For people with type 2 diabetes or prediabetes, a gram a day for three months moves the kind of lab numbers — HbA1c, triglycerides, a sugar-aging marker called CML — that don't feel like anything but mean something on bloodwork. For everyone else: a low-grade anti-inflammatory and anti-glycation effect, mechanism strong, payoff invisible from the inside. The catch most bottles leave out: most of an oral dose breaks apart in the bloodstream within minutes, so what's marketed as carnosine acting is partly its two building blocks acting in disguise. Roughly forty dollars a year to try it; the bar to expect more than "modest, slow, lab-readable" is high.
Carnosine is two amino acids — β-alanine and histidine — joined in a particular way your body builds on demand. It is stocked at high concentration in muscle and brain (where it helps soak up the acid generated during hard effort) and in tissue exposed to a lot of sugar (where it intercepts the chemistry that browns and stiffens protein over time). The same chemistry that browns bread crust also yellows the lens of an aging eye, stiffens arterial walls, and cross-links the collagen under your skin — a slow background tax called glycation. Carnosine reacts with the sugar fragments and aldehydes that drive that reaction, and with already-modified proteins, before they cross-link further Hipkiss 2010. The molecule mops up the spill before it sets.
The catch shows up the moment you swallow a capsule. Carnosine gets absorbed intact through the gut, hits the bloodstream — and a circulating enzyme called serum carnosinase cuts most of it back into its two building blocks within minutes Boldyrev et al. 2013. In healthy adults, plasma carnosine is barely detectable after a dose unless the blood sample is specifically stabilised. About one-seventh of the dose survives intact long enough to clear in the urine over five hours. How fast this happens depends on a common gene variant: some people clear carnosine slowly, others quickly, and current trials don't usually check for it. The practical implication, important to hold throughout: when a label sells you on what "carnosine" is doing, a meaningful share of the effect is actually β-alanine and histidine acting separately in their own right.
What the trials actually show
The trial that anchors the metabolic case is small and surprisingly clean.
A 0.6-point HbA1c drop is the kind of move a clinician notices on a follow-up panel — modest by drug standards, real by supplement standards. The trial is one centre, 54 people, twelve weeks; that is the strongest single piece of human evidence the substance has.
Pool nine small trials in 350 people total and you get a reproducible low-grade anti-inflammatory shift: CRP down, TNF-α down, a marker of oxidative damage to fats down, the antioxidant enzyme catalase up Saadati et al. 2024. The effect sizes are small, the certainty rating low, the populations mixed. Translation for a person: a quiet anti-inflammation and anti-glycation effect you cannot feel but can sometimes measure on a panel.
For cognition, the largest trial to date split 299 healthy adults across three age bands and gave them 2 g/day for up to 12 weeks.
The age inversion is awkward for the marketing — the people buying "anti-aging" supplements skew older, and the only cognition signal in the largest trial lives in the young. Why is not yet known. A small companion brain-imaging study did at least confirm something basic and important: oral carnosine measurably reaches the brain after dosing, peaks around an hour later, and returns to baseline by five hours Ali et al. 2025.
What is not shown in humans: a longevity effect, a measurable skin-aging effect from oral dosing, or a cataract-reversal effect from eye drops Dubois & Bastawrous 2017. The cell-culture data — fibroblasts cultured for months in carnosine show slower telomere shortening and extended division capacity — uses concentrations a swallowed capsule cannot reach systemically Shao et al. 2004. Useful for mechanism; not transferable as a promise to a person.
What the marketing oversells
"It loads muscle carnosine for performance." If the goal is more carnosine in your quadriceps for a hard interval session, the established route is β-alanine: 3 to 6 grams a day for four to ten weeks reliably raises muscle carnosine by 30 to 60% Rezende et al. 2020. Whether swallowing L-carnosine itself reaches muscle as efficiently — given the serum-enzyme problem — has not been shown in any head-to-head trial. The sports-nutrition industry quietly knows this; it is why the bottles on the gym-supplement shelf are mostly labelled β-alanine.
"It reverses cataracts when you drop it in your eye." A topical version called N-acetyl-carnosine was sold for years on the back of small Russian trials reporting 90% improvements in lens clarity. Cochrane reviewed the literature and found the supporting studies too weakly designed to support a conclusion of either prevention or reversal Dubois & Bastawrous 2017. The marketing budget has outrun the replication budget by a long way.
"It's a flagship anti-aging molecule." The mechanism — reacting with the sugar fragments behind slow protein damage — is real and elegant. The human evidence for what that translates into on a person is: a 0.6-point HbA1c shift in 54 diabetics, a few inflammation markers that move modestly, and one positive cognition trial confined to adults under 35. That is a useful low-cost adjunct. It is not a longevity drug.
If you want to try it
The doses with actual human data behind them are 1 to 2 grams a day, taken as two 500-milligram capsules with meals, for eight to sixteen weeks before judging anything. With food cuts the nausea. Give it a full quarter before deciding it works or doesn't, because the endpoints that actually move on the evidence — fasting glucose, triglycerides, an inflammation panel settling — operate on bloodwork timescales, not gym-mirror timescales.
What does the same job better, depending on the job
For muscle pH buffering in hard intervals: β-alanine, whose evidence base is decades deep and dose-response well mapped Rezende et al. 2020. For blood sugar in type 2 diabetes: metformin, an order of magnitude more potent and one of the most studied molecules in medicine. For slowing the sugar-aging of skin and tissue: cutting dietary sugar moves the same chemistry from the input side; for visible skin aging on a year scale, topical retinoids out-perform topical carnosine on every meaningful clinical endpoint. For cataract: surgery, whose outcomes are unrelated to anything in a bottle.
None of these are reasons not to take carnosine. They are reasons to be clear-eyed about which slot it actually fills — a low-cost adjunct with a narrow useful window, not a replacement for any of the above.
One narrowing worth saying out loud, because it cuts across most of what's above: the strongest evidence is in people whose metabolic numbers already have room to fall. If your fasting glucose is 95 and your HbA1c is 5.2, the same dose that moves a diabetic 13 mg/dL and 0.6 points has almost nothing to bring down. The Houjeghani trial selected its effect by selecting its population. For prediabetic and type 2 diabetic adults, the metabolic case for trying it is the strongest the substance has Houjeghani et al. 2018. For metabolically healthy adults, expect the meta-analysis-grade effect — modest inflammation and oxidative-stress numbers nudging in the right direction Saadati et al. 2024 — not the headline-grade one. Vegetarians and vegans, who eat far less of the foods that supply dietary carnosine (red meat, poultry, fish), plausibly have a bigger gradient for supplementation to close, though no trial has stratified the question directly.
What's adjacent
- β-alanine — the precursor and what the sports-nutrition aisle is actually selling when it talks about loading muscle carnosine.
- Advanced glycation end-products (AGEs) — the slow chemistry carnosine targets; lowering dietary AGE intake works the same dial from the other end.
HbA1c— the long-window blood-sugar lab number that captures the kind of slow shift carnosine seems to move in diabetics.- Topical retinoids — what to reach for if the actual goal is visibly different skin in months, not decades.
- Metformin — the substantially more potent glucose lever, for the people who need one.
Substance + claimed effects
L-carnosine is a naturally occurring dipeptide of β-alanine and L-histidine, stored at millimolar concentrations in skeletal muscle and brain tissue Boldyrev et al. 2013. Marketed as an oral supplement (typical capsules 500 mg), it is claimed to: buffer muscle pH during high-intensity exercise; inhibit non-enzymatic protein glycation and the formation of advanced glycation end-products (AGEs); lower inflammatory and metabolic markers in type 2 diabetes; protect skin collagen from sugar-driven aging; slow cellular and telomere aging; and improve cognition. This entry covers all of these holistically, and treats the central pharmacological catch — that orally administered L-carnosine is hydrolysed rapidly in human serum by carnosinase 1 (CN1) — as load-bearing throughout.
Evidence by addressing question
Mechanism
Buffering and Ca2+ handling. Carnosine's imidazole ring has a pKa near physiological pH; it accepts H+ generated during glycolysis and contributes roughly 7–15% of muscle's intracellular buffering capacity Boldyrev et al. 2013. It also acts as a diffusible Ca2+/H+ exchanger between sarcoplasmic reticulum and cell membrane, plausibly improving contraction efficiency.
Anti-glycation / carbonyl scavenging. Carnosine reacts directly with reactive carbonyls (methylglyoxal, glyoxal, 4-hydroxynonenal) and with already-glycated protein adducts, blocking the cross-linking that forms AGEs Hipkiss 2010. The histidine residue captures the carbonyl; the β-alanine residue is unusual in mammalian peptides and resists peptidase digestion in tissue.
Antioxidant / anti-inflammatory. Carnosine chelates transition metals (Cu, Zn), quenches singlet oxygen and hydroxyl radicals, and downregulates NF-κB-driven cytokine output in cell and animal models Boldyrev et al. 2013.
The bioavailability catch. Oral L-carnosine is absorbed intact via the PepT1 peptide transporter, but circulating serum carnosinase (CN1) cleaves it back to β-alanine and histidine within minutes. In healthy humans, ~14% of an ingested dose appears as intact carnosine in 5-hour urine, with plasma levels mostly undetectable unless samples are stabilised at collection Boldyrev et al. 2013. CN1 activity is set largely by a CTG repeat polymorphism in CNDP1 — homozygotes for the short (5L) allele secrete less enzyme and clear carnosine more slowly. This means a substantial fraction of the dose acts pharmacologically not as carnosine but as its precursors β-alanine and histidine — relevant for what muscle vs. systemic effects can plausibly be attributed to the supplement itself.
Brain penetration. In a 2025 healthy-volunteer study using 1H-MRS of the posterior cingulate cortex, oral carnosine raised brain carnosine signal at 10 g single doses, peaking at one hour and returning to baseline by five hours Ali et al. 2025 — the first direct human demonstration that the dipeptide (or a marker of it) reaches brain tissue after oral dosing.
Evidence
Glycemic control. The pivotal small trial: Houjeghani 2018, double-blind, 54 adults with type 2 diabetes randomised to 1 g/day L-carnosine or placebo for 12 weeks Houjeghani et al. 2018. Versus placebo, the carnosine arm dropped fasting glucose by 13.1 mg/dL, HbA1c by 0.6 percentage points, triglycerides by 29.8 mg/dL, and serum carboxymethyl-lysine (a circulating AGE) by 91.8 ng/mL; body fat fell ~1.5% and lean mass rose ~1.7%. Insulin resistance (HOMA-IR) did not change.
Inflammation and oxidative stress. A 2024 meta-analysis pooled 9 RCTs (n=350) of carnosine and histidine-containing dipeptides Saadati et al. 2024. Significant pooled reductions: CRP -0.97 mg/L, TNF-α -3.60 pg/mL, malondialdehyde -0.34 µmol/L, with catalase increased. IL-6, adiponectin, GSH, SOD, TAC did not move. GRADE certainty was rated low to very low — small trials, mixed populations, modest effect sizes.
Cognition. The NEAT trial (2025) randomised 299 healthy adults to 2 g/day oral carnosine vs placebo for up to 12 weeks, stratified by age band O'Toole et al. 2025. The 23–35 age stratum showed significant improvements in processing speed and prefrontal-cortex-linked tasks; the 36–50 and 51–65 strata showed essentially no effect. An earlier Japanese RCT in 39 elderly subjects on a 1 g/day anserine+carnosine mix found preserved verbal episodic memory and improved brain blood flow on SPECT Hisatsune et al. 2016. Across the small literature, effects on standard MCI/Alzheimer batteries (MMSE, ADAS) have not consistently emerged.
Cataract (topical N-acetylcarnosine). Cochrane reviewed the topical NAC eye-drop evidence (Babizhayev trials cited 88–90% improvement in glare, visual acuity, posterior subcapsular opacity at 6 months) and concluded that, due to design weaknesses and lack of independent replication, there is no convincing evidence that NAC drops reverse or prevent cataract progression Dubois & Bastawrous 2017.
Skin / anti-glycation. Mechanism-grade. Ex vivo human skin biopsies treated with carnosine-containing topicals show measurable AGE-formation inhibition; clinical cosmetic trials of carnosine creams over 8–16 weeks report improved firmness and elasticity — but these are small, often open-label, and industry-funded. No oral-dose RCT exists that uses photographic, instrument-validated skin outcomes as a primary endpoint.
Cellular aging. Shao et al. (2004) showed that human fibroblasts grown continuously in 20 mM carnosine had a slower telomere shortening rate and extended population-doubling lifespan; quiescent cells in carnosine accumulated less telomeric DNA damage Shao et al. 2004. The cell-culture concentration is unattainable systemically by oral dosing.
Muscle buffering (oral L-carnosine vs β-alanine). Muscle carnosine elevation is achievable — but the well-established route is oral β-alanine (~3–6 g/day for 4–10 weeks raises muscle carnosine 30–60%) Rezende et al. 2020. Whether equivalent doses of oral L-carnosine itself load muscle as efficiently is unclear: serum carnosinase hydrolyses most of the ingested dipeptide, and the resulting β-alanine is then the rate-limiter for re-synthesis inside the myocyte. The exercise-performance literature is mostly built on β-alanine, not L-carnosine.
Protocol
Most oral RCTs use 1–2 g/day in divided doses (e.g. 500 mg twice daily with food) for 8–16 weeks; the Houjeghani T2DM trial used 1 g/day for 12 weeks Houjeghani et al. 2018; the NEAT cognition trial used 2 g/day O'Toole et al. 2025. Taking with food reduces nausea. Co-ingestion with histidine-containing dipeptides (anserine, in chicken/turkey extracts) can saturate serum carnosinase and raise circulating carnosine.
Contraindications
Safety has been formally characterised: single doses up to 10 g are well tolerated; 15 g produces unacceptable rates of headache (44%), nausea (22%), and paraesthesia (22%); long-term 5 g twice daily for 4 weeks produced no adverse events or biochemical abnormalities in healthy adults Ali et al. 2025. Pregnancy / breastfeeding: no human safety data, default exclusion from supplement trials. Theoretical interactions: additive hypoglycemic effect with diabetes medications; concurrent use with metformin / sulfonylureas / insulin warrants monitoring. β-alanine release from hydrolysis can produce paraesthesia at high single doses.
Misconceptions
"Oral L-carnosine is the way to load muscle carnosine." Largely false. Oral β-alanine has decades of replicated dose-response data showing muscle carnosine increases of 30–60% Rezende et al. 2020; equivalent direct data for oral L-carnosine is sparse, partly because of serum hydrolysis. For sport performance, β-alanine is the conventional choice.
"Carnosine reverses cataract through eye drops." Cochrane: no convincing evidence Dubois & Bastawrous 2017.
"It's a dominant anti-aging peptide." Marketing extrapolates from in vitro fibroblast longevity (at 20 mM, unreachable orally Shao et al. 2004) and from senescence-accelerated mouse models. No human longevity RCT exists, and the clinical effect sizes observed (HbA1c −0.6, TG −30 mg/dL in T2DM) are useful but unspectacular.
"More dose, more effect." The dose-response above 2 g/day is poorly characterised; pharmacokinetic ceilings from carnosinase kick in early. The Ali 2025 brain-MRS data show the signal returns to baseline by 5 h post a 10 g dose Ali et al. 2025.
Failure-modes
- Carnosinase clearance. Individuals with high-activity CN1 alleles likely get less circulating carnosine per gram dosed; this isn't typically genotyped in trials and may explain heterogeneous responses.
- Wrong endpoint. Readers chasing visible skin payoff or noticeable energy expecting topical-cosmetic returns from a pill that mostly produces modest changes in glycemic and inflammatory markers.
- Wrong substance. Loading muscle for performance with L-carnosine when β-alanine is the established route.
- Dose drift. Single megadoses ≥15 g cause symptomatic side effects without added benefit Ali et al. 2025.
Audience
Strongest evidence: adults with type 2 diabetes or pre-diabetes, where the metabolic and AGE markers are elevated to begin with and have room to move Houjeghani et al. 2018. Cognition signal: only the 23–35 age band in the NEAT trial showed statistically significant effects on speed/efficiency O'Toole et al. 2025 — opposite of the demographic that buys "anti-aging" supplements. Vegetarians have lower dietary carnosine intake (red meat, chicken, tuna are the main food sources, providing ~50–250 mg/day in a mixed-meat diet) and may have a larger gradient to fill from supplementation, though no vegetarian-specific RCT has been published.
Alternatives
For muscle carnosine loading and exercise performance: β-alanine 3–6 g/day (well-established) Rezende et al. 2020. For glycemic control: metformin (orders of magnitude stronger evidence base), dietary AGE restriction, exercise. For skin glycation: dietary sugar reduction; topical retinoids (orders of magnitude stronger evidence than topical carnosine for visible skin aging). For cataract: surgical replacement, not eye drops.
Practicalities
Available OTC as 500 mg vegetarian capsules; major brands (NOW, Swanson, Life Extension, Nutricost) retail around USD 0.10–0.20 per gram. At 1 g/day, annual cost is roughly USD 30–60. No prescription required, not regulated as a drug.
Stakes / payoff
Below the felt-effect floor for most healthy users — the meaningful trial effects (HbA1c, TG, AGEs) are invisible without bloodwork. T2DM/prediabetes readers might see fasting-glucose movement on quarterly labs after 12 weeks. Cognition payoff narrow to under-35 in current data. Onset latency 8–14 weeks for the metabolic endpoints.
The credibility range
Optimist case. Carnosine sits at a real biological junction: it scavenges the reactive carbonyls that drive both protein cross-linking and chronic inflammation; mechanism is unusually clean and conserved across mammals; meta-analytic data show reproducible decreases in CRP, TNF-α, MDA Saadati et al. 2024; the Houjeghani trial's HbA1c drop at 1 g/day is the kind of effect normally costing pharmaceutical-tier intervention Houjeghani et al. 2018; in vitro the fibroblast lifespan and telomere data are striking Shao et al. 2004; brain delivery has now been demonstrated in humans Ali et al. 2025; cost and safety are both extremely favourable.
Skeptic case. Almost every human trial is small (n≤60), single-centre, and short (12–16 weeks). The Cochrane review of the most-marketed application (NAC eye drops for cataract) concluded no convincing evidence Dubois & Bastawrous 2017. The pivotal cognition trial showed effects only in the youngest stratum — the demographic least bottlenecked on cognition O'Toole et al. 2025. The pharmacokinetic problem is real: serum carnosinase hydrolyses most of the oral dose, so much of what is sold as "carnosine supplementation" is, pharmacologically, β-alanine + histidine supplementation routed through a brief carnosine pulse. The Saadati meta-analysis itself rated almost all outcomes low or very low GRADE certainty. In vitro and animal data routinely use concentrations or routes unattainable in human oral dosing.
Author's call. A modest, low-risk supplement with a real but narrow signal in type 2 diabetes / prediabetes (anti-AGE, modest glycemic, anti-inflammatory) and a tentative cognition signal in young adults; everything wider — anti-aging, skin, longevity — is currently extrapolation from mechanism and cell culture, not human evidence. Worth knowing about; not worth selling as a flagship. Evidence scored 2; controversy 2 (the field disagrees on whether oral L-carnosine adds anything β-alanine + dietary histidine doesn't).
Stakeholder + incentive map
- Supplement industry (NOW, Swanson, Life Extension, Doctor's Best): commercial; the "anti-aging dipeptide" framing is the marketing lever. Cosmetic creams using carnosine carry similar incentive.
- Sports nutrition: largely pivoted to β-alanine (CarnoSyn is the patented form), where evidence is stronger; carnosine itself is a secondary niche.
- Academic carnosine researchers (Aldini / Milan group, Hipkiss, de Courten / Monash, Boldyrev): publish on mechanism and small trials; legitimately optimistic, but the pipeline is small-trial heavy.
- Diabetes nephropathy field: CNDP1 5L-allele protective effect generated genuine clinical interest in raising carnosine in kidneys; carnosinase inhibitors are an active drug-development line.
- Skeptics: Cochrane reviewers; clinical pharmacologists pointing at the bioavailability problem.
Population variability
- CNDP1 genotype. 5L homozygotes clear carnosine more slowly and have higher steady-state plasma carnosine; conventionally untyped in trials.
- Baseline glycemia. Effect sizes for HbA1c and AGEs appear larger in T2DM than in normoglycemic adults — the marker has further to fall.
- Age. NEAT trial cognition signal localised to 23–35; older strata flat O'Toole et al. 2025. The mechanism for the age inversion is unresolved (CNDP1 activity rises with age? prefrontal substrate?).
- Diet baseline. Vegetarians and vegans run lower endogenous carnosine intake (red meat, poultry, fish are the main sources). The supplement plausibly closes more of a gap for them; no vegetarian-stratified RCT.
- Sex. Beta-alanine response is not sex-dependent Rezende et al. 2020; oral L-carnosine sex-stratified data sparse.
Knowledge gaps
- No large, multicentre, long-duration human RCT for any single endpoint. The T2DM trial that anchors the metabolic case is n=54.
- No human longevity / mortality data.
- No head-to-head oral L-carnosine vs β-alanine trial for muscle carnosine loading.
- No instrument-validated, photographic skin-aging RCT for oral carnosine.
- Mechanism of the age inversion in the NEAT cognition signal.
- Whether CNDP1 genotype-stratified dosing improves response.
- Brain pharmacokinetics beyond a single 1-hour peak (Ali 2025 was a small sub-study) Ali et al. 2025.
Brief vs entry. The input description named "glycation, muscle buffering, skin aging, and cognition." The article covers glycation (as the anti-AGE mechanism, woven through mechanism / evidence / misconceptions), cognition (the NEAT trial gets its own anchor callout in evidence), and skin aging (covered as a consequence of the anti-glycation chemistry, plus a misconception flag on topical claims, scored on beauty_cumulative). Muscle buffering is covered via the misconception that L-carnosine is the route to load muscle carnosine — the actual evidence base is β-alanine, which is the substance the sports-nutrition aisle sells when it markets carnosine. Flagging that explicitly so the gap is the honest call, not a silent drop.
Hard scoping call. The pharmacokinetic catch (serum carnosinase hydrolysing most of an oral dose within minutes) is treated as load-bearing throughout — mechanism, misconceptions, controversy score. Without it, the marketing framing rules; with it, the entry's "modest, mostly mismarketed" through-line lands honestly. Considered structuring the entry around β-alanine instead; rejected because the topic brief is L-carnosine specifically and the dipeptide does have an independent (if small) human-trial base in glycemia, inflammation, and cognition.
Rating difficulties. focus at 1 is a close call: the NEAT trial signal is statistically real but confined to one age band in one centre. Scored 1 rather than 0 because the trial is the largest cognition RCT on the substance to date (n=299) and the effect appears at multiple follow-ups on multiple tests, but the age inversion is genuinely puzzling and the score is appropriately tentative. beauty_cumulative at 2 leans on mechanism plus in-vitro telomere data plus the meta-analytic anti-inflammation signal; no oral-dosing photographic skin RCT exists, which is why it's not a 3.
Future-link candidates. Separate entries on β-alanine, HbA1c, advanced glycation end-products, topical retinoids, and metformin are pointed at in the closing pointers; once they land, wire the cross-links.
Separate-entry candidates. β-alanine is the most obvious one — bigger evidence base, distinct use case (exercise performance), reader-search overlap that would benefit from its own page. Topical retinoids likewise.
Dream narrative lever. Overall score computed at ~19/100 — well below the 40 obligation floor. Narrative was written in the relief/debunking register rather than aspirational, because the honest hook here is what the reader gets back (money, supplement-stack creep, epistemic clarity) rather than what the substance unlocks. The dek and tagline were written straight, with the relief lever shaping framing but no marketing-words lift beyond what the §1 voice rules already allow.
Excluded by choice. The CNDP1 5L-allele protective story in diabetic nephropathy is real and interesting but didn't earn article real estate — too specialist and not actionable. Animal-only and cell-culture longevity data treated as mechanism colour, not claims (Shao 2004 explicitly flagged as in-vitro / unattainable orally inside the evidence section). Anserine/carnosine combination products (the Japanese trial line — Hisatsune et al. 2016) noted in research but not pulled forward to the article, because the combination changes the substance and would confuse a reader expecting plain L-carnosine.
L-Carnosine
OTC capsules retail around USD 0.10–0.20/g; 1 g/day works out to roughly USD 30–60/year.
Daily capsule with food. Trivial.
Mechanism is anti-glycation: carnosine reacts with reactive carbonyls and AGE precursors, the chemistry that cross-links collagen and elastin over decades (Hipkiss 2010). In vitro, fibroblasts cultured continuously in carnosine show extended lifespan and slower telomere shortening (Shao et al. 2004). Real but slow contribution; no instrument-validated photographic RCT of oral dosing supports a stronger score.
Meta-analytic decreases in CRP (-0.97 mg/L), TNF-α (-3.60 pg/mL), and MDA (-0.34 µmol/L) across 9 RCTs / 350 participants (Saadati et al. 2024). In T2DM, 1 g/day for 12 weeks dropped HbA1c 0.6 percentage points, fasting glucose 13.1 mg/dL, triglycerides 29.8 mg/dL (Houjeghani et al. 2018). Healthy users feel essentially nothing — improvements are biomarker-level, not felt-experience-level.
Modest anti-AGE and anti-inflammatory mechanism (Hipkiss 2010, Saadati et al. 2024); meaningful HbA1c reduction in T2DM (Houjeghani et al. 2018). No human mortality or longevity RCT. In vitro fibroblast lifespan extension uses concentrations unattainable orally (Shao et al. 2004). Mechanistically plausible additive effect; not a flagship longevity intervention.
One small T2DM RCT (n=54) anchors the metabolic case (Houjeghani et al. 2018); one larger cognition RCT (n=299, age-restricted signal) (O'Toole et al. 2025); a 9-trial / 350-participant meta-analysis on inflammation markers with low to very low GRADE certainty (Saadati et al. 2024); Cochrane no-convincing-evidence verdict on the topical cataract application (Dubois & Bastawrous 2017). Mechanism strong; trials small and short.
NEAT trial (n=299, 2 g/day, 12 weeks) found statistically significant gains in processing speed and prefrontal-task accuracy — but only in the 23–35 age stratum; older strata flat (O'Toole et al. 2025). Tentative, age-restricted signal.