Cause of CO Poisoning, Relation to Halogenated Agents Still Not Clear

Richard E. Moon, M.D.

The above case report from Florida appears to indicate that also with one of the newest halogenated volatile anesthetics comes the rare (but previously clearly documented) association with very high levels of carbon monoxide in the patient’s blood.

Carbon monoxide (CO) is an odorless, tasteless gas that is usually produced by combustion. Common sources include internal combustion engine exhaust, house fires and barbecues. CO poisoning can also result from inhalation of methylene chloride fumes, as cytochrome P450 in the liver converts this compound to CO. There is a small endogenous production of CO from breakdown of hemoproteins, particularly hemoglobin (M). The toxicity of CO is caused by displacing oxygen from various hemoproteins, including Hb. fib binds CO with an avidity approximately 200 times greater than Oz forming carboxyhemoglobin (COHb). Formation of COHB results in a functional anemia and an additional increase in affinity for 02 of the unbound Hb (‘shift to the left” of the Hb-02 dissociation curve), resulting in decreased 02 delivery and impaired extraction of 02 from Hb at the tissue. Additional mechanisms of toxicity result from CO binding to other hemoproteins such as myoglobin and cytochrome oxidase.

Because of its high affinity for target proteins, CO is toxic even in low concentrations. Breathing 1000 parts per million (ppm) for an hour will typically result in a COHB level of around 30% and evidence of moderately severe poisoning. For an eight hour exposure the Occupational Safety and Health Administration (OSHA) has set the maximum limit at 50 ppm. Using data on the effect of low level CO exposure on anginal threshold in individuals with coronary artery disease, the Environmental Protection Agency (EPA) has set the levels at 35 and 9 ppm for one and eight hour exposures, respectively.

Symptoms and signs of CO poisoning include headache, nausea, vomiting, dizziness, motor weakness, impaired consciousness, cardiac arrhythmias and ischemia. In some instances, particularly if there are neurological abnormalities at the time of the exposure, there can be prolonged or permanent sequelae consisting of cognitive deficits, mood changes, dementia and extra-pyramidal motor abnormalities. CO poisoning can also be fatal by preventing normal oxygen delivery to the tissues.

CO poisoning can be diagnosed by measurement of COHB in peripheral blood. COHB can readily by measured using a four-wavelength cooximeter, and is reported as a percentage of total Hb. A typical non-smoker may normally have 1-2% COHB, derived from endogenously produced CO. Smokers may have around 4 7% COHB. Although there is a poor correlation between COHB level and clinical severity of CO poisoning, generally a COHB level greater than 15-20% is associated with symptoms and greater than 50% with impaired consciousness.

CO poisoning during anesthesia is unlikely to be diagnosed using commonly employed monitors. COHB is not easily detected by dual wavelength devices such as pulse oximeters. Studies in dogs’ and observations in patients (2) have indicated that high COHB levels result in only a trivial reduction in SaO2 measured by pulse oximetry. Detection of gaseous CO is also difficult. Dedicated CO analyzers most commonly use either electrochemical techniques or infrared absorption. However, the infrared absorption spectrum of CO is different from that of C02, and in concentrations likely to be present in cases of CO poisoning (0.05 0.1%), would not significantly alter the reading on clinical capnographs. CO has a molecular weight of approximately 28, and with commonly used mass spectrometers cannot be distinguished from nitrogen. The only reliable method of detection is direct measurement of blood COHB.

Treatment of CO poisoning includes removal from the source of exposure and immediate administration Of 02. Inspired 02 should be as close to 100% as feasible. Hyperbaric oxygen has been recommended for patients with neurological symptoms, cardiac ischemia, pregnancy or high COHB levels.’ In the setting of neurological abnormalities, including temporary loss of consciousness, there is evidence that hyperbaric oxygen may prevent long term sequelae.

Intraoperative Carbon Monoxide Poisoning

The case reported by Dr. Lentz is similar to a number of others which have occurred in at least three other institutions in this country. Our own experience at Duke Medical Center dates back to January of 1990, at which time a 76-year-old nonsmoking female was undergoing general anesthesia for thyroid resection. It is the policy of our Blood Gas Lab to do co-oximetry on all samples sent for blood gas analysis. An arterial catheter had been inserted preoperatively and 25 minutes after anesthesia induction, a routine ABG sample was sent to the laboratory. Carboxyhemoglobin (COHB) level was 9.1%. SaO2 by pulse oximetry was 99 100% throughout the anesthetic. Another blood gas was sent an hour after the first one and the COHB level was 28%. Upon receipt of this result, another sample was sent and the COHB level was 29%. This was confirmed on two other co-oximeters. It was assumed that either the O2 or the N20 supply had been contaminated. The gas supply lines were disconnected from the wall source and the patient was administered 100% O2 from the accessory tank. Subsequent COHB levels were lower. The patient awoke with a headache. Because of the extremely high level and the residual symptom despite 100% 02 administration, she was treated with hyperbaric oxygen with complete resolution of the headache.

Despite the suspicion of source gas contamination, blood values obtained from patients in rooms supplied from the same gas tank lines showed no similar COHB elevation. Immediate measurement of gaseous CO in the supply revealed levels <1 ppm. Additionally, exhaustive investigations in conjunction with the suppliers of 02 and N20 essentially ruled out the possibility of source gas contamination.

The second case became evident about six weeks later when a patient undergoing total hip replacement under general anesthesia had a COHB level of 24.7%. Similar investigations were carried out; no source was found. However, the anesthesia circuit had been left in place and, using an electrochemical CO monitor, it was noted that gas exiting the Sodasorb canister had a CO concentration > 500 ppm. Heating of one of the two soda lime canisters liberated high levels of CO.

Abstract Brings More Reports

A total of eight instances occurred at Duke Medical Center. After publication of an ASA abstract, we were immediately contacted by Dr. Ed Brunner at Northwestern and Dr. Chuck Ingram at Emory, reporting, respectively, three and eighteen similar cases with COHB levels ranging from 8.5 to 32%. Many of the cases had baseline measurements and therefore a documented rise in COHB during anesthesia.

At the 1990 ASA meeting, a meeting was held including representatives of W.R. Grace, manufacturer of Sodasorb, and Anaquest, manufacturer of fluorinated volatile anesthetics. Review of the cases revealed no particular distinguishing characteristics of the anesthetics. Most patients had been administered one of the fluorinated gases with nitrous oxide and some narcotic. One patient had received a spinal anesthetic and had presumably been exposed to CO while receiving supplemental 02 via the anesthetic circuit. There was, however, one interesting factor: most instances had been the first case anesthetized on a Monday morning; all cases occurred in a room which had not been used for at least two days. This raised the possibility that a slow chemical reaction, possibly involving C02 absorbent, was responsible for accumulation of CO within the anesthesia machine.

Although there was no obvious mechanism for the observed cases of CO poisoning, it seemed that the soda lime was probably the site of formation or accumulation, and a set of interim guidelines was developed by Drs. E. Brunner, C. Ingram, A. Meyer and R. Moon.,(4,5) These were: frequent replacement of used soda lime, flushing the soda lime with high flow 02 for one minute prior to each anesthetic and using high fresh gas flow (> 5 I/min). Since implementing these policies no similar cases have occurred at Duke Medical Center.

The guidelines listed above were only intended to be temporary, pending definitive elucidation of the cause. Investigations had begun at Duke Medical Center. While actual cases of CO poisoning were uncommonly discovered, in part because blood gases were measured on only about 10% of patients, ‘footprints’ of the phenomenon, in the form of measurable gaseous CO within unused anesthesia circuits, were relatively common. On Sunday afternoons dangerously high CO levels (> 1000 ppm) within the soda lime compartments of anesthesia machines were detected in over 2% of measurements (320 observations).

Solutions to the Mystery?

Several hypotheses were investigated, aided significantly by an APSF grant awarded for 1993:

Source Gas Contamination: Because of the sporadic occurrence of the cases, and repeated failure to find high significant CO concentrations, this possibility was unlikely.

Chemical Reaction Between Soda Lime and Anesthetics

Studies dating back to the early 20th century indicate that certain chlorinated halocarbons can indeed generate CO when exposed to strong bases. For example, chloroform (CHC13) undergoes a reaction with sodium hydroxide to produce CO, with formate (CHCOO-) as an intermediate.’ In the presence of soda lime trichloroethylene (Trilene: CHC12CH2CI) produces equimolar quantities of phosgene and CO.”‘ (While the phosgene production associated with the use of trichloroethylene in semi-closed circuits, and its clinical toxicity, are well described in the literature, the possibility that CO may have played a role in this syndrome seems to have been overlooked.)

Is it possible that fluorocarbons exposed to soda lime could result in analogous reactions? Studies performed in the 1950s indicated that compounds such as fluoroform (CHF3) could react in an alkaline environment in a manner similar to chloroform but at substantially slower rates.” Studies of modem fluorinated anesthetics indicated that toxic concentrations of CO did not occur when exposed to soda lime. This was confirmed by studies of our own, in which Sodasorb was exposed for a prolonged time to saturated vapors of halothane, enflurane, isoflurane and fluoroform. The highest concentrations of CO observed were after exposure of the absorbent to fluoroform (about 27 ppm), though insufficient to have produced the COHB levels observed in the three institutions.

Contamination of Volatile Anesthetics

Analysis of enflurane from the vaporizer used for one of the Duke cases was performed using a high resolution gas chromatography-mass spectrometry system. Fluorinated compounds other than the main anesthetic were detected in part per billion (ppb) concentrations, but well within the manufacturer’s specifications for purity. Further analysis revealed no detectable formate.

Endogenous CO Production

A report by Middleton published in 196511 demonstrated high CO levels within the anesthetic circuits in patients anesthetized using low flow. The authors attributed these levels to endogenous production of CO from hemoglobin breakdown. COHB measurements were not reported. Accelerated erythrocyte breakdown (e.g. hemolysis, blood transfusion) causes increased endogenous CO production and it is possible that under certain conditions exhaled CO concentrations could reach toxic levels. In particular, breathing 100% 02 results in displacement of CO from Hb and an acute increase in PCO in the exhaled gas. Studies in normal volunteers breathing from an anesthetic circuit have shown that the CO concentration in the circuit can rise to 100 ppm or more. Of course self-poisoning by one’s own endogenously produced CO would be unlikely, though this could provide a mechanism via which adsorption of CO to soda lime, accumulation and then release could occur (see below).

Soda Lime Contamination

Although the soda lime in semi-closed circuits has been implicated in the pathogenesis of ” phenomenon, the mechanism is still not understood and is under active investigation. Formulation does not appear to be important, since both Sodasorb and Baralyme have been in use during the cases collected from the three institutions mentioned above. Analysis of fresh and used Sodasorb samples has revealed traces of formate in some used samples, particularly in ones associated with CO poisoning. Since CO can readily be generated from formate, one has to suspect a possible link. Formic acid is endogenously generated and trace quantities in exhaled gas could be trapped in soda lime, providing a source for CO production. However, in our studies formate concentrations in the exhaled vapor of both normal volunteers and anesthetized patients were orders of magnitude lower than required to explain the observed levels in Sodasorb.

Production of formate in soda lime could also occur from other exhaled substances, such as methane, and is under active investigation. Although energetically unlikely, production of formate could conceivably occur from exhaled CO, perhaps catalyzed by trace amounts of heavy metals in soda lime.

Another possibility is that soda lime could act as a nonspecific adsorbent for CO. Adsorption of halothane, isoflurane, enflurane and sevoflurane to soda lime has been measured and is accentuated if the absorbent is dehydrated.

According to measurements in our laboratory, soda lime can also adsorb CO. Since adsorption is usually inversely proportional to temperature, it is possible that adsorption and accumulation of endogenously produced CO could occur, and then be released if the soda lime temperature is raised. The ‘Monday morning” phenomenon could be explained by slow diffusion of CO from the interior of granules.

Return to Low Flow?

Although high fresh gas flows appear to have played a part in reducing the likelihood of CO poisoning, the additional cost of anesthetics is substantial. At Duke Medical Center recently, the policy has been changed to remove the restriction on fresh gas flow rate, while continuing to monitor weekend CO levels. If the distribution of CO levels does not indicate greater numbers of machines with dangerous CO concentrations it may be possible to remove this most costly of the three 1990 guidelines


Because of its episodic nature, the elucidation of the cause of this rare but potentially fatal phenomenon has been difficult to establish. Since it has not (as yet) been reproducible in the laboratory, it is likely to be in part due to interaction between soda lime and some component of exhaled gas from patients. Carbon monoxide poisoning cannot be detected using standard anesthesia monitors. The guidelines for prevention, listed in the text, appear to be effective. Treatment of CO poisoning should include removal from the source, administration of 100% 02 and if neurological symptoms or signs exist, hyperbaric oxygen.

Dr. Moon, recipient of a 1993 APSF Research Grant,

is from the Department of Anesthesiology, Duke University Medical Center, Durham, NC


1. Barker SJ, Tremper KK: The effect of carbon monoxide inhalation on pulse oximetry and transcutaneous P02. Anesthesiology 66:677,1987

2. Gonzalez A, Gomez-Arnau J, Pensado A: Carboxyhemoglobin and pulse oximetry. Anesthesiology 73:573, 1990

3. Piantadosi CA. The role of hyperbaric oxygen in carbon monoxide, cyanide and sulfide intoxication. Probl Respir Care 4:215-231,1991

4. Centers for Disease Control. Elevated intraoperative blood carboxyhemoglobin levels in surgical patients Georgia, Illinois and North Carolina. MMWR 40:248-249, 1991

5. “Safety Persists as ASA Meeting Theme.” APSF Newsletter 1991;6:42

6. Kemp DS, Vellaccio F. Organic Chemistry. New York: Worth, 1980, p. 1181

7. Carden S. Hazards in the use of the closed-circuit technique for Trilene anaesthesia. Brit Med 1 1:319-320, 1944

8. Hunter AR. Complications of Trilene anaesthesia. Lancet 1:308-309,1944

9. McClelland M. Some toxic effects following Trilene decomposition products. Proc Roy Soc Med 37:526-528, 1944

10. Firth JB, Stuckey RE. Decomposition of Trilene in closed circuit anaesthesia. Lancet 1:814-816,1944

11. Hine J, Dowell AM Jr, Singley JE. Carbon dihalides as intermediates in the basic hydrolysis of haloforms. IV. Relative reactivities of haloforms. JACS 78-.479482,1956

12. Middleton VA, Van Poznak A, Artusio JF, Smith SM. Carbon monoxide accumulation in closed circle anesthesia systems. Anesthesiology 26:715-719,1965