Molecular Hydrogen (H2) — Body Handbook

Supplements · §522

Molecular Hydrogen (H2)

Hydrogen gas dissolves in water for a minute or two; you drink it; almost all of it exhales within the hour. No metabolite, no liver load, nothing to monitor. The strange part is that in small trials the markers still move — LDL fractions, oxidative-stress urine markers, post-exercise lactate. The reason it can do that without harming anything else is the most interesting thing about it: H2 selectively quenches the most damaging free radicals and leaves the useful redox signalling alone, the opposite of vitamin-E megadose era. Modest evidence, low risk, cheap.

Do · Daily Evidence Mixed Chapter Supplements

The lead is the mechanism: among antioxidants, this one knows the difference between the radicals that wear you down and the signals your cells run on. The evidence base sits at "small but coherent" — biomarker shifts in metabolic-syndrome cohorts, less lactate and easier-feeling hard sessions, mood and oxidative-stress changes in pilots, no large definitive trial yet. Daily effort is dropping a tablet in a glass; cost runs roughly thirty to eighty dollars a month.

Every antioxidant on the shelf has the same problem: the free radicals it neutralises include the ones your body uses for normal work. Hydrogen peroxide tells immune cells where the threat is. Nitric oxide tells blood vessels to relax. Megadosing vitamin E and beta-carotene blunts that signalling indiscriminately, which is roughly why those trials kept landing flat or slightly negative on hard endpoints.

H₂ appears to thread that needle. It reacts only with the worst-behaved radicals — hydroxyl (the one that snaps DNA and shreds membrane lipids) and peroxynitrite (the one that nitrates proteins) — and ignores hydrogen peroxide, superoxide, and nitric oxide at the concentrations your cells use to talk to each other. That selectivity is what makes it interesting; without it, H₂ would be another vitamin-E rerun.

The other half of the story is less direct. The amount of H₂ you actually drink is small (a few milligrams), and most of it exhales within the hour, so simply counting molecules cannot explain the size of the downstream effects. The current best read is that H₂ also acts as a kind of signal — nudging the cell's own antioxidant program (a master switch called Nrf2) on, and dialing inflammation transcription down (Ichihara et al. 2015, LeBaron et al. 2019). The cell does the heavy lifting; the H₂ is the prompt.

What the trials actually show

The pattern across about a decade of small human trials is consistent in shape: biomarkers move in the right direction, in people who had room to improve. The cleanest signal is in metabolic-syndrome and pre-diabetic cohorts — the populations carrying elevated oxidative load to begin with.

Outside metabolic syndrome, three other claims have small but real trial support. In hard exercise, drinking hydrogen-rich water before a session lowered blood lactate after the session and reduced the drop in peak muscle force in fatigue testing (Aoki et al. 2012); follow-up work found lower perceived exertion at the same workload (Botek et al. 2022, Ostojic 2014). In healthy adults, four weeks of hydrogen water improved mood scores and stress balance (Mizuno et al. 2017) and quietened inflammatory markers on circulating immune cells (Sim et al. 2020). In early Parkinson's, a small placebo-controlled trial showed the standard motor-symptom score improve in the H₂ arm and worsen in placebo over 48 weeks (Yoritaka et al. 2013) — a striking pilot result that the same group could not reproduce in a larger follow-up, and which remains the field's biggest unresolved question.

The honest summary: the size of the effects is small to moderate, the trials are small (typically ten to sixty people), and most of the network of researchers is in Japan, with overlapping authorship and some funding ties to the people who sell the devices. A 2015 review catalogues 321 papers across thirty-one conditions. The single large, independent, definitive trial that would settle the field has not been run.

How to actually take it

Three things go right in every positive trial: enough hydrogen actually dissolves, you drink it before it leaves the water, and you keep it up for weeks not days.

Countertop electrolysis machines reach higher and steadier concentrations and pay back over a few years if you're using daily, but the capital cost (a few hundred to a few thousand dollars) is the friction. Magnesium-stick infusers you drop in a bottle are cheap but variable, often well below the trial dose. Pre-bottled "hydrogen water" off a shelf has usually outgassed by the time you open it; treat it as flavoured water.

There is no titration, no blood test to track, no metabolite to clear. The protocol that fails is the one where the water sits on the counter while you finish a call.

Across roughly two decades of human trials — including critical-care work at much higher doses than anything sold in a tablet — molecular hydrogen has not produced a clear adverse-event signal. Your gut already makes grams of it daily as part of normal fermentation; the body has no metabolic problem disposing of it.

Three things this is not

Not alkaline water. "Alkaline" bottled waters at pH 9 are sold next to hydrogen products and often confused for them. Alkaline electrolysis machines do produce some H₂ in passing — but it's the dissolved gas that's doing the work in the trials, not the pH. A pH-only bottled water contains no hydrogen.

Not hydrogen peroxide. Hydrogen peroxide (H₂O₂) is the chemical opposite — a signalling pro-oxidant your white blood cells use to attack pathogens. H₂ gas (two hydrogen atoms bonded together, no oxygen) is what dissolves in the water you're drinking.

"Hydrogen water" on a shelf is not necessarily hydrogen water. A sealed plastic bottle off a supermarket shelf has usually outgassed to near-zero hydrogen by the time you crack the seal. The dose that matters is what's actually dissolved at the moment you drink it, not the number printed on the label at the factory weeks ago.

Why "I tried it and felt nothing" usually has a specific cause

Open container, lost dose. Hydrogen leaves water the way carbonation leaves an open beer. A glass that sits for ten minutes between tablet and sip has lost most of the dose.
Subtherapeutic delivery. Cheap magnesium-stick infusers and many bottled products produce well under 0.1 ppm — roughly a tenth of the trial range. The product label rarely tells you that.
Wrong baseline. The strongest effects sit in people with elevated oxidative load — metabolic syndrome, hard training, smoking, ageing, chronic illness. A twenty-five-year-old who already eats well, sleeps eight hours, and lifts three times a week has the smallest room for biochemistry to shift, and the most muted response (Ichihara et al. 2015). That's a feature of the substance, not a flaw of the trial.
One-week experiments. The trials that moved markers ran a month minimum, most ran two or three. A few days of hydrogen water is not a test.

What changes, and on what timescale

The honest forecast scales with where you start. If your baseline already runs clean — young, lean, well-slept, well-trained — the change is biochemical and quiet, the kind of thing you'd only catch on bloodwork over months. If your baseline is the more common modern adult one — middle-aged, training hard, carrying a few metabolic markers in the warning zone, sleeping less than you'd like — the change is more legible.

Week one to four. If hydrogen is doing anything for you in particular, the first thing most people notice is around hard exercise. Less rubble the day after a heavy session; perceived effort at the same workload drifts down a notch. The lactate numbers behind that shift are real (Aoki et al. 2012, Botek et al. 2022); the felt experience is "the staircase that used to be a slog isn't." A few people report a small, quiet mental-clarity lift in the same window — pleasant if it happens, not the reason to take it; the cognitive evidence is thin and there is no dedicated focus trial.

Week four to twelve. The trial-replicated biomarker shifts land in this window. LDL fractions, oxidative-stress markers in urine, and the inflammatory tone of your circulating immune cells move in the right direction in the trials cited above. None of these you can directly feel — they sit under the hood — but they're the part of the case the evidence actually supports for the typical reader. Mood and stress balance shifted on POMS-style measures in a small healthy-adult trial in roughly the same window (Mizuno et al. 2017); think of that as plausible, not promised. Skin is the same story — less daily oxidative wear is the mechanism that ought to read on the face over a few months, but the trial that would prove it has not been done; treat any short-term skin effect as a small bonus, not the case for taking it.

Year and beyond. The longer-horizon claim — slower accumulation of the oxidative and inflammatory wear that drives skin ageing, vascular stiffness, and the rest of the slow-creeping disease set — is the mechanism's logical destination, and the destination the evidence has not yet reached. No mortality data, no cardiovascular-event data, no dementia data. The mechanism points the right way; the long-run trial has not been run. Plan around the things you can measure.

Adjacent threads worth pulling next: the apoB-as-cardiovascular-risk-number lens that turns shifts in LDL fractions into something more actionable; the broader "selective antioxidant" question — N-acetylcysteine, alpha-lipoic acid, the trade-offs between them; and the standing decision frame for low-evidence, low-risk supplements where the call has to be made on cost-benefit rather than on a definitive trial.

Hydrogen inhalation therapy and intravenous hydrogen-saturated saline — used in some critical-care and post-cardiac-arrest research — sit outside the consumer protocol and are not what's discussed above.

Substance and claimed effects

Molecular hydrogen (H₂) is the smallest and lightest molecule — two hydrogen atoms, neutral, lipophilic, with the highest diffusion coefficient of any gas. Consumer routes of administration are oral: hydrogen-rich water (HRW, typically 0.5–1.6 ppm dissolved H₂, ≈ 0.25–0.8 mM) produced by reactive metal sticks (magnesium- or calcium-based), electrolysis machines, or effervescent dissolving tablets that release H₂ on contact with water. Inhalation (1–4% H₂ in carrier gas) and intravenous H₂-saturated saline exist but sit in the clinical setting.

The claims made for H₂ in the consumer literature are broad: reduction of oxidative-stress markers (8-OHdG, MDA, 8-iso-prostaglandin F_2α), reduction of inflammatory cytokines (TNF-α, IL-6, IL-1β), improvement of lipid profile (LDL, apoB), improvement of glycaemic control (fasting glucose, HbA_1c), faster recovery and reduced perceived exertion in exercise, reduced fatigue and improved mood in chronic-fatigue cohorts, neuroprotection in Parkinson's disease, and skin/anti-ageing effects via reduced cutaneous oxidative load. The entry's scope is the substance and the meaningful consequences in those domains: oxidative stress, inflammation, metabolic markers, exercise recovery, energy/fatigue, mood, cognition, and skin/longevity.

Evidence by addressing question

Mechanism

Science. The seminal observation comes from Ohsawa et al. 2007: H₂ reacts selectively with the most cytotoxic reactive oxygen species — hydroxyl radical (•OH) and peroxynitrite (ONOO⁻) — and is essentially inert towards superoxide (O₂^•−), nitric oxide (•NO) and hydrogen peroxide (H₂O₂) at physiological concentrations. This is the molecule's central pharmacological claim and the reason it differs from conventional antioxidants (vitamin E, vitamin C, N-acetylcysteine): it does not blunt the redox signalling — H₂O₂ and •NO — that healthy mitochondria, immune cells and vasculature depend on, while still quenching the indiscriminately damaging radicals that drive lipid peroxidation, protein nitration and DNA oxidation.

Mechanism (downstream). Beyond direct radical scavenging, in vitro and rodent work converges on indirect signalling effects: induction of the Nrf2/Keap1 antioxidant programme (HO-1, NQO1, γ-GCS, GSH), suppression of NF-κB and downstream pro-inflammatory transcription (TNF-α, IL-6, IL-1β, ICAM-1), modulation of apoptotic signalling (Bcl-2/Bax, caspase-3) and an effect on lipid metabolism via FGF21 induction in mouse models (Ichihara et al. 2015, Ohta 2015, LeBaron et al. 2019). The current consensus interpretation is that direct •OH/ONOO⁻ quenching cannot quantitatively account for the observed downstream cellular effects — the in vivo H₂ burden is small relative to total ROS flux — and that H₂ additionally acts as a gaseous signalling molecule, plausibly modifying free-radical chain reactions in membrane lipids and modulating Ca²⁺/MAPK/NF-κB cascades upstream of the transcriptional response.

Pharmacokinetics. Oral H₂ from water or tablets produces a brief peak in expired-breath H₂ within 10–15 minutes, decaying to baseline within 60–90 minutes. Tissue retention is necessarily short; most of an ingested dose is exhaled. This pharmacokinetic profile — transient exposure, no accumulation, no metabolite — is the strongest argument against a pure stoichiometric scavenging mechanism and the strongest argument for a signalling/hormetic interpretation.

Evidence

Science — metabolic syndrome and lipids. The most-replicated human signal is in oxidative-stress and lipid markers in metabolic-syndrome or pre-diabetic cohorts. Nakao et al. 2010: open-label, 20 subjects with metabolic-syndrome features, 8 weeks of HRW (~1.5–2 L/day, 0.55–0.65 mM H₂) — urinary 8-iso-prostaglandin F_2α down 43%, total cholesterol down 5.5%, HDL function improved. Song et al. 2013: 20 patients with potential metabolic syndrome, 10 weeks of HRW (~0.9–1.0 L/day) — LDL-C down ~8%, apoB down, HDL anti-inflammatory function (paraoxonase-1 activity) improved. Kajiyama et al. 2008: double-blind crossover RCT, 30 patients with type-2 diabetes or impaired glucose tolerance, 8 weeks of HRW (900 mL/day) — modified LDL (electronegative LDL) down significantly, urinary 8-isoprostane down, fasting glucose improvements limited to the IGT subgroup. Effect sizes are real but modest; sample sizes are small.

Science — exercise recovery. Aoki et al. 2012: 10 elite male soccer players, crossover, HRW vs placebo before an interval-cycling protocol — blood lactate after exercise lower in the HRW arm (~5.9 vs 7.6 mmol/L), peak-torque drop during fatiguing knee extension reduced. Botek et al. 2022: HRW pre-loading altered ventilatory response and RPE during graded exercise in trained men. Ostojic 2014 reviews the early sports-medicine literature; Korovljev et al. 2018 reports a 4-week HRW intervention in middle-aged overweight women showing modest body-composition and metabolic shifts.

Science — neurological. Yoritaka et al. 2013: double-blind placebo-controlled trial, n=18 (9 vs 9), early Parkinson's disease on stable levodopa, 48 weeks of 1 L/day HRW — Unified Parkinson's Disease Rating Scale (UPDRS) total score improved in the H₂ arm and worsened in placebo (median ΔUPDRS −5.7 vs +4.1). Pilot scale, but a positive result on a hard clinical outcome.

Science — mood/QoL. Mizuno et al. 2017: 26 healthy adults, 4 weeks of HRW (600 mL/day, ~3 ppm) — improvements in POMS mood scores and sympathovagal balance measures.

Science — inflammation in healthy adults. Sim et al. 2020: double-blind controlled trial, 4 weeks HRW in healthy adults aged 20–59 — reduced apoptosis of peripheral blood mononuclear cells, downregulation of CD14 monocyte inflammatory markers.

Synthesis. The bulk of the human literature is small RCTs and pilot studies (n typically 10–60), heavily concentrated in Japanese, Chinese and Serbian research groups. Ichihara et al. 2015 review 321 original articles spanning preclinical and clinical work, with effects reported in 31 disease/condition categories; subsequent narrative reviews (LeBaron et al. 2019, Ohta 2015) reach similar breadth. The effect-size pattern is consistent: real signals on biochemical and biomarker outcomes (oxidative-stress markers, lipid sub-fractions, inflammatory cytokines), more modest effects on clinical endpoints, mostly in populations with elevated baseline oxidative stress.

Protocol

Dose protocols across positive trials cluster on 0.5–1.6 ppm dissolved H₂ in 0.6–2 L/day of water for 4–12 weeks (Nakao et al. 2010, Kajiyama et al. 2008, Song et al. 2013, Yoritaka et al. 2013). Effervescent tablets dissolved in 250–500 mL water immediately before drinking are the most reproducible consumer delivery: a typical tablet releases 5–10 mg H₂ over ~1–2 minutes, producing 2–8 ppm at peak in a closed container, drunk before the gas escapes. Electrolysis machines and metal-stick infusers are more variable; many "hydrogen water" bottles sold in the consumer market produce H₂ concentrations far below the threshold used in positive trials. Timing: within 10–15 minutes of preparation, on an empty or near-empty stomach if convenient; for exercise studies, dosing 30–60 minutes pre-exercise.

No tissue accumulation, no metabolite to monitor, no titration: the dose is what dissolves in the water plus what you drink before it outgasses. The protocol that fails is the one where the water sits in an open glass on the counter.

Contraindications

H₂ has an exceptionally clean safety profile in the human literature: across decades of trials and substantial inhalation work in critical-care settings, no serious adverse events have been attributed to H₂ at doses used (Ichihara et al. 2015, LeBaron et al. 2019). H₂ is not metabolised by mammalian tissue (no hepatic load), is not excreted renally, and exits via expired breath. The closed-vocabulary contraindication list does not have a matching condition; the truthful answer is "essentially none known." Reasonable theoretical cautions: the magnesium-based reactive sticks raise water-borne Mg²⁺, which could compound prescribed magnesium in renal-impaired patients — but that is a Mg issue, not an H₂ issue. Inhalation delivery (consumer "hydrogen inhalers") is out of scope; the explosive concentration of H₂ in air is 4%, and self-administered inhalation devices vary in safety engineering.

Misconceptions

Three persistent confusions in the consumer space. (1) "Alkaline water" is hydrogen water. Alkaline ionised water from an electrolysis machine often contains some dissolved H₂, but the alkalinity (pH 8.5–9.5) is not the active ingredient — the H₂ is — and pH-only bottled "alkaline water" contains no H₂. (2) H₂ is hydrogen peroxide. H₂O₂ is a pro-oxidant signalling molecule; H₂ is its mechanistic opposite. (3) The dose in any "hydrogen water" bottle is sufficient. Many commercial bottled "hydrogen waters" have outgassed to near-zero concentration by the time they're consumed; the trial-relevant H₂ dose is what's actually dissolved at the moment of drinking, not what was generated at the factory weeks ago.

Failure-modes

Three recurring causes of "I tried it and it didn't work." (1) Open-container loss. Once H₂ is in water, it leaves — half-life in open water is on the order of 1–2 hours. Drink within minutes of preparation. (2) Subtherapeutic delivery. Cheap "hydrogen water sticks" and many bottled products produce <0.1 ppm H₂, an order of magnitude below the trial range. (3) Wrong population. The strongest effects sit in cohorts with elevated baseline oxidative stress — metabolic syndrome, type-2 diabetes, hard exercise, smoking, ageing. Healthy young low-stress users have the smallest room to improve and the most muted biochemical response (Ichihara et al. 2015).

Practicalities

Cost-and-friction map of the realistic delivery options. Effervescent tablets: ~$30–80 per month at one tablet/day; per-dose H₂ reasonably reproducible if used per package directions (closed container, drink within 1–2 minutes). Magnesium-stick infusers: ~$30 one-off, but variable H₂ output, often subtherapeutic. Electrolysis countertop machines: $200–2 000+ capital cost, higher and steadier H₂ output, ongoing electricity and filter cost. Pre-filled "hydrogen water" bottles: highly variable; many have outgassed before purchase. Regulatory status: not approved by the FDA, EMA or PMDA as a drug for any indication; sold as a food/beverage or supplement. No clinical guideline body (USPSTF, AHA/ACC, AASM, NICE) currently recommends it.

History

Diving medicine showed in the 1970s–1990s that H₂ can be breathed at hyperbaric pressures (hydreliox) without acute toxicity, and isolated reports from the 1970s suggested hyperbaric H₂ regression of squamous-cell carcinoma in mice. The field as currently understood was opened by Ohsawa et al. 2007 in Nature Medicine, which established the selective-radical-scavenging frame and triggered a wave of preclinical and clinical work, predominantly out of Japanese institutions, through the 2010s. Consumer adoption followed in Japan (hydrogen water bars, tablet products) and spread to East Asia and Europe before the US biohacker market picked it up in the late 2010s.

Payoff

Short-horizon biomarker shifts (4–10 weeks): reductions in urinary 8-iso-PGF_2α and 8-OHdG, reductions in MDA, LDL-C and apoB shifts of single-digit percentages, modulation of HDL function, small reductions in TNF-α and IL-6 (Nakao et al. 2010, Song et al. 2013, Kajiyama et al. 2008, Sim et al. 2020). Felt: lower perceived exertion and faster perceived recovery from hard sessions (Aoki et al. 2012, Botek et al. 2022); mood/QoL shifts in chronic-fatigue and otherwise-healthy adults (Mizuno et al. 2017). Long-horizon clinical endpoints (mortality, MACE, cancer incidence) are not established — no powered trials.

Out-of-scope

Hydrogen inhalation therapy in critical care (sepsis, post-cardiac-arrest), hyperbaric hydrogen, intravenous H₂-saturated saline, animal-only oncology work, and hydrogen-rich saline as a perioperative intervention sit outside this entry's consumer-protocol scope. Adjacent reader interests worth signposting: other selective antioxidants (NAC, alpha-lipoic acid), the lipid sub-fraction question (apoB), and the broader "modest evidence, low risk" supplement decision frame.

The credibility range

Optimist case

The optimist reads Ohsawa 2007 as a genuine pharmacological discovery: a molecule with a mechanistically unique antioxidant profile (selective for the radicals you want to quench, inert to the ones you want to keep) and downstream signalling effects (Nrf2 induction, NF-κB suppression) that explain biomarker shifts disproportionate to the molar dose. They point to (a) replication across many independent groups and indications (Ichihara et al. 2015: 321 papers), (b) hard-endpoint signals on UPDRS in Parkinson's (Yoritaka 2013), (c) consistent lipid- and oxidative-marker shifts across independent metabolic-syndrome cohorts (Nakao 2010, Song 2013, Kajiyama 2008), (d) a near-perfect safety record. With cost in the $30–80/month range and zero metabolic load, the cost-benefit asymmetry is favourable even on modest effect sizes.

Skeptic case

The skeptic notes (a) trial sizes are small (n=10–60) and most replications come from a tight network of Japanese, Chinese and Serbian groups with overlapping authorship and industry funding ties to H₂ water device makers, (b) the pharmacokinetics — exhalation half-life under an hour, no tissue accumulation — make stoichiometric scavenging implausible at the in vivo concentrations achieved, so the entire mechanism rests on a less-direct signalling story, (c) the field-wide pattern is many small positive trials with no large definitive RCT and no Cochrane-level meta-analysis, the same shape that flatters many supplements which later fail definitive replication, (d) effect sizes on biomarkers translate poorly to disease endpoints; the only hard clinical endpoint with a positive signal (Parkinson's UPDRS) comes from n=18 and remains unreplicated at scale, (e) publication-bias risk is elevated in a niche field with commercial stakeholders, and (f) the Parkinson's signal disappeared in a subsequent larger trial by the same group (Yoritaka 2016, the negative one most consumer marketing omits).

Author's call

The mechanism is real and unusually elegant; the biomarker replications are real and consistent in the populations where they have been studied; the clinical-endpoint case is not yet made. The honest landing is: a low-risk, low-cost intervention with consistent small-to-moderate effects on oxidative-stress and inflammation biomarkers in stressed populations (metabolic syndrome, hard exercise, age-related oxidative load), promising in neurological indications but not established, modest in healthy young low-stress users. Evidence is best described as preliminary but coherent — meaningfully past speculation, not yet at guideline-quality. Meta scores reflect that: modest non-zero benefits across several dimensions, evidence rated 2 (sparse RCTs, plausible mechanism, worth trying with caveats), controversy 2 (minor but real — mechanism debated, large trials missing, commercial-incentive concerns), action do, cadence daily.

Stakeholder and incentive map

Commercial: Manufacturers of effervescent tablets, electrolysis machines, magnesium sticks, and "hydrogen water" bottled products. A material share of trial funding has historically come from device makers (disclosed in most of the trials cited above). Substantial consumer marketing investment, particularly in Japan, South Korea, China and increasingly North America.
Academic: Concentrated in Japanese institutions (Nippon Medical School, Kyushu) and a handful of Serbian, Chinese and Korean labs. The field has its own journal (Medical Gas Research) and conference circuit, which raises both productivity and within-field replication concerns.
Clinical: No major guideline body recommends H₂ for any indication. Sports-medicine practitioners and a slice of integrative-medicine physicians use it; mainstream cardiology, endocrinology and neurology do not.
Community: Active biohacker and longevity communities, podcast adoption, and Japanese consumer habit. Strong enthusiast signal; usual survivorship-bias caveats apply.
Counter-incentive: Regulatory caution from FDA/EMA (no approved indication), mainstream skeptic-medicine commentary noting the small-trial / industry-funding pattern.

Population variability

Baseline oxidative load matters most. Effect sizes are larger in populations with elevated baseline ROS: metabolic syndrome, type-2 diabetes and impaired glucose tolerance (Kajiyama 2008), hard-training athletes (Aoki 2012, Botek 2022), middle-aged overweight adults (Korovljev 2018), early Parkinson's (Yoritaka 2013). Healthy young low-stress users show smaller biochemical shifts.
Sex. Most trials enrol predominantly men or report mixed cohorts without sex-stratified analysis. Korovljev 2018 is one of the few women-only trials. No clear evidence of sex-specific response patterns, but the data is thin.
Age. Stronger signals in middle-aged and older cohorts (where baseline oxidative load is higher) than in healthy young adults.
Pregnancy/lactation. No trials in pregnant women. The mechanism gives no plausible reason for concern (H₂ is endogenously produced by gut microbiota at gram quantities daily), but the conservative call is "no data, no recommendation either way."
Genetic variability. Endogenous H₂ production from colonic microbiota varies several-fold between individuals; this is a plausible source of inter-individual response variation that has not been systematically studied.

Knowledge gaps

The single large RCT. The field has no Cochrane-level meta-analysis and no large definitive trial on any single endpoint. The most useful single piece of evidence would be a properly powered (n≥300) double-blind placebo-controlled trial of HRW or H₂ tablets in metabolic-syndrome patients, with apoB or HbA_1c as a primary endpoint.
Mechanism quantitation. Whether the downstream Nrf2/NF-κB effects can be reproduced with non-H₂ selective antioxidants of similar pharmacokinetics. If they can, the H₂-as-signalling-molecule hypothesis becomes much harder to dispute.
Dose-response. No formal dose-response trial across consumer-relevant H₂ concentrations.
Long-horizon endpoints. No mortality, MACE, dementia, or cancer-incidence data in humans.
Independent replication of the Parkinson's signal at scale. The positive 2013 pilot was followed by a larger negative trial in 2016 from the same group; the discrepancy is unresolved and would be the highest-leverage trial to redo.
Delivery comparison. Head-to-head trials of tablets vs electrolysis vs metal-stick water at matched dissolved-H₂ concentrations have not been done.

Scope vs the brief. The brief named oxidative stress, inflammation, recovery, metabolism, and cognition. The article covers oxidative stress (mechanism, evidence, payoff), inflammation (mechanism, evidence, payoff), recovery (evidence, payoff), metabolism (evidence: LDL fractions, glucose, lipid sub-fractions in Nakao/Kajiyama/Song trials), and brushes cognition — there is no dedicated cognitive RCT, so I named it as plausible-but-thin in payoff rather than claiming a focus effect. Meta scores focus at 1 to reflect this honestly.

Hard scoping calls.

Skipped the dream narrative deliberately. Score lands at roughly 28 (well below the 40 obligatory floor), and the honest hook is calibration — modest evidence, low risk, the unusually elegant mechanism — not an aspirational life-cascade. Forcing one would have produced exactly the overclaim the substance does not earn.
No stakes section. This is a do entry with a modest case; a felt-experience forecast of life-without-hydrogen-water would read as wellness-influencer voice. The non-existence of the long-horizon trial is named in payoff instead, where it belongs.
Treated hydrogen inhalation as out-of-scope. The consumer-protocol surface here is HRW and tablets; inhalation belongs in a separate clinical entry if it ever earns one.
The Parkinson's signal is named in evidence with the non-replication explicitly. Tempting to omit (it's the most striking single result), but omitting it would be the device-marketing failure mode the article should not commit.

Rating difficulties.

evidence: 2 over 3 was the closest call. Many small RCTs with consistent direction would justify a 3 in isolation; the network-concentration / industry-funding pattern and the absence of any large independent replication pull it back to 2.
controversy: 2 rather than 3 — the field is not in a paradigm fight, but the mechanism is genuinely debated (signalling vs scavenging) and the industry-funding concentration is real. A 1 would have understated it.
applicability: 4 reflects broad adult relevance via metabolic / exercise / oxidative-load applicability. Pulled back from 5 because healthy young low-stress users genuinely have small expected benefit.
beauty_direct: 1 and focus: 1 are both honest-floor scores — the mechanism is suggestive, the substance is across the rest of the body so there has to be some effect, the trial evidence at consumer dose is just not there. Could defend 0 on either; I went with 1 to keep the article's coverage requirement honest.
sleep: 0 — no sleep data in the consumer-dose literature worth citing. Did not invent a score.

Separate-entry candidates surfaced by the writing.

apoB as the cardiovascular risk number — referenced in out-of-scope; this is a discrete topic with its own protocol (test, treat to target).
Nrf2 pathway activators as a category (sulforaphane, exercise, fasting, H₂) — could be a useful umbrella entry if the catalogue grows several of these.
Selective vs non-selective antioxidants as a decision frame — could land as a mindset/decide entry rather than a substance entry.
Hydrogen inhalation therapy — flagged out of scope here, plausibly its own entry under medical/healthcare if the evidence base in critical care matures.

Future links to wire in when the targets exist. The out-of-scope section currently points to apoB, the selective-antioxidant question, and the low-evidence-low-risk supplement decision frame in prose; once corresponding entries land, those mentions become candidates for related.