The most serious clinical setting hydrogen has been tested in is not a wellness context at all. It is the intensive care unit, in patients who have been resuscitated after their heart stopped outside hospital. A multi-centre randomized trial called HYBRID II tested inhaled hydrogen gas in exactly that population[1].
The trial missed its primary endpoint. Its secondary outcomes, including 90-day survival, favoured hydrogen. Understanding why those two sentences can both be true, and what weight to give each, is the point of this page.
One thing to establish immediately: this research has nothing to do with any consumer hydrogen product. It tested medical-grade hydrogen gas delivered by ventilator in intensive care.
The clinical problem
When the heart stops, the brain loses blood flow. When circulation is restored, the returning blood carries oxygen into tissue that has been starved of it, and that reperfusion generates a burst of reactive oxygen species which causes further injury. The resulting condition is called post-cardiac arrest syndrome, and brain injury is the main determinant of whether a patient who survives to hospital recovers a life they would recognise.
The theoretical case for hydrogen here is more specific than in most of this literature. Reperfusion injury is a defined burst of oxidative damage at a known moment, in a patient already in intensive care where a gas can be administered under full control. If a selective antioxidant helps anywhere, this is the kind of setting where it should be detectable.
The first-in-human study
Before the randomized trial, a pilot study established feasibility[2]. Between January 2014 and January 2015, of 107 patients with out-of-hospital cardiac arrest, 21 achieved spontaneous return of circulation. After exclusions, 5 patients underwent hydrogen inhalation together with target temperature management.
No undesirable effects attributable to hydrogen inhalation were observed, and 4 of the 5 patients survived 90 days with a favourable neurological outcome[2].
Five patients cannot tell you whether a treatment works. What a study like this establishes is that hydrogen can be delivered safely alongside cooling in a critically ill patient, which is the precondition for running a real trial. The 4-of-5 survival figure is not an efficacy signal and the authors did not present it as one.
HYBRID II
HYBRID II was a prospective, multi-centre, randomized, double-blind, placebo-controlled trial run at 15 hospitals in Japan between February 2017 and September 2021[1]. Patients aged 20 to 80 in a coma following cardiogenic out-of-hospital cardiac arrest were randomly assigned one-to-one, using blinded gas cylinders, to supplementary oxygen with 2 percent hydrogen or to oxygen alone, for 18 hours.
Blinded gas cylinders are a genuinely elegant solution to a hard problem. Hydrogen is odourless and colourless, so identical unlabelled cylinders make it possible to blind clinicians delivering the gas, which is not achievable for most critical-care interventions.
The primary outcome was the proportion of patients with a Cerebral Performance Category of 1 or 2 at 90 days. CPC is a five-point scale where 1 means good cerebral performance and 2 means moderate disability with enough function for independent daily activities. Categories 1 and 2 together are the standard definition of a good neurological outcome in cardiac arrest research.
The trial was stopped early
Enrolment was terminated prematurely because of restrictions imposed during the COVID-19 pandemic[1]. Of 429 patients screened, 73 were randomized, 39 to hydrogen and 34 to control.
This matters more than any single result in the trial. A study is designed with a target sample size calculated to give it a reasonable chance of detecting an effect of a specified size. Stopping at 73 patients means the trial ran with less statistical power than planned, which makes a non-significant primary result much harder to interpret. It does not distinguish between “hydrogen does not work” and “the trial was too small to tell”.
The primary result
A CPC of 1 or 2 at 90 days was achieved by 22 of 39 patients (56 percent) in the hydrogen group and 13 of 34 (39 percent) in the control group, with a relative risk of 0.72, a 95 percent confidence interval from 0.46 to 1.13, and p equal to 0.15[1].
A 17 percentage point absolute difference in good neurological outcome would be a large clinical effect if it were real. At p equal to 0.15 it did not reach statistical significance, and the confidence interval crosses 1, meaning the data are compatible with a meaningful benefit and also with no benefit.
The trial did not demonstrate that hydrogen improves neurological outcome after cardiac arrest. That is the primary result and it is the correct headline.
The secondary outcomes
Several secondary endpoints did reach significance[1]. Median modified Rankin Scale score was 1 with an interquartile range of 0 to 5 in the hydrogen group, against 5 with an interquartile range of 1 to 6 in the control group, at p equal to 0.01. An mRS score of 0, meaning no symptoms at all, was achieved by 18 of 39 (46 percent) in the hydrogen group against 7 of 33 (21 percent) in the control group, at p equal to 0.03. The 90-day survival rate was 85 percent (33 of 39) with hydrogen against 61 percent (20 of 33) in the control group, at p equal to 0.02.
A 24 percentage point survival difference is not a subtle finding. But secondary outcomes in a trial that missed its primary endpoint occupy a specific and limited epistemic position. When several endpoints are tested, some will reach significance by chance, and the primary endpoint is designated in advance precisely so that the trial has one pre-committed test rather than many opportunities. Significant secondaries after a failed primary generate hypotheses. They do not establish efficacy.
The authors state this correctly. Their interpretation is that the increase in participants with good neurological outcomes was not statistically significant, and that the secondary outcomes suggest hydrogen inhalation may increase 90-day survival without neurological deficits[1]. “Suggest” and “may” are load-bearing.
The trial was funded by Taiyo Nippon Sanso Corporation, an industrial gas company[1], which is disclosed in the paper.
The post hoc temperature analysis
A subsequent analysis examined whether hydrogen’s effect depended on how cold patients were cooled[3]. Target temperature management after cardiac arrest is delivered at different target ranges, and the analysis split patients into those managed at 32 to 34 degrees Celsius and those at 35 to 36 degrees.
Among 72 patients with outcome data, the colder range was used in 25 of 39 hydrogen patients (64 percent) and 24 of 33 controls (73 percent), with p equal to 0.46[3]. Within that colder subgroup, 17 of 25 (68 percent) in the hydrogen group achieved good neurologic outcomes against 9 of 24 (38 percent) in the control group, giving a relative risk of 1.81 with a 95 percent confidence interval from 1.05 to 3.66, at p less than 0.05. In adjusted analysis, hydrogen combined with the colder range was independently associated with good neurologic outcomes, with an odds ratio of 16.10 and a 95 percent confidence interval from 1.88 to 138.17, at p equal to 0.01.
That odds ratio deserves scepticism rather than excitement. A confidence interval running from 1.88 to 138.17 is enormously wide, which is what happens when you fit an adjusted model to a subgroup of 49 patients. The point estimate of 16.10 is not a reliable estimate of anything.
And the same analysis found that hydrogen plus the colder range did not improve survival compared with the colder range alone, with an adjusted hazard ratio of 0.22, a 95 percent confidence interval from 0.05 to 1.06, and p equal to 0.06[3]. So the subgroup analysis that strengthens the neurological finding simultaneously fails to reproduce the survival finding that was the most striking secondary result of the main trial.
This is a post hoc subgroup analysis of an underpowered trial that was stopped early. That is close to the weakest form of clinical evidence that still gets published in a good journal. Its value is in generating a specific, testable hypothesis for the next trial: that hydrogen and deeper cooling may interact.
What this means for anyone reading about hydrogen products
Nothing here transfers to consumer use, and the reasons are worth being explicit about.
The intervention was 2 percent hydrogen gas inhaled continuously for 18 hours through a ventilator circuit in an intensive care unit. That is a fundamentally different exposure from dissolved hydrogen in a glass of water, both in quantity and in route.
The population was comatose patients after cardiac arrest, receiving temperature management and full critical care. The biology being targeted is an acute reperfusion injury occurring over hours.
The setting requires medical gas supply, ventilator delivery, blinded cylinders and continuous monitoring. None of this is replicable outside a hospital, and attempting to inhale hydrogen at home is not a scaled-down version of this treatment.
The bottom line
HYBRID II is the most rigorous hydrogen trial that exists, and it did not meet its primary endpoint. Good neurological outcome at 90 days was 56 percent with hydrogen against 39 percent with control, at p equal to 0.15, in a trial stopped early at 73 of a planned larger sample[1].
Its secondary outcomes, including 85 percent against 61 percent survival at 90 days, favoured hydrogen and are hypothesis-generating rather than conclusive. A post hoc analysis suggests the effect may depend on cooling depth, while failing to reproduce the survival benefit in the same subgroup[3].
The reasonable summary is that inhaled hydrogen in post-cardiac arrest care is a serious, unresolved research question that deserves a properly powered trial. It is not a demonstrated treatment, and it has no bearing on whether anyone should drink hydrogen water.
References
- Efficacy of inhaled hydrogen on neurological outcome following brain ischaemia during post-cardiac arrest care (HYBRID II): a multi-centre, randomised, double-blind, placebo-controlled trial. EClinicalMedicine, 2023
- Feasibility and Safety of Hydrogen Gas Inhalation for Post-Cardiac Arrest Syndrome: First-in-Human Pilot Study. Circulation Journal, 2016
- Combination of Hydrogen Inhalation and Hypothermic Temperature Control After Out-of-Hospital Cardiac Arrest: A Post hoc Analysis of the Efficacy of Inhaled Hydrogen on Neurologic Outcome Following Brain Ischemia During Post-Cardiac Arrest Care II Trial. Critical Care Medicine, 2024
These statements have not been evaluated by the Food and Drug Administration. This information is not intended to diagnose, treat, cure, or prevent any disease. Content is for informational purposes only and is not medical advice; consult a qualified healthcare provider before starting any supplement. As an Amazon Associate we earn from qualifying purchases.
