piceatannol is a hydroxylated relative of resveratrol, while pinosylvin has a hydroxyl-group pattern different from simple mono-O-demethylated pterostilbene. Thus, equating “3-O-demethylpterostilbene = pinosylvin” is incorrect. Even when a metabolite is detected in a microsomal experiment, its plasma proportion and independent contribution to biological effects require separate measurement. The next question is whether the aglycone or metabolite concentration actually reached can alter a particular cellular SIRT1 reaction and predict a human outcome; these levels cannot be reduced to a marketing formula of “systemic superiority.”

When a stilbene changes the SIRT1 reaction

SIRT1 is an NAD⁺-dependent deacetylase whose catalytic core, with a nicotinamide-binding fold and zinc-binding region, works with regulatory segments on either side of the protein. An N-terminal region around Glu230 participates in responses to several small molecules, while the C-terminal ESA sequence helps maintain catalytic activity. In the elementary reaction, the enzyme binds an acetylated protein lysine and NAD⁺, releasing nicotinamide, deacetylated lysine, and 2′-O-acetyl-ADP-ribose: $\mathrm{Protein{-}Lys(Ac)+NAD^+\rightarrow Protein{-}Lys+nicotinamide+2′{-}O{-}acetyl{-}ADPR}$. NAD⁺ consumption is stoichiometric, but global cellular NAD⁺ does not equal local SIRT1 activity: compartment, availability of the particular protein substrate, nicotinamide inhibition, and competing enzymes all matter.

The idea of sirtuin-activating compounds (STACs) arose from experiments using resveratrol and a fluorescent model peptide. The objection was substantial: removing the bulky fluorophore eliminated observed activation in some assays. Hubbard and colleagues did not invalidate that artefact; they explained the conditional nature of the allosteric effect: hydrophobic amino acids near acetylated lysine permit responses in certain unlabelled substrates, while another peptide without the appropriate local context is not activated. Position +1 was important in their experiments, but a single Trp at +1 without a suitable neighbouring residue was insufficient. Accordingly, “any natural substrate with Trp/Phe/Tyr/Leu at +1 or +6 is activated” is too broad. The result applies to the named peptides, STAC concentrations, and nicotinamide or O-acetyl-ADP-ribose measurement methods.

Replacing Glu230 with Lys or Ala weakened SIRT1 responses to many activators studied by Hubbard while preserving basal activity in those tests. This supports the N-terminal region's role in allosteric coupling, but does not demonstrate one universal contact between resveratrol or pterostilbene and every proposed pocket residue. Structures with certain activators define the geometry of a particular enzyme complex; transferring those coordinates to another stilbene requires its own binding experiment. In particular, an exact Glu230–Arg446 ion pair, an invariant 3–8-fold decrease in $K_m$, and a pterostilbene advantage of $\Delta\Delta G_{bind}=-1{,}8$ kcal/mol cannot be declared established for every human protein substrate. $K_m$ is an operational enzyme-assay parameter, rather than the concentration of nuclear PGC-1α; comparison between peptides under different conditions does not predict deacetylation in vivo.

There is also an indirect pathway studied for resveratrol in cellular and animal models: inhibition of cyclic-nucleotide phosphodiesterases can increase cAMP, activate Epac1, affect calcium signalling and CaMKKβ, and then AMPK phosphorylation at Thr172. There is no need to insert an obligatory “CaMKKβ phosphorylates LKB1, which only then activates AMPK” step: CaMKKβ and LKB1 are separate upstream AMPK kinases under the relevant conditions. AMPK activity can accompany changes in NAD⁺ metabolism and interact with SIRT1, but positive feedback is not a self-sustaining engine independent of energy status. NAMPT participates in nicotinamide salvage, and its expression and pathway flux depend on tissue and time. Co-occurring changes in p-AMPK, NAD⁺, and protein acetylation do not identify whether direct SIRT1 allostery, PDE-dependent signalling, or secondary energy stress caused them.

In Hubbard's study, effects of selected STACs on cellular mitochondrial measures disappeared when the corresponding mouse SIRT1 residue was replaced; at the concentrations they used, the authors found no explanation through PDE or SIRT1-independent AMPK phosphorylation. This supports a direct mechanism in that system, without disproving Park's PDE route with a different compound, dose, or cell type. This distinction underpins a sound reading of the literature: one positive enzyme experiment cannot establish that pterostilbene activates SIRT1 in the human brain, and an artefact in an older fluorescent assay cannot establish that allostery never occurs.

Beyond SIRT1: an effector network, rather than a ready-made “rejuvenation” program

PGC-1α is a transcriptional coactivator involved in adaptation of energy metabolism. Its function is regulated by acetylation, phosphorylation, partner availability, and tissue context; the acetyltransferase GCN5 and SIRT1 can shift the acetylation balance in opposite directions. Deacetylating PGC-1α in a particular model may enhance coactivation of PPARα/ERRα and associated fatty-acid oxidation genes. However, a scheme in which a stilbene invariably deacetylates all “13 lysines” and instantly switches a cell from glycolysis to β-oxidation combines incompatible systems and undefined sites. CPT1A and acyl-CoA dehydrogenase expression are only part of the program; actual β-oxidation flux depends on acyl-group transport, substrate supply, and the respiratory chain. The CPT1 carnitine shuttle explains why a higher CPT1A transcript cannot be taken as a direct measurement of fatty-acid oxidation.

PGC-1α interacts with nuclear respiratory factors NRF1 and GABP/“NRF2” in historical nomenclature. The latter should not be confused with the antioxidant-response factor NFE2L2, also commonly called NRF2. Increased transcription of TFAM and nuclear respiratory-complex genes may accompany TFAM import through TOM/TIM, nucleoid organization, and mitochondrial-DNA maintenance. The human mitochondrial genome encodes 13 OXPHOS proteins, 22 tRNAs, and two rRNAs, but assembling functional complexes requires many nuclear proteins, transport, folding, and adequate energy. Therefore, increased TFAM, mitochondrial-DNA copy number, or stained “mitochondrial mass” individually do not demonstrate new mature cristae, an improved respiratory-control ratio, or increased human ATP. The acetyl-L-carnitine review provides a useful comparison: metabolic measures there likewise do not equal clinical neuroprotection.

FOXO1 and FOXO3a help redistribute cellular responses between defence, metabolism, and apoptosis. SIRT1-dependent deacetylation can change their transcriptional selectivity experimentally; SOD2, catalase, Sesn3, ATG5/ATG7, LC3, BNIP3, and GADD45 belong to possible downstream programs, but their induction need not occur together or in one direction. SOD2 converts superoxide to hydrogen peroxide, which is subsequently removed by several peroxidase systems and catalase in the appropriate compartments; claiming complete “prevention of cardiolipin peroxidation” from one transcriptional marker is excessive. Assessing mitophagy requires measurements beyond LC3 or BNIP3, including mitochondrial delivery flux to lysosomes and eventual functional recovery.

NF-κB connects inflammatory signals to cytokine and enzyme expression, but many stages determine its activity: IκB retention, nuclear translocation, RelA/p65 modification, and chromatin accessibility. In biochemical studies, SIRT1 deacetylated p65 at Lys310, altering transcriptional activity. This does not establish that oral resveratrol or pterostilbene invariably suppresses TNF-α, IL-1β, IL-6, ICAM-1/VCAM-1, iNOS, and COX-2 simultaneously in patients, still less that it resolves insulin resistance. Exposure, target acetylation, inflammatory biomarkers, and clinical outcomes require separate measurement.

Endothelial eNOS/NOS3 produces NO in the presence of L-arginine, O₂, BH4, calmodulin, and coupled electron transfer. SIRT1-dependent eNOS deacetylation and AMPK-dependent phosphorylation are two studied regulatory mechanisms; their interaction depends on cellular state. Yet increased Ser1177 phosphorylation or altered Lys496/Lys506 acetylation does not ensure available NO: BH4 deficiency or oxidative stress may alter reaction coupling. A cellular diagram therefore cannot imply reduced blood pressure, arterial stiffness, and thrombosis as obligatory effects of both stilbenes. Clinical vascular data need their own assessment.

What human trials showed

In Timmers and colleagues' double-blind crossover study, 11 men with obesity received resveratrol 150 mg/day or placebo for 30 days. The authors measured energy expenditure, metabolic variables, liver lipids, and muscle markers, observing changes resembling some adaptations to calorie restriction. The small sample and short duration cannot establish slowed ageing, diabetes prevention, or increased mitochondrial biogenesis in every organ. Moreover, changes in SIRT1 protein content and AMPK phosphorylation in a biopsy do not show that resveratrol directly occupied the SIRT1 allosteric pocket at the achieved free concentration. Absolute percentages across many biomarkers must be read alongside the prespecified endpoint and correction for multiple comparisons, rather than added into a single “therapeutic score.”

Type 2 diabetes studies used different formulations, doses, durations, and background glucose-lowering medication. A small Brasnyó study reported improved insulin-sensitivity measures, while Bhatt's work assessed HbA1c and other metabolic variables alongside standard treatment. These findings do not attribute every change specifically to SIRT1 or make resveratrol a replacement for established diabetes treatment. Interpretation requires the randomized between-group difference, baseline HbA1c, duration, weight changes, and prespecified primary endpoint; a within-treatment-group “before/after” HbA1c comparison can overestimate the effect. In steatotic liver disease, Chen's study reported ultrasound and biochemical changes, but ultrasound and transaminases do not directly measure fibrosis or long-term liver events. MASLD reflects current terminology; participants in historical studies may have entered under different NAFLD criteria.

Single-dose vascular-response studies, including Wong's work, measured brachial-artery flow-mediated dilation (FMD). FMD depends on ultrasound methodology, baseline diameter, haemodynamics, and time after administration. Thus, an acute FMD increase at particular doses does not establish sustained reductions in blood pressure, vascular stiffness, or myocardial-infarction incidence. Kennedy's study measured cerebral blood-flow variables during task performance; regional perfusion and cognitive performance are different endpoints. The direct chain “stilbene → eNOS deacetylation in a human vessel → more NO → clinical thrombosis prevention” requires measurements at every level and is not established by these experiments alone.

For pterostilbene, Riche's small four-group double-blind trial is especially informative: 80 adults with elevated total cholesterol and/or LDL received pterostilbene 125 mg twice daily, 50 mg twice daily, pterostilbene 50 mg plus grape extract 100 mg twice daily, or placebo for 6–8 weeks. In a baseline-adjusted model, LDL rose by 17.1 mg/dL with monotherapy, whereas this increase was not seen in the combination group; systolic and diastolic blood pressure were respectively 7.8 and 7.3 mmHg lower in the high-dose group. The trial was small and short, participants differed in lipid-lowering medication use, and the combination comparison does not prove that grape extract reliably prevents an LDL increase in other patients. A PPARα mechanism was not established as the cause of the lipid change in this group. The result does not indicate that pterostilbene must be combined with resveratrol or grape extract, or prescribed to treat hypertension.

Experiments in ageing mice and amyloid-pathology models provide separate signals concerning memory, tau, Aβ, and inflammatory markers. Even if pterostilbene outperforms equimolar resveratrol in a particular model, species, diet composition, exposure timing, and brain analytics limit translation. A Doppler observation in volunteers does not measure dementia prevention. Neither compound therefore currently supports a claim of established human neuroprotection or pterostilbene superiority for cognitive outcomes. “More bioavailable” does not mean “more effective”: greater free exposure may strengthen beneficial and adverse actions alike.

Genotype and metabolism: plausibility is not yet personalized treatment

Regulatory SIRT1 variants, including rs7895833 and rs12778366, have been studied in associations with metabolic traits. However, an association between genotype and baseline expression or disease does not demonstrate a “genotype × stilbene treatment” interaction in a randomized trial. Such a pharmacogenetic recommendation requires a prespecified hypothesis, adequate genotype-subgroup sizes, measured active aglycone, a clinical outcome, and replication. A higher pterostilbene dose or a promise of reduced visceral fat cannot be prescribed to a carrier of one rs7895833 allele. HIC1/CtBP, miR-34a, and miR-217 are important cellular regulators of SIRT1 expression, but changing them in culture does not produce a diagnostic supplement-sensitivity test.

UGT1A1*28 is associated with reduced UGT1A1 transcription and Gilbert syndrome in the appropriate genotype and context. Yet resveratrol's fate is distributed among several UGTs, SULTs, transporters, and elimination routes; a 28/28 genotype therefore does not guarantee twice or three times the free-resveratrol exposure, still less an enhanced clinical effect. UGT1A93 and SULT1A12 variants can change enzyme activity in a biochemical assay, but their contribution to systemic concentration at a particular dose and formulation requires a dedicated PK study. Using “mutation” without distinguishing a functional variant from a pathogenic state overstates the significance of common polymorphisms.

PRKAA2 encodes the catalytic α2 AMPK subunit, rather than a regulatory subunit. Without functional and clinical confirmation, its genetic variants cannot predict responses to the PDE–cAMP–AMPK axis or select an “activator combination.” Allele rs2796498 cannot be used to prescribe alpha-lipoic acid, berberine, or the two stilbenes. For scientific discussion, genotype may serve as a stratification variable in a future study, but it is not a ready-made personalization algorithm.

Changing the formulation: which barrier does it remove?

For crystalline resveratrol, dissolution in intestinal fluid may limit the rate at which molecules reach the epithelial surface. Under a diffusion approximation, the Noyes–Whitney equation can be written as $\frac{dM}{dt}=\frac{DA}{h}(C_s-C)$, where $M$ is dissolved mass, $D$ the diffusion coefficient, $A$ particle surface area, $h$ diffusion-layer thickness, $C_s$ saturation concentration at the particle, and $C$ bulk concentration. Expressing concentration change $dC/dt$ at constant volume requires additional division by the medium volume; omitting it conflates dimensions. Micronization increases total surface area at equal mass and can speed dissolution, but does not eliminate conjugation after absorption. The pilot oncology study of micronized SRT501 measured its own exposure; differences from another formulation should not become an invariant “3–5-fold” advantage for all micronized products.

Phospholipid complexes, self-emulsifying drug-delivery systems (SEDDS), and liposomes address different tasks: dispersion, dissolved fraction, stability, and membrane contact. Each formulation requires measurements of particle size and distribution, isomer content, dissolution, and food interaction, followed by PK measurements of free and conjugated aglycone. A nanodroplet does not guarantee lymphatic transport: that route depends especially on incorporation into chylomicron assembly, lipid composition, and absorption physiology. Even if some administered compound avoids first-pass hepatic metabolism, intestinal UGTs/SULTs remain present. Therefore, no claim can be made that the whole liposomal or “phytosomal” dose enters arterial blood unchanged or reaches the brain.

Coadministration of piperine or quercetin has been studied to modify conjugation and transport. In Johnson and colleagues' experimental model, the increased plasma resveratrol signal with piperine depends on species, dose ratio, and analytical method; it does not establish a human regimen. Competition for SULT1A1 and PAPS consumption are mechanistically possible in an appropriate enzyme system, but quercetin itself is metabolized and selective clinical “switching off” of resveratrol sulfation has not been demonstrated. Nonselective interference with enzymes and transporters can alter drug exposure, so “higher bioavailability” requires a parallel safety assessment.

Doses, combinations, and the circadian argument

The 150 mg/day in Timmers and the 100 or 250 mg/day in Riche are regimens studied in specific populations, rather than approved therapeutic ranges for ageing prevention. Tolerability in a short trial does not make resveratrol 250–500 mg or pterostilbene 50–150 mg an “optimal” daily dose. Instructions to take them only in the morning, necessarily with fat, strictly in micronized form, or in a fixed stilbene ratio require direct regimen comparisons that the supplied literature does not provide. Food and formulation may alter dissolution and exposure, but the direction and size of the effect are determined by studies of the specific formulation.

SIRT1 uses NAD⁺ as a cosubstrate and an acetylated protein as its other substrate; a universal Michaelis–Menten formula cannot be substituted for this two-substrate mechanism without specifying binding order, product inhibition, and experimental constants. If $[A]$ denotes peptide and $[B]$ NAD⁺, velocity may depend on both, but changes in total NAD⁺ after NMN/NR do not necessarily increase the nuclear free fraction available to SIRT1, and an activator-induced change in one peptide's $K_m$ does not extend to every protein. CD38 and PARP consume NAD⁺ in particular contexts; an age-related association does not mean that it is “critically depleted” in every tissue. “STAC + NR/NMN” therefore remains a testable hypothesis with questions about tissue, biomarkers, and clinical endpoints, rather than a proven synergistic protocol. Adding piperine or quercetin introduces further potential interactions without resolving that uncertainty.

Circadian factors CLOCK:BMAL1 and NAMPT connect daily rhythms to NAD⁺ metabolism, but rhythm phase depends on tissue, sleep, feeding, and individual schedules. Rhythmic NAMPT expression does not establish a universal SIRT1 “receptivity window” of 07:30–09:30 or demonstrate harm from evening stilbene intake. Chronopharmacological conclusions require trials of the same dose at different times with exposure and clinical-endpoint measurements; the reviewed source set does not supply such data.

High doses should not be considered harmless because the molecule comes from plants. In the micronized SRT501 study in relapsed/refractory multiple-myeloma patients receiving bortezomib, serious adverse events included kidney injury; severe underlying disease and concomitant treatment prevent automatic attribution of the entire risk to resveratrol in healthy people, but equally prevent declaring gram-level doses safe. Pterostilbene safety is also limited by human-study size and duration; the LDL rise in Riche deserves attention, while failure to detect an event over a few weeks does not demonstrate absence of long-term risk.

Resveratrol has been studied as a modifier of drug-metabolizing enzymes, including CYP1A2, CYP2C9, and CYP3A4; Chow's human study assessed enzyme activity after administration. Translation to a particular drug requires consideration of free enzyme-site concentration, duration, formulation, isoenzyme, active metabolites, and therapeutic index. Biochemical CYP2C9 inhibition does not establish that every resveratrol dose necessarily raises INR in warfarin users, but combining it with a narrow-therapeutic-index drug requires individualized assessment and usual clinical monitoring. Automatically replacing warfarin with a direct anticoagulant or preventively lowering a statin dose on the basis of one in vitro constant is unsafe. Human data on pterostilbene–CYP interactions are even less certain.

The comparison supported by the evidence

A similar chemical scaffold does not make the molecules interchangeable, and greater exposure does not determine the direction of clinical effects. Future comparisons would benefit from equimolar doses, analytical separation of $(E)$/$(Z)$ isomers, free aglycone and conjugates, target confirmation in an accessible tissue, and prespecified outcomes. For now, the strongest conclusion is that resveratrol has studied but contextual SIRT1 allostery and limited clinical signals; pterostilbene differs chemically and pharmacokinetically, and one small human randomized trial showed a possible vascular effect alongside an adverse lipid shift. None of these steps justifies a ready-made SIRT1-activation regimen for healthy people.

How this review was prepared

This is a narrative review of the saved literature on stilbene chemistry, pharmacokinetics, and SIRT1; the set contains 30 publications. The editorial package was recovered on 8 October 2026 from the previously prepared Russian text and original bibliography. Bibliographic inconsistencies for Hubbard and Howells were corrected directly against PubMed records; this does not constitute renewed verification of every result or full text. No systematic search, meta-analysis, formal risk-of-bias assessment, or GRADE evaluation was conducted.

Enzyme allostery, cellular signalling, tissue concentration, and clinical outcomes are considered distinct levels of evidence. Numbers apply to the studied formulation, dose, population, duration, and measurement method. Animal pharmacokinetics do not establish human bioavailability; a positive surrogate measure does not demonstrate lower disease incidence or longer life. Transfer between molecules, isomers, and formulations requires its own comparative evidence.