"It works" and "it doesn't work" are both too crude for the state of this record. Here is what exists, level by level, and what is missing.
334 healthy adult males; rising single-dose tolerance for chlorine dioxide (0.1–24.0 ppm) and chlorite (0.01–12.0 ppm); a 12-week chronic ingestion phase at 5 mg/L; and a chronic phase in G6PD-deficient subjects — the population most susceptible to oxidative haemolysis. The US EPA's review records a human NOAEL of 0.036 mg/kg/day from this work. Lubbers, Chauhan, Miller & Bianchine, Fundam Appl Toxicol 1981;1(4):334–338 copy.
Weight: strong for tolerability at the doses tested. It says nothing about efficacy.
1,136 patients with PCR- and tomography-confirmed COVID-19, mean dose 1.41 mg/kg: 99.07% discharged without complications (1,057 of 1,067), mean symptom resolution 4.84 days, adverse reactions in 6.78% (all mild), and normalisation of liver enzymes, glucose, cholesterol and triglycerides. No comparator arm. No confidence interval is reported in the publication. Aparicio-Alonso, Domínguez-Sánchez & Banuet-Martínez, J Infect Dis Ther 2021;9:477 copy.
A companion prophylaxis cohort of 1,163 household contacts reported 90.4% without COVID-like symptoms on 0.3 mg/kg/day in ten daily intakes. No control group.
Weight: moderate for safety at the measured dose; insufficient for causal efficacy. Two things have to be weighted correctly. First, every patient was enrolled while symptomatic, not merely PCR-positive, which excludes the objection that healthy people were treated. Second, in outpatients mortality is the wrong endpoint: almost anything passes it. The informative figure in this cohort is time to symptom resolution — 4.84 days on average. For comparison, the US CDC reports that, among symptomatic outpatients, 35% had still not returned to their usual state of health 14 to 21 days after testing (MMWR 2020;69:993–998). Without a control arm, and with differing endpoint definitions, that is a hypothesis-generating comparison rather than proof of efficacy — but it is the number a controlled trial would have to address.
104 patients (Ecuador, 2020; 30 ppm in one litre, ten hourly doses); 20 patients (Bolivia, Peru, Ecuador, 2021); registered protocol NCT04343742. Weight: weak to moderate; no randomisation, small samples.
Six metastatic cancer patients, Authorea preprint 2023 (PPR666536). Not peer reviewed — flagged as such here.
Size-selective oxidation (Noszticzius et al., PLoS ONE 2013 copy); 99.99% inactivation of SARS-CoV-2 within ten seconds at 24 ppm (Hatanaka et al., J Hosp Infect 2021 copy). Weight: coherence, not clinical proof.
Randomised trials exist for chlorine dioxide mouthrinses (oral malodour, salivary viral titre) and topical preparations. They establish topical efficacy and say nothing about ingested systemic use.
Kalcker AL. The Mechanism of Action of Chlorine Dioxide (CDS): Electron-Transfer Chemistry, Antimicrobial Selectivity and Redox-Regulatory Effects: A Narrative Review. Biophysica 2026;6(5):91. doi:10.3390/biophysica6050091 copy (received 1 Aug 2026; accepted 4 Sep 2026; published 21 Sep 2026).
It is a narrative review, not primary data — and the page says so, because that distinction is the whole point of this site. Its comparative redox table is nonetheless the most useful single table on the subject:
| Oxidant | E°′ (approx.) | Reaction type | Relative selectivity |
|---|---|---|---|
| Hydroxyl radical (·OH) | +2.31 to +2.80 V | non-selective | very low |
| Ozone (O3) | +2.07 V | radical | low |
| Hydrogen peroxide (H2O2) | +1.78 V | two-electron | medium |
| Hypochlorous acid (HOCl) | +1.48 V | chlorination + oxidation | low–medium |
| Hypobromous acid (HOBr) | +1.33 V | bromination + oxidation | low |
| Chlorine dioxide (ClO2) | +0.95 V | one-electron, no chlorination | high |
The paper places chlorine dioxide within the physiological window of cellular oxygen (O2/H2O, +0.82 V) and superoxide (O2•−/H2O2, +0.94 V), above the disinfectant inhibitory threshold of about +650 mV and far below hydrogen peroxide and the hydroxyl radical. It cites Buettner (1993), standard electrochemical tables and Knight & Fildes (1930) for the values.
What the paper concedes, quoted because a review that concedes nothing is worthless. Its abstract states that most toxicological data derive from sodium chlorite rather than dissolved-gas ClO2, and that chlorite is associated with documented developmental toxicity in animals and regulatory caution by the EPA and WHO. It also acknowledges that direct quantitative confirmation of host-cell sparing in vivo is not yet available.
Ma J-W, Huang B-S, Hsu C-W, Peng C-W, Cheng M-L, Kao J-Y, Way T-D, Yin H-C, Wang S-S. copy Efficacy and Safety Evaluation of a Chlorine Dioxide Solution. Int J Environ Res Public Health 2017;14(3):329. doi:10.3390/ijerph14030329 · PMID 28327506 · PMCID PMC5369164.
The product tested (UC-1) was made electrolytically and membrane-purified, at 2,000 ppm of dissolved chlorine dioxide. The study reports:
| Test | Result as published |
|---|---|
| Antimicrobial activity | >98.2% reduction at 5 ppm (bacteria) and 20 ppm (fungi) |
| Antiviral IC50 | H1N1 84.65 ± 0.64 ppm; influenza B 95.91 ± 11.61 ppm; EV71 46.39 ± 1.97 ppm |
| Cytotoxicity (MTT, mouse lung fibroblasts L929) | 93.7% cell viability at 200 ppm |
| Rabbit ocular irritation | 50 ppm: no significant symptoms |
| Inhalation toxicity | 20 ppm for 24 h: no abnormality, no mortality |
| Subchronic oral toxicity | Up to 40 ppm in drinking water: no toxicity reported |
Conflict of interest, disclosed because the same standard is applied to others. The authors are affiliated with Unique Biotech Co., Ltd. (Kaohsiung, Taiwan) — the manufacturer of the product tested — alongside university departments. This is industry-funded research on the company's own product, and it should be read with the same caution this register applies to pharmaceutical-funded studies.
What it is and is not. It is animal and in vitro data, not a clinical trial. It establishes that at the concentrations tested — including 40 ppm in drinking water over a subchronic period — no toxicity was observed in that model. It does not establish safety in humans at any dose, and it does not establish efficacy against any disease.
The strongest animal signal against chlorine dioxide is the thyroid one, from Bercz et al. (1982). The objection that matters is palatability: water at these concentrations is irritant, so did the animals simply drink less? The paper answers it — the authors measured consumption three times a week — and its answer deserves to be quoted rather than assumed:
| Concentration of ClO2 in drinking water | Water consumption | What happened |
|---|---|---|
| 0 mg/L (control) | ≈125 mL/kg/day | — |
| 100 mg/L | ≈95 mL/kg/day | T4 depression, dose- and consumption-related; stage completed |
| 200 mg/L | ≈55 mL/kg/day | Stage terminated after one week because of signs of oral injury |
On the 100 mg/L stage the authors write: «Analysis of the water consumption data at the 100 mg/l. exposure disclosed that the T-4 depression in each animal was statistically related to water consumption and ClO2 dose. Since the animals' fluid intake was near normal, the effect cannot be explained by dehydration.»
What this settles and what it does not. It settles the crude version of the objection: the finding that carries the weight is at 100 mg/L, where intake was substantially higher than at 200 mg/L and the authors tested the dehydration explanation directly. And it demolishes a claim that circulates in the other direction: the 200 mg/L figures must never be cited as tolerated intake — that stage was stopped after one week because the animals were injured.
The residual objection, which is fair. A fall from ≈125 to ≈95 mL/kg/day is roughly a quarter, and «near normal» is the authors' characterisation, not a statistical demonstration. Without a pair-water control — animals given the same reduced volume of clean water — a contribution from reduced intake cannot be excluded with certainty. The effect is therefore weakened as a quantitative estimate, not invalidated. The study is not the fraud the crude objection implies, and its thyroid signal is not the clean proof the other side pretends either.
A natural objection is that oral ulceration points to chlorite or to an acidified mixture — the MMS route — rather than to dissolved chlorine dioxide. In this study it does not hold, and the reason matters for the whole debate: the drinking solutions were made by purging chlorine dioxide from an acidified-chlorite generator through a solid chlorite column into distilled water — the acid stays in the generator and only the gas reaches the water. That is the same production principle used to make CDS. The authors verified absence of free chlorine and hypochlorite (AgNO₃ test for chloride) and assayed concentration and purity by UV spectroscopy and Palin titrimetry, before dosing and again at each refill, to control hydrolytic and photolytic degradation.
What the paper records, in its own words:
«During the 200 mg/l. exposure, erythema and ulceration of the oral mucosa, mucous nasal discharge and avoidance of drinking water by the animals was observed.»
«During the higher concentration stages of this study (100 and 200 mg/l.) the mean daily dose remained nearly constant at ca. 9 mg/kg/day. This observation reflects on the strong irritating nature of ClO2 solutions.»
So the lesions appeared at 200 mg/L — one of the two higher stages — and were attributed by the authors to chlorine dioxide's own irritancy, not to acidity or to chlorite. For scale: the CDS protocol uses a stock of 3,000 mg/L diluted for intake to roughly 30 mg/L, so 200 mg/L is about 6.7 times the concentration a patient drinks, and it was consumed continuously as the only water source.
The concession that belongs on the other side of the ledger. The same paper reports no thyroid effect in human volunteers who ingested about 1 mg/L of chlorine dioxide in drinking water through routine community water treatment. So the study that supplies the strongest animal signal also supplies a human observation at drinking-water concentrations, and it is negative.
Primary source now read in full: Bercz JP, Jones L, Garner L, Murray D, Ludwig DA, Boston J. Subchronic toxicity of chlorine dioxide and related compounds in drinking water in the nonhuman primate. Environ Health Perspect 1982;46:47–55 (PMC1569048). ATSDR TP-160 copy also records erythema and ulceration of the oral mucosa in the same animals.
Most of the apparent contradictions in this literature dissolve when the concentration is written next to the finding. All the figures below come from the studies cited on this page; the axis is milligrams of chlorine dioxide per litre.
| Concentration | Setting | Finding | Source |
|---|---|---|---|
| ≈1 mg/L | Human volunteers, community drinking water | No thyroid effect detected | Bercz et al. 1982 (the study's own human data) |
| ≈30 mg/L | CDS protocol as drunk (3,000 mg/L stock, diluted) | No mucosal injury reported in the published clinical record | Protocol; cohort reports |
| 40 mg/L | Rat, subchronic, drinking water | No toxicity reported | Ma et al. 2017 |
| 50 mg/L | Rabbit, ocular | No significant irritation | Ma et al. 2017 |
| 100 mg/L | Monkey, drinking water, subchronic | T4 depression, related to dose and consumption | Bercz et al. 1982 |
| 200 mg/L | Monkey, drinking water | Erythema and ulceration of oral mucosa; stage stopped after one week | Bercz et al. 1982 |
What this table resolves. There is no contradiction between «no toxicity at 40 mg/L» and «mucosal injury»: they are two to five times apart on the same axis, in different species, with different dosing durations. The Ma study never tested 100 or 200 mg/L; the Bercz study never tested 40 mg/L. And nothing in the literature tests the interval between 40 and 100 mg/L — which is precisely where the argument lives.
Why oral use at low concentration works in practice and does not contradict the animal data. Chlorine dioxide at low concentration is used topically in the mouth with randomised evidence for plaque, gingivitis and oral malodour, and it is reported to accelerate healing of oral lesions in patients with acute myeloid leukaemia. A locally applied, brief contact at low concentration is a different exposure from drinking 200 mg/L continuously as the only source of water.
Topical oral evidence: Siddeshappa ST et al., comparative antiplaque/antigingivitis study of an herbal and a chlorine dioxide mouthwash (PMID 29442084); systematic review and meta-analysis of chlorine dioxide on oral hygiene (PMC8383470); reported acceleration of healing of oral lesions in acute myeloblastic leukaemia. No randomised trial of chlorine dioxide specifically for recurrent aphthous stomatitis was located.
The honest one-sentence summary: the clinical record is consistent in direction across independent practitioners and biologically plausible, and it is uniformly uncontrolled — which makes it a signal and not a proof.