H₂ inhalation therapy in patients with moderate COVID-19 (H2COVID): a prospective ascending-dose phase I clinical trial
Salomez-Ihl C, Giai J, Barbado M, Paris A, Touati S, Alcaraz JP, Tanguy S, Leroy C, Lehmann A, Degano B, Gavard M, Bedouch P, Pavese P, Moreau-Gaudry A, Roustit M, Boucher F, Cinquin P, Brion JP. · Antimicrob Agents Chemother. 2024;68(8):e0057324.
What kind of evidence is this?
Human
Prospective open-label single-ascending-duration phase I study
Respiratory health
Inhaled H₂
Safety/tolerability signal; efficacy indeterminate.
Phase I safety and tolerability outcomes with exploratory clinical measures.
Methods at a glance
Hospitalized adults with moderate COVID-19 receiving standard care.
12 participants in one-, three- and planned six-day duration cohorts.
One, three or planned six days.
3.6% H₂/96.4% N₂ at 1 L/min, combined with wall oxygen adjusted to clinical need.
Inhaled H₂-containing gas.
3.6% H₂ and 96.4% nitrogen at 1 L/min.
1 L/min H₂ mixture; wall oxygen up to 6 L/min as clinically needed.
Supplemental oxygen was adjusted independently according to clinical need.
No concurrent control; ascending-duration cohorts.
Outcomes and reported result
Maximum tolerated duration, adverse events, oxygen needs and exploratory clinical and laboratory observations.
The maximum tolerated duration was at least three days; two serious adverse events were attributed to COVID-19 rather than H₂, and the design cannot establish efficacy.
The full article methods, intervention, outcomes, positive and null findings, funding and conflict statements were checked.
A reported association, difference or mechanism is not automatically a clinical benefit. Null findings, outcome type, study design and precision all matter.
Cautions and applicability
Very small, open-label, noncomparative phase I study. Gas cylinders and the delivery device were supplied by Air Products.
Hospitalized adults with moderate COVID-19 under this monitored protocol.
ROBINS-I — preliminary
Serious concerns
Appropriate for early safety exploration, but nonrandom cohorts, no comparator, small numbers and changing duration prevent efficacy inference.
Selection and allocation checked · Confounding and co-interventions remain possible · Missing observations checked where reported · Outcome measurement varies by endpoint · Selective reporting cannot be excluded
No single-study GRADE certainty rating is assigned. Certainty is assessed by important outcome across a complete eligible evidence set, not by attaching a final grade to one source.
Sources and record status
PMID: 39016593 · DOI: 10.1128/aac.00573-24
A free publisher or repository full article was checked.
Publisher or repository full article and bibliographic record; checked 10 August 2026.
2026-08-10