Nitrous Oxide

Nitrous oxide is the only substance in this batch whose sought-after effect and feared harm have almost nothing to do with one another. The effect lasts a minute and rests on a measured NMDA antagonism. The harm is a slow chemical reaction, unconnected to psychoactivity: the gas oxidises the cobalt of vitamin B12 and puts an enzyme of one-carbon metabolism out of service. Nothing warns at the moment of taking it, the first alert arrives weeks later in the form of pins and needles, and the blood test everyone thinks of for reassurance is the wrong one.

The literature is very unbalanced, and this has to be said at the outset. It is solid on the biochemistry of B12 and on myeloneuropathy, with real hospital cohorts. It is thin on almost everything else: the phenomenology rests on a handful of experiments from the 1990s and on self-selected online surveys, no human functional imaging under recreational use was found, dependence potential is not settled, and there is no established toxicity threshold.

Neurochemical action

The only source consulted that puts figures on effects at isolated receptors is Yamakura and Harris, on recombinant receptors expressed in Xenopus oocytes, with gas at 0.58 atmospheres. NMDA receptors are inhibited by 31%, the most sensitive of the glutamate receptors. Nicotinic receptors carrying a beta-2 subunit are inhibited more, 39% for the alpha-4 beta-2 assembly against only 7% for alpha-4 beta-4. 5-HT3 receptors are slightly inhibited, with no figure. Glycine and GABA-A are potentiated, far less than by isoflurane, GABA-C inhibited. Inhibition of NMDA and of nicotinic receptors is non-competitive. A two-P-domain potassium channel, TREK-1, is clearly activated: several secondary sources write inhibition of TREK-1, and that is the opposite of the published result.

What this study did not test counts just as much. The protocol covered only ligand-gated ion channels. The serotonin transporter and monoamine oxidase were not studied, neither there nor elsewhere. One must therefore write not tested, never no documented effect: an absence of measurement is not a negative result. In the rat at 75%, no change in brain serotonin or 5-HIAA content, with the reservation that these are regional tissue contents and not a measure of synaptic release. What does move is dopaminergic.

Analgesia goes through a separate pathway, documented by blockade: it is abolished by naloxone in the mouse, the rat and the human, and by opioid antagonists microinjected into the periaqueductal grey matter, whereas the intrathecal route has no effect. Cross-tolerance with morphine is one-way, animals tolerant to morphine are tolerant to the gas and the reverse is false, and that very asymmetry argues for a release of endogenous opioids rather than for a direct action on the receptors. Anxiolysis, for its part, has a benzodiazepine-type profile, blocked by flumazenil.

The gas is therefore, like ketamine and like dextromethorphan, an NMDA antagonist. One has to resist drawing a warning from that: no source proposes combining nitrous oxide with dextromethorphan as a risk, and the only available datum points the other way, Jevtović-Todorović et al. explaining the good safety record of the gas by its usual combination with GABAergic anaesthetics that counter its neurotoxic potential. An unexamined question, not an established danger.

The minute the effect lasts

Two onset ranges circulate and they do not contradict each other. From a balloon, the effect sets in within about ten seconds and peaks within the 30 seconds following the last inhalation; under a mask in anaesthesia, the delay is 2 to 5 minutes. The gap comes from what is being measured, the sensation of a concentrated hit of a very poorly soluble gas, or the rise towards an equilibrium alveolar concentration under controlled flow.

The only controlled experimental study found gives the rest. Twelve healthy volunteers, randomised blind crossover design against placebo, four concentrations: peak within 30 seconds, plateau of about one minute, return close to baseline at 5 minutes. The rise in the high and stoned scales is dose-dependent, but so is the rise in anxious and that must not be smoothed away, this is not a pure euphoria. Free recall of words presented in the minute following inhalation is significantly reduced at the highest concentration. In the largest self-report corpus, 27.8% of past-year users report hallucinations and 23.9% confusion, the authors noting that these are perhaps the sought-after effects.

Clinically, about 25% is enough to reduce pain, hypnotic effects appear above 60%, unconsciousness around 70%. Elimination is almost entirely pulmonary, the expired concentration falling from 66 to 70% down to 6 to 9% in 5 minutes and 2 to 4% in 30 minutes. Subjective recovery is complete within a few minutes, but lingering behavioural effects can persist for up to 30 minutes.

The gas carries an old philosophical literature, Davy in 1800, Blood in 1874, James in 1882, of which no text has been read here. For James, two incompatible paginations circulate without any original having been consulted: we cite him without a page, or not at all.

The toxic mechanism: an oxidised cobalt

The gas oxidises the cobalt of cobalamin(I), a highly reducing form generated during the catalytic cycle, into cobalamin(II), which is unable to accept a methyl group. Methylcobalamin, the cofactor of methionine synthase, is no longer regenerated and the enzyme is inactivated. The precise destructive step remains a hypothesis proposed by its authors, on the Escherichia coli enzyme and not on the human enzyme. The cascade, for its part, is well described: the conversion of homocysteine into methionine collapses, homocysteine accumulates, folate stays trapped in methylated form, the synthesis of methionine and of myelin is compromised.

The only reliable rates come from the primary source. In the rat at 50%, an inactivation half-time for hepatic methionine synthase of 5.4 minutes. In the human, 22 patients undergoing laparotomy under 70%, mean half-time 46 minutes. Three points without which the figure misleads: it is a half-time and not a delay to complete inactivation, the enzyme measured is hepatic on biopsy and not a whole-body measure, and the two species did not receive the same concentration. The authors conclude that the metabolic consequences depend critically on the duration of exposure and are unlikely to be significant below 40 minutes. That 40 is a threshold of clinical relevance and not a rate; a much-cited review turned it into a rate, and that figure must not be repeated.

The rise in homocysteine outlasts the exposure: return to baseline between 12 and 24 hours in the adolescent, still raised at 24 hours in the adult, with elevations documented for up to a week. Since homocysteine is itself an agonist at the glutamate site of the NMDA receptor, the authors propose a window in which the excitotoxic effect is no longer counterbalanced by the antagonism of the gas. A hypothesis given as such.

What is consumed, and the absence of a threshold

A siphon charger weighs about 8 g, capped at 8.6 g per sale in France since 2023, which is about 4.1 L of gas under normal conditions: this is not a citation but a direct calculation from the molar mass. A canister holds 580 to 660 g, about 600 g according to the London series, which sets the equivalence of one canister to 75 chargers. That conversion is exactly the ratio of the masses, and so arithmetically neutral, and the uncertainty bears solely on the nominal masses declared by manufacturers. The change in packaging is the central fact of the past ten years: it is no longer the same product that is being consumed.

Ordinary use remains modest, fewer than 10 balloons per episode and fewer than 10 episodes a year for about 80% of the users in a Dutch survey. The figures from the hospital series describe something else entirely and are never reference doses: median of 318 chargers a week in London, Birmingham and Manchester, range 1 to 2,800, quantity documented in only 78 of the 119 patients; 400 to 6,000 g in a single day in Amsterdam.

There is no established safety threshold, and that is the most important point of this section. The two best sources write it: it is not known whether there exists a minimum quantity that must be consumed before the threshold of subacute combined degeneration is reached. Cases have occurred after relatively few chargers, and the authors see in this not the proof that a small quantity suffices but the sign of individual susceptibility factors. The only quantified dose-lesion correlation covers 22 patients, and it is absent for B12, homocysteine and mean corpuscular volume. That is too little to make a benchmark of.

Dependence: the question is not settled

A 2015 review writes that use does not appear to lead to dependence and that no legal measure is called for. That position is dated. A 2024 review setting the literature against the ICD-10 criteria concludes that the data are inconsistent, with indications of tolerance and of neglect of other interests, and settles on a dependence that is above all psychological, physical dependence not being convincing as things stand. In France, 90% of the subjects reported to the addictovigilance centres have a dependence disorder, and 59% of the 2024 cases are daily users. The gap between population surveys and clinical cohorts comes in part from the fact that these are not the same people.

Harm reduction

Asphyxia is the leading cause of death, and it is a matter of inhalation method. Recreational gas is pure, with no oxygen. The main identified cause of death is the use of a mask strapped to the face and connected to a source, or of a bag placed over the head. Also to be proscribed, direct inhalation in contact with the charger and any use during a safety-sensitive activity, driving first of all. Sitting, not standing: 1.2% of past-year users report an accidental injury, with an apparent relation to the number of hits per session.

Diffusion hypoxia extends the danger beyond the hit. When inhalation stops, the very poorly soluble gas leaves the blood en masse towards the alveolus and dilutes alveolar oxygen, which is why 100% oxygen is given after stopping in clinical practice. The most hypoxaemic moment is therefore also just after the hit, and running balloons quickly one after another prevents the recovery breathing that would compensate. Breathing plain air between two balloons is not optional.

Frostbite comes from the expansion of the compressed gas, which brings the gas and the metal down to about -40 °C. A Swedish series over 22 months documents 5 patients, mean age 23 years, all previously in good health. Commonest mechanism, direct contact with the canister, four times out of five; commonest site, the inner thighs, the patients holding the canister between their legs. Mean delay before consultation 17 days, four patients out of five operated on, two re-operated, one having undergone a contracture release 21 months later. Do not hold the canister between the thighs, do not inhale in contact with the metal, go through a balloon.

Myeloneuropathy is the principal danger and its first signs are inconspicuous. A functional B12 deficiency, reaching the spinal cord and the peripheral nerves. Across 119 patients, paraesthesia is the presenting symptom in 85% of cases and the only complaint in 40%, gait ataxia affects 67%, urinary disturbance 18%. MRI shows subacute combined degeneration in 73% of usable scans, most often from C3 to C5, but 19.5% of MRIs are normal and a normal MRI does not rule out the diagnosis. The picture can present acutely and mimic Guillain-Barré syndrome, to the point that patients have been wrongly treated with immunoglobulins.

The signs that call for stopping immediately and seeking medical advice:

• pins and needles or numbness in the hands and feet, in a glove and stocking distribution, persisting after the effect has ended; this is the first sign in 85% of cases;
• unsteadiness in walking, a sense of the ground moving, worse in the dark or with the eyes closed;
• weakness in the legs, difficulty lifting the front of the foot;
• urinary, bowel or erectile problems; the authors explicitly recommend asking about them, because neither patients nor clinicians think of them;
• Lhermitte's sign, an electric shock down the spine on flexing the neck;
• an unusual and persistent headache, chest pain, breathlessness, pain or swelling in a calf, visual disturbance: possible thrombosis;
• delusional ideas, paranoia, persistent hallucinations, change in behaviour: possible encephalopathy, the form most easily missed.

A normal vitamin B12 level is not reassuring. This is the most counter-intuitive message on this page. In the series of 119 patients, serum B12 is low in only 53.5% of cases, whereas homocysteine is raised in 80% of the 20 patients tested and methylmalonic acid in 78% of the 41 tested. In Frankfurt, 95% of patients have a normal or low-normal B12, against 100% with raised homocysteine and methylmalonic acid, on 9 and 19 patients tested respectively. In Sydney, mean B12 is squarely normal and mean homocysteine is 51 µmol/L against a normal range of 5 to 15. The markers to ask for are homocysteine and methylmalonic acid, not B12 alone.

A course of B12 does not make consumption safe. Supplementation while consumption continues is unlikely to mitigate the lesions, neurotoxicity is described in spite of it, and the authors write that it does not fully protect. Worse, it normalises homocysteine and methylmalonic acid and can falsely reassure the clinician without protecting the patient. The same authors suggest considering B12 in the heavy user who does not stop, saying exactly to what end: to delay the appearance of symptoms so as to leave time to reconsider the practice. B12 buys time, it does not buy safety. The only reliable means remains stopping.

Thromboses are the second serious danger and the one users know least about. In the largest hospital in Amsterdam, out of 326 patients reporting recreational use, 17 had a serious thrombotic event, 5 arterial including three acute coronary syndromes, 12 venous including ten pulmonary embolisms, with a median homocysteine of 125 µmol/L. The actionable fact lies elsewhere: half of these patients had already had neurological symptoms before the vascular event. Pins and needles are a vascular warning sign too.

The treatment, as given by the only source consulted that offers a full protocol. Stopping exposure, which is its foundation. Cobalamin 1,000 µg intramuscularly or subcutaneously every day for 1 to 2 weeks, then 1,000 µg a week parenterally or 2,000 µg a day orally until resolution. Methionine 1 g orally three times a day for at least 4 to 6 weeks. Do not give folate before B12 repletion, a counter-intuitive instruction motivated by a risk of worsening symptoms and delaying recovery. Rehabilitation in case of disability. The authors specify that efficacy data are limited and that the optimal dose and duration remain open.

The prognosis diverges widely between series, with a follow-up bias acknowledged by the authors. One synthesis gives 95 to 97% of patients at least partly improved, with more than a third of those admitted to hospital keeping residual symptoms after months. The series of 119 patients sees only 38 people again, of whom only 4 without symptoms, but its authors note that patients with sequelae come back for consultation more readily. What does recur across several series, on the other hand, is that progression is seen in those who continue.

The interactions, in order of documented seriousness, and they are metabolic.

• Methotrexate is the best-supported interaction. The gas blocks B12 and folate at the same time, and the authors conclude that its use in the child with leukaemia at the same time as methotrexate must be avoided, all the more so as replacement anaesthetics exist. The experimental side points the same way in the rat, where methotrexate becomes strongly antileukaemic in combination with the gas: it is the same mechanism that, out of control, makes the neurotoxicity. To be flagged to anyone on methotrexate, including in rheumatology.

• Drugs that lower B12, metformin, proton pump inhibitors, H2 blockers. A cumulative effect, but it must not be inflated: in a matched cohort of 11,200 people with diabetes, adding a proton pump inhibitor to metformin brings the adjusted risk ratio for B12 deficiency to only 1.18, 95% confidence interval 1.02 to 1.35. This is a reduced starting reserve, not a multiplier.

• Respiratory depressants, opioids, sedatives and hypnotics, whose combination is explicitly advised against in obstetrics. Alcohol does not figure in that source and must not be attributed to it; it remains a contextual risk factor on another basis, two of the five patients in the Swedish frostbite series being under the influence of alcohol at the time of the injury.

• The commonest chain in this corpus is an emetic plus a sedative. Nitrous oxide is one of the three sedatives in the batch, along with dextromethorphan and diphenhydramine, and several other substances reliably induce vomiting, Syrian rue, nutmeg, morning glory and woodrose seeds. Vomiting while drowsy is the best-established lethal chain here. It has a published anchor, and it comes from another substance in the batch: the only published death on DPT is a deep aspiration of gastric contents, in a man of 20 who died neither of hyperthermia nor of agitation but of having lost the protection of his airway. The published order is the one to retain, collapse, then seizure, then vomiting: the emesis comes last, in a person already unconscious, and monitoring must therefore not wait for the first vomit to begin. What nitrous oxide adds to this picture is not a toxicity, it is sedation, and that is exactly what is enough to make it more likely. No case documents this combination: the reasoning bears on the mechanics of inhalation, not on a pharmacology. There is nothing to administer against this chain, and the gas leaves the only means of acting to those around: someone who is not taking a balloon, and who knows to turn the person who no longer responds onto their side.

• The seizure signal that runs through this corpus, carried by ten of its fourteen substances, does not pass through it: no source consulted attributes to nitrous oxide any lowering of the seizure threshold, and nothing here should manufacture one for it. Combining two carriers does indeed add two lowerings of the threshold, but what the gas adds to a combination remains what the previous bullet says, sedation, and the seizure that precedes the vomiting in the DPT death is only a reminder of who the emesis arrives on.

• A fatal co-exposure exists and it must be cited without over-reading it. A 17-year-old who died in 2001 after 30 to 35 mg of insufflated 2C-T-7 had also taken ecstasy, nitrous oxide and an ephedrine-based product. No source attributes anything whatever to the gas in that death: presence, not attribution.

• Serotonin syndrome is not the danger of this molecule. There is a single case report, a two-page letter that is permanently inaccessible, and its confounder is not a hypothesis but a published case: the pairing of buspirone and fluoxetine, with no nitrous oxide at all, produces a documented serotonin syndrome on its own. The two drugs present in the case are enough to account for it. Added to which, the serotonin transporter and monoamine oxidase have never been tested for this gas. Do not assert the interaction, do not write either that it is impossible, and do not claim that the question has been settled: it has not been asked.

• Pharmacologically, nitrous oxide is the orphan of this corpus. None of the other substances in the batch touches folate, B12 or homocysteine, and its elimination by monoamine oxidase has never been tested. The question that structures the risk on the other pages, MAO inhibition by Syrian rue, does not arise here as the data stand.

The contraindications follow first of all from the gas diffusing into closed gas-filled cavities and expanding them: pneumothorax, small bowel obstruction, middle ear surgery, retinal surgery with intraocular gas. To which are added pulmonary hypertension, severe cardiac disease, severe psychiatric disorders, head and neck procedures with cautery, the first trimester of pregnancy and altered consciousness. These contraindications are written for anaesthesia, but those bearing on closed cavities and on altered consciousness hold just as much in recreational use.

On pregnancy, beyond the first-trimester contraindication, the 2025 European delegated regulation classifies nitrous oxide as a reproductive toxicant 1B, hazard statement H360Df, may damage the unborn child and suspected of damaging fertility, applying from 1 February 2027. And the French medicines agency recorded in 2023, for the first time, reports concerning two newborns presenting with neurological disorders after exposure in utero.

Some people are more exposed than others. Pre-existing B12 deficiency, often unrecognised: pernicious anaemia, atrophic gastritis, bariatric surgery, strict vegan diet, advanced age. Treatments that lower B12. Genetic abnormalities of one-carbon metabolism, of which the most severe published case is an infant carrying a methylenetetrahydrofolate reductase deficiency, exposed twice to 60% gas, who died within two months; the role of common polymorphisms remains hypothetical and the primary text of that case has not been read. Baseline hyperhomocysteinaemia above 15 µmol/L.

In the emergency department, say what has been taken. There is no direct biological test usable in routine practice, neither in the emergency department nor in roadside testing, because of the speed of pulmonary elimination; a fraction is excreted unchanged in the urine, but the data are lacking and a urinary bacterium can produce it, which removes any specificity. All the dose data in the literature rest on retrospective self-report. Naming nitrous oxide to the clinician is the only way to steer the diagnosis, and it is what separates a recognised myeloneuropathy from a Guillain-Barré wrongly treated.

What mortality says, and does not say

In the United States, on CDC data from 2010 to 2023 in 15 to 74 year olds, 1,240 deaths in total, 23 in 2010 and 156 in 2023. The correct reading is not the ratio between the two ends: the trend analysis shows a significant rise up to 2018, then no detectable rise from 2019 to 2023, and the authors explicitly describe a plateau. Reservation given by them, the code used also covers other binary compounds of oxygen and nitrogen. In England and Wales, 56 deaths registered between 2001 and 2020, of which 45 since 2010, the third volatile substance mentioned on a death certificate but a long way behind butane at 324 and propane at 123.

The French signal is not about deaths but about harms: reports to the addictovigilance centres multiplied by 4.35 between 2020 and 2024; in 2024, mean age 22 years, more than 80% neurological disorders and about 40% serious conditions of the spinal cord or of the peripheral nerves among the cases reported. The gas was identified as the leading cause of neuropathy in 20 to 25 year olds in Île-de-France in 2021. Sale has been regulated in France since 2021 and capped since 2023, the Netherlands added the gas to list II of the Opiumwet on 1 January 2023, and the United Kingdom classified it as class C on 8 November 2023. On that last point, a useful correction for a user: simple possession does not straight away carry two years in prison, the announced regime being an unlimited fine, a visible unpaid work sentence or a warning, the two years being aimed at serious repeat offenders. The authors of the main public health review warn against criminalising the user, which risks worsening the harm by deterring people from seeking care; their article has been read in abstract only.

What is not known

The toxicity threshold, which is the most important question for a reader in harm reduction, has no answer: the two best sources leave the existence of a minimum quantity as an unresolved question. Individual susceptibility factors are suspected and not demonstrated. The frequency of macrocytic anaemia varies from 6 to 50% across series, and the reader should know that the largest direct measurement available is also the lowest. The optimal dose and duration of B12 and of methionine remain open, the long-term prognosis is poorly known for want of follow-up, and no study has systematically explored long-term cognitive effects in users. Lastly, the actual composition of the gas consumed is unknown: industrial processes may introduce other substances, of unstudied biological effect.

Sources

• C. De Halleux and D. N. Juurlink (2023), diagnosis and management of toxicity related to recreational nitrous oxide use, CMAJ (best single clinical source: treatment protocol, biomarkers, prevention).
• D. Mair et al. (2023), nitrous oxide induced myeloneuropathy, series of 119 cases, J Neurol Neurosurg Psychiatry (largest published series; clinical picture, MRI, laboratory findings, prognosis).
• M. G. Caris et al. (2022), extreme homocysteine and thrombosis in the young adult, Amsterdam series, J Thromb Haemost.
• Tshagharyan et al. (2025), neurological disorders related to recreational use, Frankfurt cohort, Neurol Res Pract.
• G. Swart et al. (2021), nitrous oxide myeloneuropathy, Sydney series, Eur J Neurol.
• E. Gernez et al. (2023), pharmacology, pharmacokinetics, toxicity and metabolic impact, Toxics (excellent on the biomarkers and on elimination kinetics; wrong on receptor pharmacology, it confuses NMDA, AMPA and kainate and contradicts itself on the kappa receptors).
• Y. Liu et al. (2024), frostbite related to recreational use, Swedish series, JPRAS Open.
• T. Yamakura and R. A. Harris (2000), effects of nitrous oxide and xenon on ligand-gated ion channels, Anesthesiology (only source putting figures on the inhibition percentages; tested neither the serotonin transporter nor MAO).
• V. Jevtović-Todorović et al. (1998), nitrous oxide, an NMDA antagonist, neuroprotectant and neurotoxin, Nature Medicine.
• D. E. Emmanouil and R. M. Quock (2007), advances on the actions of nitrous oxide, Anesthesia Progress (opioid mechanism, one-way cross-tolerance, benzodiazepine profile).
• M. Gruss et al. (2004), two-P-domain potassium channels, a target of the anaesthetic gases, Molecular Pharmacology (TREK-1 is activated, not inhibited).
• M. Murakawa et al. (1994), activation of the cortical and medullary dopaminergic systems in the rat, Anesthesia and Analgesia (no effect on brain serotonin or 5-HIAA content).
• B. D. Royston et al. (1988), rate of inactivation of human and rodent hepatic methionine synthase, Anesthesiology (primary source of the half-times, 46 min in the human at 70%, 5.4 min in the rat at 50%).
• J. T. Drummond and R. G. Matthews (1994), degradation of nitrous oxide by methionine synthase, Biochemistry (E. coli enzyme; the radical step is proposed, not demonstrated).
• I. Chanarin (1982), effects of nitrous oxide on cobalamins and folates, Critical Reviews in Toxicology (methylfolate trap).
• S. Savage and D. Ma (2014), the neurotoxicity of nitrous oxide, facts and putative mechanisms, Brain Sciences (homocysteine kinetics, vulnerable populations). ⚠️ To be treated with suspicion on three points: its chemistry of the cobalt bond is wrong, its running text contradicts its own table on the rate of inactivation, and its description of the rhesus macaque gives the impression that the gas alone damaged the brain whereas its table says the opposite, the apoptosis requiring isoflurane in combination.
• K. Knuf and C. V. Maani, Nitrous Oxide, StatPearls (blood/gas coefficient, MAC, contraindications, diffusion hypoxia).
• J. P. Zacny et al. (1994), time course of the effects of brief inhalations in the healthy volunteer, Addiction (only controlled experimental study found on the kinetics of the effects of a hit).
• J. van Amsterdam, T. Nabben and W. van den Brink (2015), recreational use, prevalence and risks, Regul Toxicol Pharmacol (dated on dependence and on public policy).
• S. J. Kaar et al. (2016), international survey of contemporary use, J Psychopharmacol (self-selected online survey, not representative of the general population).
• K. J. Pavone et al. (2016), slow and delta oscillations induced by nitrous oxide, Clinical Neurophysiology (figures cross-checked on the abstract, body of the article not read).
• D. Diesing and P. Neu (2024), nitrous oxide and the ICD-10 dependence criteria, Der Nervenarzt.
• V. J. Forster et al. (2016), drug interactions as a risk factor for methotrexate neurotoxicity, Cancer Chemother Pharmacol (explicitly recommends avoiding the gas in patients on methotrexate) and J. Abels et al. (1990), antileukaemic potential of methylcobalamin inactivation, Am J Hematol (experimental side in the rat).
• C. Jung, S. Park and H. Kim (2025), metformin and proton pump inhibitors, risk of B12 deficiency, J Diabetes Investig (adjusted risk ratio 1.18, a real and modest effect).
• G. H. Manos (2000), possible serotonin syndrome after adding buspirone to fluoxetine, Ann Pharmacother (the key piece of the serotonergic question: the pairing in the incriminated case produces the syndrome with no nitrous oxide at all).
• P. S. Myles et al. (2025), antidepressant effects of nitrous oxide, phase 2b trial, Biol Psychiatry Glob Open Sci (primary endpoint missed at p = 0.051; only 40 patients received the gas, and the article does not say how many were taking a serotonergic agent), with P. Nagele et al. (2015), Biological Psychiatry and P. Nagele et al. (2021), Science Translational Medicine (in 2021, no difference detected between 25 and 50% in 24 patients, which is not a non-inferiority).
• P. S. Myles et al. (2014), ENIGMA-II, The Lancet and K. Leslie et al. (2015), one-year follow-up, Anesthesiology (no excess cardiovascular mortality, with the reservation written by the CMAJ authors themselves: these results are not generalisable to chronic users, in whom thromboses are documented).
• R. A. Yockey and R. A. Hoopsick (2025), nitrous oxide related mortality in the United States, JAMA Network Open (1,240 deaths, a rise up to 2018 then a plateau; the figure of +578% in circulation does not come from the article and contradicts its conclusion).
• Office for National Statistics, deaths related to volatile substances, helium and nitrogen, England and Wales, 2001-2020 (56 deaths, against 324 for butane and 123 for propane).
• S. A. Zaloum et al. (2025), addressing the growing burden of public health harms, The Lancet Public Health. Abstract only, article inaccessible.
• European Commission, delegated regulation (EU) 2025/1222 of 2 April 2025, Annex VI of the CLP regulation (Repr. 1B H360Df, STOT RE 1 H372 nervous system, STOT SE 3, Ozone 1; applying from 1 February 2027).
• Home Office, Nitrous oxide ban: guidance, gov.uk and Rijksoverheid, ban on nitrous oxide as a recreational product (British and Dutch statuses verified on the government pages).
• É. Poumirol (rapporteur), Proposition de loi visant à réserver la vente de protoxyde d'azote aux seuls professionnels, report no. 416, Sénat, 2025-2026 session, 18 February 2026. Best source of French data. Read by automatic extraction; several figures, among them the prevalences among secondary school pupils, the masses and the prices of the packagings, could not be cross-checked on the pages consulted.
• J. Allan, J. Cameron and J. Bruno (2022), systematic review of recreational use, Int J Environ Res Public Health (notes the absence of published harm reduction strategies) and K. Kor et al. (2025), co-production of resources with young users, Harm Reduction Journal (they prefer practical and actionable advice to an abstinence message).
• Erowid, Nitrous Oxide Vault (corpus of user reports, to be read on site; its licence forbids reproduction, and no Erowid content has been consulted or used here, the pages having returned HTTP 403).
• L. M. Volz, B. L. Muldowney, M. C. Trawicki (2016), a case of serotonin syndrome attributed to nitrous oxide in an anxious adolescent, Journal of Clinical Anesthesia 32:27-28. Text not read and permanently inaccessible: neither PubMed nor Europe PMC has an abstract of it, and the publisher's index requires authentication. It is the only existing case report, and its clinical content is known to no one in this file.
• R. R. Selzer, D. S. Rosenblatt, R. Laxova, K. Hogan (2003), adverse effect of nitrous oxide in a child carrying a methylenetetrahydrofolate reductase deficiency, New England Journal of Medicine 349(1):45-50. Text not read, with no abstract available; the clinical details come via Savage and Ma 2014.
• World Health Organization, Expert Committee on Drug Dependence, 46th meeting, pre-review report on nitrous oxide. Not read, available as PDF only. It is probably the most complete synthesis in existence and it deserves a manual reading.
• H. Davy (1800), B. P. Blood (1874), W. James (1882). No primary text read. ⚠️ For James, two incompatible paginations circulate without any original having been consulted: write no pagination. The attribution to Davy of the phrase laughing gas remains unverified.