Biologic Effects of Nitrous Oxide: A Mechanistic and Toxicologic Review

The minimum alveolar concentration (MAC) of nitrous oxide is reported as 104% with the initial study completed in hyperbaric conditions on seven human volunteers exposed to tetanic electrical impulses as the noxious stimulus (rather than the standard incision)16; therefore, nitrous oxide is the least potent inhalational anesthetic agent in current practice. The predominant theory of the molecular mechanism to account for nitrous oxide’s anesthetic action is noncompetitive inhibition of the NMDA subtype of glutamate receptors.13The importance of this receptor has recently been confirmed using sophisticated pharmacogenomic techniques in Caenorhabditis elegans  14and in mice.15Whether there is also a role for non-NMDA glutamate receptors remains unclear, but inhibitory effects on these receptors have also been reported.17Therefore, inhibition of excitatory glutamatergic neurotransmission is central to the thesis for the anesthetic actions of nitrous oxide. Another potential target contributing to nitrous oxide’s anesthetic action are the two pore domain potassium channels, such as the TREK-1 channel, which, when activated, increases potassium conductance and thereby hyperpolarizing neurons away from their firing threshold.18Although the role of γ-aminobutyric acid type A (GABA^A) receptors in anesthetic states produced by intravenous (propofol, etomidate, and pentobarbital) and halogenated volatile (halothane and enflurane) anesthetics seems well established,19nitrous oxide exerts insignificant effects on these inhibitory receptors.17Despite expanding knowledge of the molecular mechanisms of nitrous oxide, clarification of the specific neural substrates and supraspinal loci for the hypnotic effect of nitrous oxide is required.

The mechanisms of analgesic action have been recently reviewed,20and therefore we will merely summarize and update the findings described elsewhere. Nitrous oxide acts supraspinally to induce analgesia by activation of opioidergic neurons in the periaqueductal gray matter and noradrenergic neurons in the locus ceruleus, A5 and A7 areas of the brainstem.21,22Critical to the activation of the locus ceruleus (and potentially the other loci) appears the hypothalamic release of corticotropin-releasing factor23(fig. 1), which may be provoked by antagonism of the NMDA receptor.24Supraspinal administration of antagonists for the corticotropin-releasing factor and opioid receptors attenuates the antinociceptive effects of nitrous oxide in rats.23Opioid release in the brainstem inhibits activity of γ-aminobutyric acid–mediated (GABAergic) interneurons, thereby removing the inhibitory tone on the descending noradrenergic inhibitory pathways (fig. 1).25,26Upon activation of these descending pathways, the antinociceptive effects of nitrous oxide are transduced via α¹and α²adrenoceptors within the dorsal horn of the spinal cord.21,27The identity of these adrenoceptors was determined by intrathecal administration of selective antagonists to attenuate nitrous oxide’s antinociceptive action. Depletion of spinal norepinephrine also attenuated nitrous oxide’s antinociceptive action.28Nitrous oxide was also ineffective as an antinociceptive agent (but not as a hypnotic–anesthetic agent) in transgenic mice lacking functional α^2Badrenoceptors, establishing a role of this receptor subtype for its analgesic action.21The importance of the descending inhibitory neurons was further established after the findings that nitrous oxide lacks antinociceptive effects when the spinal cord is transected in adult rodents28and in young animals29at ages when the descending inhibitory neurons lack functional connectivity with their targets. Whether this lack of efficacy in young animals pertains to young humans is currently being investigated.

Nitrous oxide is an effective inhalation analgesic in multiple settings; furthermore, we now understand much of how nitrous oxide transduces its analgesic effect. This understanding also informs us further about the paradoxical interaction nitrous oxide displays with agents that potentiate the GABA^Areceptor.

Nitrous oxide’s analgesic action is dependent on both the inhibition of supraspinal GABA^Areceptors as well as activation of spinal GABA^Areceptors26(fig. 1). Therefore, agents that activate the supraspinal GABA^Areceptor may interfere with nitrous oxide analgesia by inhibiting the activation of the descending inhibitory neurons.26This may explain the reported antagonism of nitrous oxide’s antinociceptive effects by the administration of volatile anesthetics,30,31midazolam,26and propofol32in animals.

Some clinical evidence supports this view; in healthy volunteers inhaling subanesthetic concentrations of sevoflurane and nitrous oxide, the sevoflurane element decreases nitrous oxide’s analgesic efficacy33(fig. 2). However, the interaction at subanesthetic and anesthetic doses may not be the same. To further dissect the importance of nitrous oxide’s analgesic mechanism to nitrous oxide’s clinical anesthetic role, we have compared MAC data from humans at time points when descending inhibitory neurons are functional (adults) and when they are not functional (infants). If functional descending inhibitory neuron activation is required for nitrous oxide’s additivity with volatile anesthetics, any interaction would be predicted to be subadditive in infants. On the contrary, if descending inhibitory neuron activation is irrelevant to nitrous oxide–volatile anesthetic additivity, an additive interaction in both childhood and adulthood would be expected. In fact, the latter is true; nitrous oxide does interact additively with volatile anesthetics in human infants34and adults.35This evidence suggests that activation of descending inhibitory neurons is not critical for the anesthetic action of nitrous oxide when coadministered with a volatile anesthetic agent (as occurs in clinical practice).

To verify these extrapolations, we have examined data from a clinical study of 1.66 MAC desflurane with or without 60% nitrous oxide (in the absence of opioids). The authors found a similar effect on blood pressure responses to electrical titanic stimulation between the groups (though heart rate responses were lower in the nitrous oxide group36). Because adrenergic responses are frequently used as a measure of pain responses intraoperatively, these findings suggest that the use of nitrous oxide during maintenance of anesthesia as an additional analgesic  may be confounded and that any analgesic effect, in this context, should be regarded as minimal.

In hyperbaric conditions, nitrous oxide can induce anesthesia itself,16and therefore one would suspect analgesia to play a role in nitrous oxide’s ability to induce anesthesia. However, in hyperbaric conditions, nitrous oxide predominantly inhibits ventral horn, rather than dorsal horn, responses to supramaximal noxious stimulus37similar to other anesthetics38to provoke immobility. If, rather than immobility, sedation is taken as an endpoint, we can also dissect the sedative and analgesic actions of nitrous oxide; in transgenic mice lacking α^2Badrenoceptors, antinociceptive responses to nitrous oxide are abolished, but sedative responses are unaffected.21Lesioning of the locus ceruleus also attenuated the antinociceptive effects of nitrous oxide without hindering its sedative effects. Therefore, the immobilizing and sedative actions of nitrous oxide should be considered discrete from its analgesic effects.

These preclinical and clinical data suggest that nitrous oxide analgesia is not important for its anesthetic action when administered clinically, i.e. , in combination with a GABAergic agent. This is clinically relevant and is supported by the extensive preclinical mechanistic work. The relevance of the mechanism of nitrous oxide’s analgesic action to clinical anesthetic practice can be appreciated in different clinical settings. A prime example is its use for the provision of general anesthesia for cesarean delivery (a period of risk of awareness), when clinicians may include nitrous oxide in the anesthetic regimen for extra “analgesia’”; however, there is no evidence that addition of nitrous oxide to a volatile anesthetic enhances the analgesic  effects of the anesthetic—merely the depth of anesthesia.

Having explored this complicated interaction with anesthetic agents that affect GABA^Areceptors, we will discuss the importance of drugs that modulate opioid signaling on nitrous oxide’s analgesic effect. Nitrous oxide activates supraspinal opioid receptors (fig. 1) via  corticotropin-releasing factor.23An interaction between nitrous oxide, which indirectly stimulates opioid receptors, and exogenous opioids could be expected. Consistent with this, nitrous oxide reduces the MAC-sparing effect of morphine on isoflurane anesthesia in rats,39suggesting overlap between morphine and nitrous oxide’s effects. Unfortunately, a dose–response curve for the effect of morphine was not constructed; hence, it is not confirmed that the reported “antagonism” exists between morphine and nitrous oxide39; in fact, this effect may represent additivity between opioids and nitrous oxide.

In an attempt to probe this interaction further, we have compared the clinical work of Ghouri and White40and Sebel et al. ,41though the reader should note that these were independent studies and therefore we are restricted to a purely qualitative comparison of the data. In these two studies, a MAC value of 2.6% was found when 3 μg/kg fentanyl was coadministered with desflurane (in oxygen without nitrous oxide),41and a MAC value of 3.0% desflurane was found when 3 μg/kg fentanyl and 60% nitrous oxide (in oxygen) was given.40This comparison suggests that if sufficient opioid is administered, the addition of nitrous oxide would not act to further decrease MAC.

However in another study, addition of 60% nitrous oxide reduced the MAC of sevoflurane from 3.96% to 1.2% during target-controlled analgesia with 1 ng/ml remifentanil, and this sevoflurane-sparing effect was still detectable at 3 ng/ml remifentanil.42The apparent discrepancy in the findings with fentanyl and remifentanil may stem from a difference in the actions of fentanyl and remifentanil on the NMDA receptor, because remifentanil potentiates the NMDA receptor.43Nitrous oxide, as an NMDA antagonist, may have blocked remifentanil’s activating effect on the NMDA receptor and thus continued to be a MAC-sparing drug in the presence of remifentanil. Through a similar mechanism, nitrous oxide may be able to inhibit opioid-induced postoperative hyperalgesia,44though further studies are required to elucidate whether this action of nitrous oxide persists in the presence of a volatile anesthetic agent.

We still remain relatively uninformed about the additivity of different anesthetic agents (be they GABAergic, NMDA antagonistic, or opioidergic) on MAC, especially when used in combination, as frequently occurs in clinical practice. We hasten to point out that the clinical data support an additive interaction between nitrous oxide and halogenated volatile anesthetics; however, the addition of nitrous oxide in this context should not be as an analgesic but for the same indication as other anesthetic agents. The interaction with opioids is complicated and requires further study, but the overlap in their mechanism of action explains both their profound analgesic capabilities and why they may have an interchangeable role in anesthesia. These discussions highlight the importance of studying the mechanisms of anesthesia and consequently understanding the ways we should use current agents and develop new ones.

Aside from the biologic actions of nitrous oxide transducing its anesthetic and analgesic actions, nitrous oxide also affects methionine synthase function. Nitrous oxide oxidizes the cobalt I (Co⁺) form of cobalamin (vitamin B¹²) to Co³⁺11:

A further reaction rapidly ensues:

However, an alternative reaction has also been proposed to account for nitrous oxide’s oxidation of cobalamin involving the generation of hydroxyl radicals that irreversibly oxidize cobalamin12:

By either of the above mechanisms, the resulting oxidized cobalt cation prevents cobalamin (vitamin B¹²) acting as a coenzyme for methionine synthase. Methionine synthase is a ubiquitous cytosolic enzyme that plays a crucial role in the generation of methyl groups (via  the active intermediary s-adenosylmethionine) for the synthesis of DNA, RNA, myelin, and catecholamines, among other products. After methyl group donation, s-adenosylmethionine is converted to homocysteine, which can then reenter the methionine pathway or be metabolized to cystathione (fig. 3). In addition to cobalamin, methionine synthase requires 5-methyltetrahydrofolate to function as a coenzyme. One carbon transfer by folates plays a crucial role in the biosynthesis of pyrimidines and purines and in serine and glycine metabolism. This latter interconversion between serine and glycine produces the methyl groups to be added to homocysteine to produce methionine. The methyl group is initially bound to tetrahydrofolate (via  an intermediary) to give 5,10-methylenetetrahydrofolate, which is then reduced to 5-methylenetetrahydrofolate. This is of importance because this pathway represents the only way to produce 5-methylenetetrahydrofolate. The methyl group is then transferred to cobalamin, producing methylcobalamin, the final methyl group donor for methionine synthesis.

This pathway is critical to cellular function, and decreased methionine synthase activity can result in both genetic and protein aberrations.45Certain patient groups may be particularly susceptible to reduced methionine synthase activity, including those deficient in cobalamin, such as patients with pernicious anemia or ileal disease, alcoholics, the elderly, and the malnourished45,46(table 1).

In preclinical studies, Sharer et al.  47showed that a 24-h exposure to concentrations of 860 parts per million (ppm) or lower of nitrous oxide did not significantly change methionine synthase function in Sprague-Dawley rats. However, at anesthetic concentrations in rats, methionine synthase activity is inhibited rapidly; 50% nitrous oxide exposure decreased methionine synthase activity within 30 min, and the activity was virtually undetectable after 6 h.48Because nitrous oxide readily crosses the placenta, this inhibition has been recorded in both the fetus and the dam, with 60 min of nitrous oxide (50%) inhibiting methionine synthase activity to 11% and 18% in maternal and fetal livers in rats, respectively.49Extrapolations from these findings to the clinical situation are hindered because of species differences; notably, rats are more sensitive to nitrous oxide–induced methionine synthase inhibition than humans are.50Nonetheless, patients anesthetized with nitrous oxide (70%) exhibited a reduction in methionine synthase activity measured in liver biopsies, with a time to half methionine synthase activity of 46 min.50After 200 min of nitrous oxide anesthesia, methionine synthase activity approaches zero. Koblin et al.  51demonstrated that nitrous oxide (70%) inhibited methionine synthase activity in liver biopsies with a 50% reduction in activity predicted after approximately 1.5 h. Typically, the enzyme’s function will recover within 3–4 days after exposure to nitrous oxide.48,51Consistent with this effect, the duration of nitrous oxide exposure is correlated with increased homocysteine levels,52which can be prevented by preoperative vitamin B complex therapy.53 

An important enzyme in the folate cycle, 5,10-methylenetetrahydrofolate reductase (MTHFR), also plays an important role in the conversion of homocysteine to methionine by generating 5-methyltetrahydrofolate (fig. 3). Two single-nucleotide polymorphisms are known (677 cytosine–thymidine and 1298 adenosine–cytosine) and are associated with reduced enzyme activity.54,55Mutations in this gene are relatively common, e.g. , one homozygous mutation of MTHFR gene (677C–T) is present in 0–3% of African Americans, 9–11% of white Americans, and 32% of Mexicans.56In Europe, the highest prevalence seems to be in Southern Europe, with a prevalence of 20–26% in Southern Italy.55Homozygosity for many of these mutations is associated with increased homocysteine levels and reduction in MTHFR enzyme activity.54Interesting recent data show that patients who receive nitrous oxide anesthesia with a homozygous mutation in MTHFR have a greater postoperative increase in plasma homocysteine levels than heterozygotes or wild type.57As yet, the functional consequences of this homocysteine increase are unknown, but it has been hypothesized to play a risk in increased perioperative myocardial morbidity (see section titled Patient Exposure Risk, Myocardial Effects).

Nitrous oxide has been associated with varied effects on the immune system, with decreased proliferation of human peripheral blood mononuclear cells58and increased and decreased neutrophil chemotaxis reported.59,60Reports such as these have led to concern that nitrous oxide may detrimentally affect immune function. A recent multicenter study of 418 colorectal surgical patients evaluated the effect of nitrous oxide anesthesia (65% nitrous oxide vs.  65% nitrogen) on wound infection and healing61and found no difference in infection rate, wound healing, hospital stay, or mortality. We must highlight that this trial was powered to find a doubling in the relative risk of wound infection, and therefore a more subtle effect may have been missed.

The ENIGMA trial demonstrated some postoperative benefits of high concentrations of intraoperative oxygen (80%) in nitrogen (20%) when compared with nitrous oxide (70%) in oxygen (30%).5Significant differences were noted in the incidences of postoperative complications between the two groups including wound infection, fever, pneumonia, and atelectasis. With 2,050 patients enrolled, this was a very large multicenter randomized controlled trial of patients undergoing major surgery that was powered to show a difference in hospital stay (from 4 days to 3.5 days; β power = 0.9) and a reduction in wound infection from 14% to 10%.

Before further discussion of the results, we wish to highlight a few caveats to the reader regarding the clinical trial design. Structured follow-up occurred on the first postoperative day with a telephone consultation and medical record review at 30 days; in the intervening phase, diagnosis of patient complications was a responsibility of the surgical team, without standardized assessment or stated prospective guidance. Although this may reflect standard medical practice, this may have resulted in underdiagnosis of complications. The study was also not blinded, because those individuals diagnosing wound infection and deciding on discharge from the hospital were aware of the treatment group (because the anesthetic record was available in the notes). Exposure to volatile anesthesia was not standardized, and the nitrous oxide–free group received significantly more propofol to maintain anesthesia. Also, it is unfortunate that biochemical and hematologic monitoring of the patients was not standardized despite the well-known effect of nitrous oxide on methionine synthase. Furthermore, it is also unclear on what criteria the surgeons selected radiologic investigation to establish the presence of atelectasis, making this a questionable endpoint for the study. Finally, as the authors concede, the large number of comparisons in the trial “increases the chance of a type I error; the secondary, exploratory and subgroup analyses should be treated with caution.”

The ENIGMA trial found no difference between the groups for the primary endpoint of the trial; the median duration of hospital stay did not differ between groups, though it did narrowly miss significance (7.0 vs.  7.1 day; P = 0.06). However, the trial did show a significant reduction for severe nausea or vomiting (which persisted after controlling for increased propofol use in the nitrous oxide–free group) and increased incidences of wound infection (odds ratio [OR], 0.72; 95% confidence interval [CI], 0.52–0.98), pneumonia (OR, 0.51; 95% CI, 0.27–0.97), and atelectasis (OR, 0.55; 95% CI, 0.40–0.75) in the nitrous oxide group. Indeed, the finding of a significant difference in wound infection is remarkable because the incidence was reduced from 10% to 7.7%, with 29 fewer patients getting infected in the nitrous oxide–free anesthesia group. However, these are all secondary endpoints that are vulnerable to type I error.

We consider that the positive results from the ENIGMA trial, with the methodologic and statistical caveats already mentioned, could equally be due to the positive effects of high-dose oxygen therapy or any inherent toxicity of nitrous oxide. The experimental study design of this “pragmatic” clinical trial makes scientific interpretation of the results difficult. The two groups are essentially incomparable because both oxygen concentration and balance gas (nitrogen or nitrous oxide) vary. Therefore, with two variables changing, one cannot definitively know which of the two is the critical factor or whether the abandonment of nitrous oxide plus the adoption of high concentrations of oxygen is necessary to reduce postoperative complications (which has been discussed further5,6). In the absence of a clear experimental approach, interpretation of the study becomes speculative as to the critical causative factors involved. Nonetheless, we consider in sum that there is no evidence to suggest that nitrous oxide induces more immunosuppression than other anesthetics. Hence, we have sought to balance these interpretations about the toxicology of nitrous oxide with the available data from the literature in this review.

It is also of interest that in a follow-up trial,§patients will be randomly assigned to receive either 70% nitrous oxide or nitrogen plus 30% oxygen, which allows more direct comparison between the groups. Unfortunately, a group receiving 70% nitrogen with 30% oxygen was not included in the ENIGMA study because it “is not often used clinically.” Nonetheless, we believe that a trial of this nature has the potential to address the association between nitrous oxide and increased postoperative complications.

Inhibition of methionine synthase may lead to significant hematologic complications such as megaloblastic anemia.45Indeed, even short periods of nitrous oxide exposure (2–6 h) in seriously ill patients can cause megaloblastic bone marrow changes.62Another vulnerable group at risk from further methionine synthase suppression may be the elderly, because up to 20% are deficient in cobalamin.46Deleu et al.  63followed elderly patients randomly assigned to propofol or nitrous oxide anesthesia. In the nitrous oxide group, postoperative serum but not erythrocyte folate levels were lower than in the propofol group. Hematocrit and hemoglobin levels also decreased, and mean cell hemoglobin levels increased slightly but significantly in the nitrous oxide group (preoperatively vs.  postoperatively), but no difference was reported between the two anesthetic groups. Three patients experienced neurologic problems such as a sore tongue, mild ataxia, and peripheral polyneuropathy (two were deficient in erythrocyte folate preoperatively), which resolved with folate therapy. In contrast, Koblin et al.  64did not find any changes in two markers of methionine synthase function (urinary formic acid and formiminoglutamic acid urinary excretion) after 3 h of nitrous oxide exposure during hip replacement in elderly patients. A minor increase in these markers was found in younger patients undergoing longer surgery for resection of acoustic neuromas, with a mean duration of anesthesia of 9.3 h. Therefore, whether the elderly represent a population vulnerable to nitrous oxide–induced methionine synthase inhibition requires further investigation, though care should be taken with patients known to be folate or cobalamin deficient.

In another study, Waldman et al.  did not find hematologic abnormalities in orthopedic and neurosurgical patients exposed to nitrous oxide, apart from a smaller perioperative increase in leukocyte count.65Other studies have shown that durations of exposure to nitrous oxide in excess of 6 h are required to provoke hematologic changes that are preventable by preoperative folinic acid supplementation.66,67However, patients deficient in cobalamin and substrates for methionine synthase are at potential risk even from exposure to short durations of nitrous oxide. A recent case report showed that after nitrous oxide–based anesthesia, a child with preoperatively unrecognized cobalamin deficiency developed neurologic symptoms and pancytopenia that resolved with vitamin supplementation.68Nonetheless, there is no evidence that individuals who are not deficient in cobalamin or folate are vulnerable to hematologic complications of nitrous oxide if exposed for less than 6 h.

Reduced methionine synthase function may lead to myelinopathies such as subacute combined neurodegeneration of the spinal cord,45which has also been reported to occur with long-term nitrous oxide abuse.69,70More severe neurologic complications have been reported after nitrous oxide exposure in patients with rare congenital gene deficiencies.71,72Myelin degeneration after chronic nitrous oxide exposure has been observed in monkeys73and pigs74with some protection afforded by methionine supplementation. Interestingly, in vitamin B¹²–deficient fruit bats, pretreatment with folates accelerates the development of nitrous oxide–induced neurotoxicity.75This has led to a hypothesis that a rapid increase of methyl-folate concentrations in the nervous system with simultaneous inactivation of vitamin B¹²may produce nitrous oxide’s neurotoxic effects.45 

Neurologic injury has also been noted after a routine nitrous oxide–based anesthetic in patients with cobalamin deficiency, such as those with pernicious anemia, although the injury did not become apparent for many weeks.76In another report, a 6-month-old girl developed hypotonia and collapsed with metabolic acidosis, reduced serum cobalamin, and diffuse cerebral atrophy after nitrous oxide anesthesia.72The reason for the patient’s cobalamin deficiency was unknown, although the mother was noted to also be deficient, having avoided dairy and meat products while breast-feeding.

In contrast, consumption of large quantities of nitrous oxide is required in the nondeficient individual to induce a similar toxicity such as that which occurs in the setting of abuse.70Nitrous oxide abusers can present with altered mental status, paresthesias, ataxia, and weakness and spasticity of the legs. A recent case report described a 21-yr-old man who used nitrous oxide at home for 4 months for postoperative analgesia after gastrointestinal tract surgery and developed a demyelinated cervical spinal cord lesion. After treatment with cobalamin and folate, the neurologic symptoms resolved.

Nitrous oxide may also prove toxic in certain rare congenital disorders encountered in pediatric practice. A child exposed twice for elective anesthesia presented 25 days postoperatively with seizures and episodes of apnea.71The patient died at 130 days of age, 46 days postoperatively, as a result of respiratory arrest. Postmortem examination showed MTHFR deficiency, which is a rare autosomal recessive disorder characterized by progressive hypotonia, convulsions, and psychomotor retardation.

As described in the section titled Patient Exposure Risk, Hematologic Effects, Deleu et al.  63reported on two patients who presented for ophthalmic surgery as part of a randomized controlled trial with reduced erythrocyte folate levels and subsequently developed neurologic symptoms after nitrous oxide exposure. Furthermore, cobalamin-deficient patients are at risk71,72,76from nitrous oxide toxicity, indicating that both should be tested before nitrous oxide use when deficiency is suspected. Because serum cobalamin levels may correlate poorly with tissue levels, a normal serum level does not exclude deficiency. Assessment of methylmalonic acid and homocysteine levels has been suggested to improve the sensitivity of detection of cobalamin deficiency and may be useful in this context.77Consequently, we suggest that it is important to replace both cobalamin and folate in the malnourished before nitrous oxide anesthesia to reduce the potential risk of toxicity. Despite being used to treat nitrous oxide neurologic injury previously,63folate monotherapy should not be used in this context because of fears that it may exacerbate neurologic injury.75However, we stress that nitrous oxide is unlikely to precipitate neurologic injury secondary to demyelination in the majority of cases; this is exemplified both by the prevalence of nitrous oxide use and by the relative scarcity of the case reports reporting this form of injury.

In accord with most other drugs that inhibit the NMDA receptor, nitrous oxide provides both neuroprotection and neurotoxicity.13Nitrous oxide is a weak neuroprotective agent against excitotoxic or hypoxic–ischemic brain injury in adult rodent models,13,78though there is also evidence disputing this effect.79Currently, there is insufficient evidence to suggest an indication of nitrous oxide as a neuroprotective agent. Furthermore, in certain situations where a neuroprotective agent may prove advantageous, such as cardiac and neurosurgery, nitrous oxide is avoided because of effects on cerebral blood flow and the potential to exacerbate air embolism. Nitrous oxide also causes reversible neurotoxicity in the retrosplenial and posterior cingulate cortices with vacuolation of pyramidal neurons in rats.13This morphologic neurotoxicity occurs at doses in excess of those normally administered in clinical practice (i.e. , in hyperbaric conditions) and resolves within 3 h in rats. However, markers of cerebral injury are detectable at clinical doses when the morphologic injury is not observable.80The functional correlate of this injury is thought to be the psychomimetic reactions that can occur with this class of agent. Coadministration of a GABAergic agent, such as isoflurane or propofol, as usually occurs in anesthetic practice, can attenuate this toxicity.13,81Female rats seem especially vulnerable to the injury produced by NMDA antagonism, in keeping with the susceptibility of human females to the anesthetic effects of this class of agent.82Interestingly, despite both being NMDA antagonists, xenon does not induce the same toxicity as nitrous oxide,80which may be related to differences in dopamine release.83Corroborating this, nitrous oxide neurotoxicity is also inhibited by the dopamine antagonist haloperidol,80further suggesting that the toxicity may be related to nitrous oxide–induced dopamine release.

Recently, exposure to anesthetic agents has been associated with a disturbing pattern of neurodegeneration in the neonatal rat, which has led to concerns regarding the vulnerability of the developing human brain.3NMDA antagonists, such as ketamine and MK-801, have been shown to provoke widespread apoptotic neuronal death in the neonatal rat brain84; again, it is interesting that xenon does not seem to induce this toxicity.85Similar to xenon, nitrous oxide exposure alone does not induce apoptosis in the neonatal rat brain at concentrations up to 75%. However, unlike xenon, it does exacerbate the injury provoked by isoflurane (0.75%).3,85Whether equianesthetic dosing of nitrous oxide–isoflurane or sole isoflurane anesthesia provokes the same injury is unknown, and any additivity should be sought to clarify this matter.

Recently, data from 5-day-old rhesus monkeys have been published suggesting that NMDA antagonist–induced injury in the young may be of less clinical importance than initially feared. Whereas exposure to ketamine for 24 h produced robust neurodegeneration in the brain, 3 h of ketamine anesthesia did not.86Currently, the evidence of injury from nitrous oxide or other NMDA antagonists is insufficient to condemn them from pediatric anesthetic practice. Indeed, nitrous oxide is a useful agent for gaseous inductions in the young because of its odorless quality and the second gas effect, and there is no evidence currently that its use in this circumstance will negatively affect a child’s neurodevelopment.

Another putative neurologic complication of anesthesia is postoperative cognitive dysfunction (POCD). The contribution of general anesthesia to this complication has recently been questioned because there is equal incidence of POCD in patients who have undergone general and regional anesthesia.87The direct effect of nitrous oxide on POCD was explored in an animal model4; aged rats exposed to 70% nitrous oxide showed impaired choice accuracy in a radial maze (a surrogate of spatial working memory) after exposure (with no difference in error rate and time to complete the maze). This cognitive dysfunction was preceded by a reduction in cortical methionine synthase activity. However, in a clinical study published recently, no association was found between POCD and nitrous oxide exposure, although it was underpowered for this endpoint.88No difference in the acute confusional state, delirium (powered primary endpoint), was found either. Data from our recent study suggest that a neuroinflammatory response to surgery is responsible for POCD in an animal model,89which would explain the lack of difference in POCD between different anesthetic techniques. On balance, experimental data do not suggest that the phenomenon of POCD is attributable to nitrous oxide; although anesthesia is still a possible risk factor, other factors, such as age and the surgery undertaken, are more likely to be of importance. However, further studies are warranted because hyperhomocysteinemia, as well as reduced folate and cobalamin levels, has been implicated in the pathogenesis of both vascular and Alzheimer dementia; therefore, a role for nitrous oxide remains biologically plausible, especially with high prevalence of cobalamin deficiency in the elderly.46Nonetheless, there are currently no clinical data to suggest that nitrous oxide anesthesia is a significant risk factor for the development of POCD.

Nitrous oxide has been associated with increased myocardial risk in the perioperative period.90Badner et al.  52initially studied the effects of nitrous oxide administration on plasma homocysteine levels because the conversion of homocysteine to methionine is methionine synthase dependent; they found a significant increase with nitrous oxide administration. As discussed earlier, certain genotypes with point mutations in the MTHFR gene may be at even greater risk of this increase.57Increased homocysteine levels are an independent risk factor for cardiac morbidity, potentially through causing endothelial dysfunction and procoagulation.91,92In a study of 90 patients undergoing carotid endarterectomy randomly assigned to isoflurane or isoflurane–nitrous oxide anesthesia, a higher incidence of ischemia with significantly longer ischemic events during the first 24 and 48 h was observed in the nitrous oxide group.90Unfortunately, the group did not report on enzyme changes or follow-up of the patients, making the true extent of the problem unclear. Although intraoperative ischemia has been noted previously with nitrous oxide administration, this did not result in cardiac morbidity.93In an earlier study by Kozmary et al. ,94no effect on enzyme changes was noted with nitrous oxide, though the study may have been underpowered for this endpoint. It should be noted that although plausible, causation between any increased risk from nitrous oxide and increased homocysteine levels remains unproven. A further exploration by Myles et al.  95(in this issue of Anesthesiology) showed a correlation between nitrous oxide exposure, increased homocysteine level, and impaired flow-mediated dilatation of the brachial artery. The duration of nitrous oxide exposure further correlated with this marker of endothelial dysfunction, increasing the likelihood of causation. If hyperhomocysteinemia secondary to nitrous oxide exposure indeed contributes to cardiac toxicity, populations such as the elderly with elevated baseline homocysteine levels45or those with certain MTHFR polymorphisms57may be particularly at risk. Another explanation is that the group that did not receive nitrous oxide was exposed to higher concentrations of the preconditioning agent isoflurane, and so the study has revealed, clinically, the lack of preconditioning effect of nitrous oxide that has been established in animals.96 

Contrariwise, nitrous oxide exhibits less cardiodepressant properties than other volatile agents and therefore may improve intraoperative hemodynamic parameters.8Further studies are required to elucidate the impact of nitrous oxide–based anesthesia on perioperative cardiac outcomes; however, it is interesting that in the recent ENIGMA study, nitrous oxide was not associated with adverse cardiac outcomes,5though it was underpowered for this endpoint. As described in the section titled Patient Exposure Risk, Immune Effects, a follow-up study, the ENIGMA II trial, plans to enroll 7,000 patients to evaluate the risk of nitrous oxide anesthesia in patients with coronary artery disease.

Concerns about the reproductive effects of exposure to anesthetic doses of nitrous oxide were founded in part by preclinical studies concentrating on exposure to concentrations of nitrous oxide greater than 50% for 24 h on day 1 of gestation.97,98However, the clinical relevance of these studies remains doubtful because of the extremely long duration of exposure and high doses used. Interestingly, in these settings, halothane prevented the nitrous oxide toxicity, including both fetal loss and neuroanatomical abnormalities99; isoflurane also prevented fetal losses but not the skeletal abnormalities.98These protective effects of volatile anesthetics have been ascribed to enhanced uterine blood flow; however, it remains unknown whether these effects are important clinically.

Some clinical studies have evaluated the risks associated with general anesthesia during pregnancy, but they are also confounded by the surgery undertaken and the pathology necessitating surgical intervention; furthermore, mixed anesthetic gas exposure was used, thereby precluding evaluation of the teratogenicity solely on the basis of nitrous oxide. Obtaining adequate controls is a problem because it would be unethical to randomly assign patients to an anesthesia-only group. However, the effects of clinical anesthesia itself on fetotoxicity have been evaluated with nonexperimental studies, and the results to date suggest that anesthesia is nontoxic.100–102Mazze and Kallen101addressed this issue in a large registry study of 5,405 patients and found no association between different methods of anesthesia and adverse outcomes. An earlier study by Duncan et al.  102found no teratogenic effect and no difference in the incidence of spontaneous abortion with controls. Subgroup analysis from this study did show a relative risk associated with certain procedures, such as those performed by obstetricians and gynecologists.102 

In summary, the fetotoxic effects of anesthetic agents remain obscure and require further investigation, but the lack of appropriate control groups continues to confound research in this area. Available clinical data do not suggest an association between anesthesia and fetotoxicity; prospective studies collecting data on patients undergoing standardized (nonobstetric, nongynecologic) surgical procedures with different anesthetic techniques should be considered to help elucidate the safest anesthetic technique. Preclinical work is also needed to ensure that anesthesia is not associated with more subtle neurodevelopmental effects not investigated in earlier studies.3,85 

Nitrous oxide is used in many types of medical practice, including anesthesia, pain medicine, obstetrics, emergency medicine, and dentistry. The recognition of the potential problems of occupational exposure of anesthetics on health and performance has led to the introduction of occupational exposure limits (OELs; table 2), which are expressed as an 8-h time-weighted average with which scavenging techniques must comply. In addition, in certain countries, a short-term exposure limit to augment the OEL has been adopted to ensure that concentrations beyond a certain threshold are never experienced, even for short periods. The OEL for nitrous oxide of 100 ppm used in the United Kingdom and Sweden has been challenged by the American Conference of Industrial Hygienists, which has advocated an OEL of 50 ppm. The National Institute for Occupational Safety and Health has suggested an OEL of only 25 ppm.

In the prescavenging era, environmental concentrations of nitrous oxide were routinely 1,000–2,000 ppm.103Anesthetic waste gas scavenging has significantly reduced these values in the operating room,104,105though a recent study showed peak nitrous oxide concentrations in excess of 1,000 ppm recorded for short periods.106Some labor ward environments have proved difficult to scavenge efficiently, resulting in midwifery staff being exposed to nitrous oxide levels that exceed the current OELs,107,108although safe levels can be achieved in purpose-built facilities.109 

Inhibition of methionine synthase by nitrous oxide is dose and duration dependent; therefore, it is plausible that environmental nitrous oxide exposure may induce an occupational risk. The majority of animal studies investigating potential effects on reproduction have used very high doses with prolonged administration, limiting the relevance of their results to the clinical setting. Studies of doses of nitrous oxide that are consistent with occupational exposure have shown increased fetal loss and impaired postnatal development with 1,000 ppm but no effect with 500 ppm or lower110,111when administered for the duration of pregnancy. These toxic effects are reduced by folinic acid99and methionine,112indicating the involvement of the methionine synthase pathway. Interestingly, exposure to nitrous oxide up to 10,000 ppm (1%) throughout gestation does not affect the behavioral development of young rats113; again higher doses have an effect,114,115but the clinical relevance of this is unclear.

Nitrous oxide–induced fertility problems do not occur at 1,000 ppm or lower in animals,113although they have been shown to be present at 5,000 ppm in some116but not all studies117(though these differences may represent interspecies variation). Thus, there is preclinical evidence that nitrous oxide can induce fetotoxicity and fertility defects through inhibition of methionine synthase, but there is no evidence that this toxicity occurs with concentrations below the OELs. The animal data suggest that doses of 500 ppm are a threshold for this toxicity (i.e. , 5 times the United Kingdom’s or 10 times the United States of America’s OELs). Furthermore, because rats seem more susceptible to nitrous oxide’s inhibitory effect on methionine synthase than humans do,50the preclinical data suggest that the existing OELs for human exposure are appropriate.

Deriving conclusions from the relevant clinical data is primarily hampered by the retrospective nature of the questionnaire-based evidence, which is therefore confounded by reporter bias. For example, in a large study of occupational exposure, the response rate in the control group was 42%, whereas in the exposed group it was 76%.118Most of the occupational exposure studies regarding anesthetics also predate modern scavenging and operating room ventilation. These studies also addressed exposure to all anesthetic gases and not specifically nitrous oxide, and often the anesthetic gas exposure levels were not quantified. Therefore, the relevance of many of the studies to current practice is doubtful (table 3).

For example, in one questionnaire study of midwifery staff in Sweden, 1,125 second-trimester pregnancies were analyzed, with approximately half of mothers exposed to nitrous oxide.119Nitrous oxide exposure was associated with higher ORs for low-birth-weight (OR, 3.4; 95% CI, 0.9–3.4) and small-for-gestational-age babies (OR, 3.0; 95% CI, 1.2–7.2). However, there were only 43 cases of low birth weight in the whole data set (because of the small sample size). Moreover, the whole data set showed slightly higher birth weight than the national average. Other important confounders include the participation of those with shift work patterns, the lack of quantification of nitrous oxide exposure, and the wide variation in results.119This is important because Axelsson et al.  found an association of spontaneous abortions with night and shift work (OR, 1.63; 95% CI, 0.95–2.81) but not with nitrous oxide exposure (OR, 0.95; 95% CI, 0.62–1.47) involving 1,717 pregnancies in 3,985 midwives.120In this latter study, the nitrous oxide exposure was within OELs.

Boivin121performed a meta-analysis of 19 studies completed between 1971 and 1995 and found a relative risk of spontaneous abortion with nitrous oxide exposure of 1.48 (95% CI, 1.4–1.58). When the data were refined to use only the studies with the best design, the relative risk increased to 1.9 (95% CI, 1.72–2.09). However, meta-analysis is always limited by the data it seeks to analyze; specifically in this case, the studies suffered from poor response rates, reporter bias, inadequate controls, and lack of nitrous oxide exposure data. We therefore consider this an inappropriate method to analyze these data. Notably, in the most highly rated study included in this meta-analysis, 10 occupations were surveyed in Denmark.122In the subgroup exposed to nitrous oxide (dental assistants), a 94% response rate was noted, and the OR for spontaneous abortion was 1.0 (CI, 0.8–1.2) and therefore the group exposed to nitrous oxide was not at increased risk. In another study, Ericson and Kallen123analyzed the official birth registry in Sweden to circumvent the problems associated with reporter bias. While again anesthetic exposure was not quantified, this study design did not reveal any differences in the incidence in perinatal deaths or malformations in anesthesiology/operating room nurses.

In a questionnaire study attempting to address concerns about female fertility, data were sought from 7,000 dental assistants and included an estimation of their exposure to nitrous oxide as well as their time to conception.124However, only 69% responded, and analysis was possible on only 459 replies because of exclusion through insufficient data or predefined criteria. From this limited data set, the authors concluded that unscavenged  nitrous oxide exposure (which they estimated to exceed 1,000 ppm) for greater than 5 h per week was associated with reduced ability to conceive. Exposure to nitrous oxide in scavenged  clinical settings seemed safe. Follow-up data showed that although exposure to nitrous oxide in an unscavenged  setting was associated with an increased risk of spontaneous abortion, this was attributable to a group of 20 dental assistants exposed to 5–9 h of unscavenged nitrous oxide per week. Interestingly, no greater risk of spontaneous abortions occurred with unscavenged  exposure beyond 10 h or after occupational exposure in scavenged settings.125With the caveat that these are retrospective questionnaire data, these studies confirm the occupational safety of nitrous oxide in the scavenged workplace.

Therefore, concerns about the reproductive toxicity of occupational exposure to nitrous oxide at levels below the OELs have not been substantiated by the available data, which unfortunately do not include prospective epidemiologic studies. Only data from adequately powered prospective studies with quantified nitrous oxide exposure levels and appropriate control groups will finally settle this issue; until then, it seems prudent to ensure that occupational exposure levels do not exceed the stated OELs.

Nitrous oxide–induced inhibition of methionine synthase can interfere with transmethylation reactions, folate metabolism, and thereby purine and pyrimidine production (fig. 3). Therefore, genotoxic effects of nitrous oxide are biologically plausible, and there is some in vitro  evidence to suggest that volatile anesthetics or nitrous oxide (50% for 72 h) can induce sister chromatid exchange, a marker of genotoxicity.126Of course, the extreme dose and duration of exposure limit the translatability of these findings to the clinical context. Indeed, there is little clinical evidence of genotoxicity after occupational exposure to nitrous oxide as a sole agent.127More recent studies have focused on mixed anesthetic gas exposure, demonstrating an association between occupational exposure and markers of genotoxicity (sister chromatid exchange, DNA strand breaks, and micronuclei formation) in lymphocytes.128,129In a study of 50 physicians (25 anesthesiologists and 25 unexposed controls), occupational exposure to sevoflurane (8.9 ± 5.6 ppm) and nitrous oxide (119 ± 39 ppm) was associated with increased levels of sister chromatid exchange.130Two months of leave from the operating room resulted in the sister chromatid exchange levels returning to normal, indicating reversibility in the effect. Although the levels exceeded OELs in this study, an earlier study showed a similar effect of occupational exposure to nitrous oxide and isoflurane even when the levels were below the OELs.127Therefore, there is a little evidence to suggest that occupational exposure to nitrous oxide alone induces genotoxic effects; however, when anesthesiologists are exposed to mixed gases (nitrous oxide plus a volatile anesthetic), genotoxic effects are apparent. Again, longitudinal cohort studies, with quantified gas exposure, are required to assess the impact of this occupational exposure on the health of anesthesiology staff.

The effect of occupational exposure on the peripheral nervous system has been studied less intensely, but a questionnaire study (with the previously discussed limitations) of 60,000 dentists and their assistants showed that high exposure (greater than 6 h a week for 10 yr) was associated with neurologic symptoms such as tingling, numbness, and weakness (1.5% vs.  control rate of 0.4%).131However, the unknown incidence of nitrous oxide abuse, lack of scavenging, and responder bias confound interpretation of these data.

A few studies in the 1970s attempted to address the potential for occupational exposure to nitrous oxide (with or without halothane) to affect neurocognitive performance. Exposure to 500 ppm for 4 h was associated with decreased performance on 1 cognitive test, the Digit Span Test, out of 12.132The same group also showed an effect at 50 ppm on audiovisual reaction time in one study133but not at higher concentrations in others. These unblinded results have not been reproduced.134,135Subsequent studies have suggested that somewhere in the region of 5–10% MAC (around 52,000–105,000 ppm) of nitrous oxide is required to affect neurocognitive performance.135 

The hematologic toxic effects occur at analgesic and anesthetic doses and not at levels consistent with occupational exposure. There seems to be little risk from occupational exposure to nitrous oxide, with hematologic abnormalities not apparent below 1,800 ppm,136no effect at 860 ppm,137and some potential effect on lymphocytes at 10,000 ppm (when associated with halothane138), although there are conflicting reports on this issue.139 

Nitrous oxide’s role in anesthesiology has again been challenged by the ENIGMA study; unfortunately, the study’s design prevents clear conclusions being drawn. We support Dr. Hopf’s interpretation that the biologic plausibility for the findings rests primarily with a protective effect of high concentrations of oxygen.6However, nitrous oxide administration is not without risk. Inhibition of methionine synthase in vulnerable populations anesthetized with nitrous oxide may induce hematologic and neurologic effects. Associated increases in plasma homocysteine levels have been correlated with increased perioperative myocardial ischemia in patients anesthetized with nitrous oxide. Although perioperative myocardial ischemia is associated with an increase in long-term mortality,140no study has followed randomized patients receiving either nitrous oxide or nitrogen to determine whether nitrous oxide has a long-term effect on survival. We hope that the ENIGMA II trial is designed to ascertain this information. The reader should also note that we have focused on nitrous oxide’s ability to provoke toxicity; however, nitrous oxide may still have a role in some patients, e.g. , nitrous oxide does not provoke malignant hyperthermia like other volatile anesthetic agents and may be useful in patients at risk from this disorder.141 

Further understanding of the anesthetic and analgesic actions of nitrous oxide will help us to identify clinical situations where it has most benefit. As mentioned previously, its use in the very young29or in combination with a GABAergic anesthetic agent26,33for analgesia  may be confounded, and therefore we may not realize its profound analgesic effects in these settings. However, nitrous oxide is used by many pediatric anesthesiologists for gaseous induction where its nonirritant quality and the second gas effect are unique attributes. Although attempts are being made to find alternatives to nitrous oxide in the labor ward, it seems that the routine use of nitrous oxide in this setting will continue for the foreseeable future; in these settings, attention to OELs is important.

At the recommended OELs, there is no conclusive evidence for reproductive, genetic, hematologic, or neurologic occupational toxicity from nitrous oxide exposure, though the potential genotoxicity recorded from mixed anesthetic gas exposure needs further attention. Facilities to ensure adequate scavenging and ventilation are imperative to ensure the occupational health of medical staff. Good prospective epidemiologic evidence is required to fully evaluate the risks of different anesthetics in the modern operating/anesthetic room environment.

Nitrous oxide has been used for 150 yr for analgesia and anesthesia and in the main has proven safe and efficacious. Exclusion from clinical practice is not warranted with the current level of evidence. Nitrous oxide currently has a niche role as an inhalational analgesic and sedative (table 4). Whether the contraindications for nitrous oxide should go beyond effects on expanding gas-filled spaces and increasing intracranial pressure is a source for future research that we heartily endorse. Finally, nitrous oxide is cheap and will continue to be used in the developing world where healthcare resources are more limited; therefore, further investigations are indicated to define the future of this drug in anesthetic and medical practice.