Mechanistic Aspects of Carbon Monoxide Formation from Volatile Anesthetics 

Enflurane, isoflurane, desflurane (Ohmeda), sevoflurane, methoxyflurane (Abbott Laboratories, Abbot Park, IL), halothane (Ayerst Laboratories, New York, NY), and chlorotrifluoroethylene (Lancaster, Wyndham, NH) were used as supplied. Anesthetic structures are provided in Figure 1. Sodium and potassium deuteroxide (NaOD, KOD; 40% in D²O), Aliquat 336, sodium carbonate, difluorodibromomethane, alpha-methylstyrene, zinc metal, iodine, and 2-methoxyethyl ether (diglyme, anhydrous) were obtained from Aldrich (Milwaukee, WI). Reagents were the highest quality available and were used without further purification. Deuterium oxide (D²O, 99.9% D) and water enriched with¹⁸O (>or= to 70%) were from Cambridge Isotope Laboratories (Andover, MA). Solutions (1.0 M) of NaOD and KOD were prepared by dilution with D²O. Ba(OD)²[middle dot] 8D²O was prepared by adding 1.5g BaO to 4 ml D²O. Then 0.173g Ba(OD)²[middle dot] 8D²O was added to 1 ml D²O to make a 1 M solution. Barium hydroxide lime (Baralyme, Chemtron Medical Division, Allied Healthcare, St. Louis, MO) and soda lime (Sodasorb, W. R. Grace & Co., Cambridge, MA) were dried to constant weight. [10]A CO standard of 952 ppm in air (Matheson Gas Products, Montgomeryville, PA) was used to prepare calibration curves for quantitative analysis.

Perdeuterated and selectively deuterated enflurane were synthesized by base-catalyzed exchange, as described previously. [12]Analysis by gas chromatography-mass spectrometry (GC-MS) showed deuterium incorporation on the difluoromethyl and chlorofluoroethyl ([Greek small letter beta]-ethyl) carbons was: DF²C-O-CF^2-CDC1F(perdeuterated, 90 and 97%), DF²C-O-CF^2-CHC1F(methyl deuterated, 92 and 13%), HF (²) C-O-CF^2-CDC1F(ethyl deuterated, 13 and 94%). Perdeuterated desflurane was prepared similarly, however, it was necessary to substitute 0.1 M NaOD and 0.1 M NaOH to synthesize the monodeuterated forms. Analysis by GC-MS showed that deuterium incorporation on the difluoromethyl and [Greek small letter alpha]-ethyl carbons was DF²C-O-CDF-CF³(perdeuterated, 86 and 89%), DF²C-O-CHF-CF³(methyl deuterated, 62 and 14%), and HF²C-O-CDF-CF³(ethyl deuterated, 48 and 92%). The smaller difference in relative acidities of the two carbons of desflurane, compared with enflurane, precluded more selective monodeuteration.

A modification of previous methods was used. [13]Zinc metal was activated by sequentially washing powdered zinc in a filter funnel with dilute HCl, water, acetone, and diethylether, dried under high vacuum to remove residual solvent, stored in a sealed container, and used within two days of activation. A stir-bar and activated zinc (1.5 g; 22.8 mmol) were placed into a dry 250-ml flask and 2 to 3 crystals of iodine were added. The flask was then sealed with a rubber septum and purged with nitrogen. Diglyme (40 ml) and [Greek small letter alpha]-methylstyrene (0.9 g, 7.6 mmol) were added via syringe. The last 10 ml of diglyme was mixed with difluorodibromomethane (4.78 g, 22.8 mmol) and transferred to the reaction flask slowly over 10 min. The reaction was stirred on ice for 24 hr. Two volumes of ether were added, the mixture was filtered through Whatman #40 paper, and the filtrate evaporated to a white film that was redissolved in ether. The final product (54% yield) was characterized by GC-MS and¹⁹F NMR. GC-MS:(t^R= 20.1 min; see below for chromatographic conditions) m/z 168 ([M]⁺, 28%), 153 ([M-CH³]⁺, 100%), 133 ([M-CH^3-HF]⁺, 35%), 103 ([M-CH^3-CF2]⁺, 18%), 77 (Ph⁺, 15%), 51 (CF²H, 9%); Lit: m/z 168 (78%), 153 (100%)[14] ¹⁹F NMR:(CDCl³, [Greek small letter delta] relative to hexafluorobenzene) 24.4 ppm (dd, J^FF= 150 Hz, J^FH= 12 Hz), 27.5 ppm (dd, J^FF= 151 Hz, J^FH= 13 Hz) Lit: dd, J^FF= 150 Hz, J^FH= 13 Hz; dd, J (^FF)= 150 Hz, J^FH= 12 Hz. [13] 

Experiments were performed in triplicate at room temperature (23 [degree sign]C) in sealed 20.7-ml vials containing 4 g desiccated barium hydroxide lime or soda lime and a small strip of Whatman #1 filter paper, onto which anesthetics were injected to promote volatilization and prevent direct contact between anesthetic and absorbent. For each replicate a pair of vials was prepared and samples were collected alternately to avoid oversampling of the headspace volume from a single vial. A blank gas sample (0.5 ml) was drawn from one vial of each pair and injected into a 12-ml sealed headspace vial for analysis. Anesthetic (either at equimole or equiMAC concentration) was then injected into each 20-ml vial, the vials were placed on a wheel revolving at 60 rpm, and a 0.5 ml headspace sample was drawn at timed intervals (1, 4, 7, 10, 20, 30, 45, 60, 120, 180, 240 and 300 min after anesthetic introduction alternating between the vial pairs for each replicate) and injected into a fresh, sealed headspace vial. Carbon monoxide was quantified by GC-MS, as described previously. [10]Results were not corrected for minor dilution caused by repetitive sampling. The detection limit was approximately 2 ppm and because all samples were diluted 24-fold, the actual detection limit was approximately 50 ppm. For experiments at equivolume (approximately equimole concentrations), 5 [micro sign]l of each anesthetic was injected, yielding desflurane (38 [micro sign]mol), enflurane (41 [micro sign]mol), isoflurane (39 [micro sign]mol), sevoflurane (38 [micro sign]mol), methoxyflurane (43 [micro sign]mol), and halothane (47 [micro sign]mol). For experiments at equiMAC concentrations (1.5 MAC), the following amounts were added: desflurane (10.2 [micro sign]l), enflurane (2.6 [micro sign]l), isoflurane (2.0 [micro sign]l), sevoflurane (3.4 [micro sign]l), methoxyflurane (0.3 [micro sign]l), and halothane (1.0 [micro sign]l). Experiments with deuterated anesthetics were conducted with 5 [micro sign]l anesthetic, and CO formation was determined after 300 min. Anesthetic concentrations in the vials were confirmed by GC-MS and were within 5% of predicted.

Experiments were performed to evaluate the source of the oxygen atom in the CO resulting from anesthetic degradation. Absorbent labeled with (¹⁸) O was prepared by rehydrating desiccated barium hydroxide lime (2 g) with 300 [micro sign]l H²¹⁸O, then again drying to constant weight. This was repeated three more times, followed by final desiccation to constant weight. Formation of CO was determined as described before after adding 5 [micro sign]l enflurane or desflurane and sampling 1, 3, and 5 hr after anesthetic addition. Samples were analyzed for C¹⁶O and C¹⁸O by GC-MS, monitoring m/z 12 and 30, respectively. In addition, m/z 16, 18, and 28 were also monitored.

Base-catalyzed proton abstraction from the methyl, ethyl, or isopropyl carbons of the volatile anesthetics was determined by deuterium exchange. Sodium, potassium or barium deuteroxide (1.0 M, 1.0 ml) and 13 [micro sign]l of the phase transfer catalyst Aliquat 336 were added to a 20-ml vial. The vial was sealed and 100 [micro sign]l unlabeled anesthetic was added through the septum via syringe. A 0.5-ml headspace sample was immediately drawn and injected into another sealed vial. Experimental vials were continuously shaken at 23 [degree sign]C, and headspace samples were withdrawn 5, 10, 20, 30, 45, 60, and 90 min after anesthetic addition. The extent of deuterium incorporation into the anesthetics was determined by GC-MS, using a Hewlett-Packard (Wilmington, DE) 5890 II+ GC/5971 mass selective detector with 7694 headspace sampler and a DB-VRX column (30 m x 0.32 mm x 1.8 [micro sign]m; J&W Scientific, Folsom, CA). The GC injector and detector temperatures were 150 and 250 [degree sign]C, respectively, and the column head pressure was 2.5 psi. For anesthetic analysis, the headspace sampler parameters were high agitation, 0.5-min sample equilibration, 0.05-min vial pressurization, 0.35-min loop fill, 0.15-min loop equilibration, 0.35-min sample injection time = 0.35 min, and oven, loop, and transfer line temperatures were 70, 75, and 85 [degree sign]C, respectively. The GC oven temperature was held at 30 [degree sign]C for 3 min, increased by 40 [degree sign]C/min to 185 [degree sign]C, and held at 185 [degree sign]C for 5 min. The following ions were monitored: desflurane m/z 52/51 [CF²D]⁺/[CF²H]⁺and 102/101 [CFDCF³]⁺/[CFHCF³](⁺); enflurane m/z 52/51 [CF²D]⁺/[CF²H]⁺and 118/117 [CF²CDC1F]⁺/[CF²CHC1F]⁺; isoflurane m/z 52/51 [CF²D]⁺/[CF²H]⁺and 118/117 [CD-C1CF³]⁺/[CHC1CF³]⁺; sevoflurane m/z 34/33 [CFDH]⁺/[CFH²]⁺and 152/151 [C(CF³)²D]⁺/[C(CF³)²H]⁺; halothane m/z 118/117 [CF³CC1D]⁺/[CF³CC1H]⁺; methoxyflurane m/z 82/81 [CF²OCDH²]⁺/[CF²OCH³]⁺and 84/83 [CDC1 (²)]⁺/[CHC1²]⁺. Proton abstraction was determined by deuterium incorporation into the anesthetic, calculated by dividing the peak height of the (M + 1)⁺ion by the sum of the peak heights of the M⁺and (M + 1)⁺ion and multiplying by 100 to obtain percent exchange. Baseline levels of the M⁺and natural isotope abundances (M + 1)⁺ions were determined for all the anesthetics and subtracted from the experimental values.

The hypothesis that difluorocarbene is a reactive intermediate in CO formation was tested. Specifically, experiments sought to trap and to identify this reactive intermediate. A 12-ml vial containing 0.5 g desiccated barium hydroxide lime, 10 [micro sign]l Aliquat 336, and a stir-bar was sealed with a rubber septum, then tetrahydrofuran (2 ml) and the carbene trap [Greek small letter alpha]-methylstyrene (65 [micro sign]l) were added via syringe. A blank sample (25 [micro sign]l) was withdrawn and 270 [micro sign]l desflurane was added via syringe. For enflurane, a 25-ml flask containing 1.5 g desiccated barium hydroxide lime, 30 [micro sign]l Aliquat 336, and a stir-bar was sealed, then diglyme (6 ml) and [Greek small letter alpha]-methylstyrene (160 [micro sign]l) were added, a blank sample was withdrawn, and 900 [micro sign]l enflurane was added via syringe. Reaction mixtures were stirred before withdrawing a sample, and heated to facilitate [Greek small letter alpha]-methylstyrene reaction with any formed intermediate. Duplicate liquid samples (25 [micro sign]l) were withdrawn 10, 20, 30, 45, 60, 75, 90, 120, 150, 180, and 240 min after anesthetic addition and injected into a sealed headspace vial. All samples were analyzed by electron ionization GC-MS for the presence of the difluorocarbene adduct 1,1-difluoro-2-methyl-2-phenylcyclopropane. Analyses were performed on the described headspace GC-MS using a DB-VRX column. The GC injector and detector temperatures were 150 and 250 [degree sign]C, respectively and the column head pressure was 2.5 psi. The headspace sampler parameters were high agitation, 3.0-min sample equilibration time, 0.05-min vial pressurization, 0.4-min loop fill, 0.15-min loop equilibration, 0.4-min sample injection, and oven, loop, and transfer line temperatures were 110, 115, and 120 [degree sign]C, respectively. The GC oven was 30 [degree sign]C for 3 min, increased by 10 [degree sign]C/min to 200 [degree sign]C, and held for 5 min. The difluorocarbene adduct eluted at 20.1 min. For time course experiments, selected ion monitoring at m/z 133, 153, and 168 was used.

Results are expressed as the mean +/− standard deviation. Correlations were examined by linear least squares regression analysis.

Carbon monoxide formation catalyzed by desiccated barium hydroxide lime and soda lime was measured at equiMAC and at equivolume (approximately equimole concentrations). The former is more clinically relevant, whereas equimole concentrations are appropriate for mechanistic inferences. The rank order of CO formation at equiMAC concentrations was desflurane > enflurane > isoflurane with barium hydroxide lime (Figure 2A), and desflurane = enflurane > isoflurane with soda lime (Figure 2B). The rank order of CO formation at approximately equimole concentrations was enflurane > desflurane > isoflurane with barium hydroxide lime (Figure 3A) and soda lime (Figure 3B). No CO was detectable from sevoflurane, halothane or methoxyflurane using either carbon dioxide absorbent.

Carbon monoxide concentrations from barium hydroxide lime degradation of perdeuterated or selectively deuterated enflurane and desflurane were determined to assess the presence and magnitude of an isotope effect (Figure 4). Carbon monoxide formation from enflurane after 5 hr was reduced by approximately 50% by perdeuteration or by selective deuteration of either the difluoromethyl or the [Greek small letter beta]-ethyl carbon. Carbon monoxide formation from desflurane was also reduced by perdeuteration or selective deuteration of the difluoromethyl or [Greek small letter alpha]-ethyl carbon. The magnitude of the isotope effect was less than with enflurane, consistent with the lower deuterium incorporation in desflurane. Effects of perdeuteration and selective deuteration were similar, both for enflurane and desflurane.

Proton (methoxy or ethyl) abstraction is a likely initial step in anesthetic degradation to CO. The ability of the strong bases in carbon dioxide absorbents to abstract a proton from the volatile anesthetics, reflecting relative hydrogen acidities, was measured by deuterium exchange. Anesthetics, strong base, and a phase transfer catalyst were reacted in deuterium oxide. Recombination of an anesthetic anion with a deuterium atom from the medium reflects proton abstraction by the strong base.

Deuterium exchange on the methoxy carbon of enflurane, desflurane, isoflurane, sevoflurane, and methoxyflurane catalyzed by potassium (Figure 5A) and sodium (Figure 5B) deuteroxide was examined first. The rank order of proton abstraction was enflurane > desflurane > isoflurane for both potassium and sodium deuteroxide. No proton abstraction from the methoxy carbon of sevoflurane or methoxyflurane was observed with either base. Proton abstraction was greater by potassium than by sodium deuteroxide for enflurane, but was similar for desflurane and isoflurane. No methyl group proton abstraction by barium deuteroxide was observed for any anesthetic.

Deuterium exchange on the ethyl carbon of enflurane, desflurane, isoflurane, methoxyflurane, and halothane, and the isopropyl carbon of sevoflurane, catalyzed by potassium and sodium deuteroxide, is shown in Figure 6A and Figure 6B, respectively. In contrast to methoxy protons, abstraction of the ethyl (or isopropyl) proton from all anesthetics was rapid and extensive, with no substantial differences observed between most drugs. The exception was proton abstraction from desflurane by sodium deuteroxide, which was less extensive. No proton abstraction by barium deuteroxide was observed for any anesthetic.

Relationships were examined between CO formation catalyzed by carbon dioxide absorbents and proton abstraction by the strong bases that they contain (Figure 7). There was a significant linear correlation between the extent of CO formation with barium hydroxide lime and methyl carbon deuterium exchange catalyzed by potassium deuteroxide (r = 1.0, P < 0.001)(Figure 7A). Similar correlations were found between the extent of CO formation catalyzed by soda lime and the extent of deuterium exchange catalyzed by potassium and sodium deuteroxide (r = 0.99, P < 0.001). Initial linear rates of CO formation and proton abstraction from the methyl carbon were also compared (Figure 7B). There was a significant correlation between the rates of CO formation with barium hydroxide lime, and deuterium exchange catalyzed by potassium deuteroxide, and linear correlations between the rates of CO formation by soda lime and rates of deuterium exchange catalyzed by both strong bases (r=0.95-1.000). No such correlations were found between CO formation and deuterium exchange from the ethyl carbon of any anesthetic, with either absorbent or either strong base (not shown).

Trapping experiments tested the hypothesis that difluorocarbene is an intermediate in the base-catalyzed decomposition of difluoromethoxy-containing anesthetics, analogous to the reaction mechanism affording carbon dihalides as intermediates in the base-catalyzed hydrolysis of halomethanes, [15]which are identified by trapping of the reactive intermediate. [16][Greek small letter alpha]-Methylstyrene was used to trap the putative reactive difluorocarbene before it could combine into a more stable species. [13] Figure 8A shows the spectrum of synthetic 1, 1-difluoro-2-methyl-2-phenylcyclopropane (t^R20.1 min), the stable product of difluorocarbene reaction with [Greek small letter alpha]-methylstyrene (reaction shown in Figure 8D, inset). Analysis by GC-MS of the reaction between desiccated barium hydroxide lime and enflurane or desflurane showed a peak at 20.1 min, which was not present in the absence of anesthetic. The mass spectrum of this peak is shown in Figure 8B (enflurane) and Figure 8C (desflurane). Based on identical retention times and mass spectra of the trapped product and synthetic standard, the trapped intermediate was identified as difluorocarbene. Using selected-ion mode to improve sensitivity (Figure 8D), time-dependent formation of the difluorocarbene adduct trapped during enflurane degradation by barium hydroxide lime was demonstrated (Figure 8E).

The oxygen in CO arising by anesthetic degradation may theoretically originate from the anesthetic, the absorbent, molecular oxygen, or water. The source of such oxygen is unknown, yet its identification can provide inferences into the mechanism of CO formation. Experiments tested the hypothesis that oxygen in CO derives from the absorbent, by labeling the absorbent with H²¹⁸O and measuring C¹⁸O formation. The ion chromatogram of a C¹⁶O standard is provided in Figure 9A. Carbon monoxide was detected by monitoring m/z 12 (the carbon fragment, which represents approximately 1.6% of the abundance of the m/z 28 ion of CO). This ion was routinely monitored because m/z 28 is confounded by nitrogen (m/z 28). The small peak at m/z 30 is attributed to the natural abundances of¹³C (1.1%),¹⁷O (0.04%), and¹⁸O (0.2%) in CO (i.e.,¹³C¹⁷O and C¹⁸O), and the ratio of the m/z 30 and 12 peaks observed is consistent with this natural abundance. Figure 9B shows the ion chromatogram of CO formed from enflurane and desiccated barium hydroxide lime. The chromatogram of CO from desflurane and desiccated barium hydroxide lime was similar (not shown). The relative abundance of the m/z 30 and 12 peaks is similar to that for the C¹⁶O standard. The ion chromatogram of CO formed from either enflurane or desflurane, and desiccated barium hydroxide lime labeled with H²¹⁸O, is shown in Figure 9C and Figure 9D, respectively. The ratio of m/z 30 to 12 is considerably greater than in Figure 9A and Figure 9B, evidence that considerably more C¹⁸O is present than in either the C¹⁶O standard or the CO formed from enflurane or desflurane and native (i.e., H²¹⁶O-containing) absorbent. Thus, the oxygen in Co originates, at least in part, from oxygen in the CO²absorbent.

Chloroform is degraded to CO by strong bases in carbon dioxide absorbents, a specific example of the well-described generation of CO from numerous halomethanes. [17]Base-catalyzed proton abstraction and formation of dihalocarbene intermediates are key steps in this reaction. [15,16,18]Carbenes are uncharged divalent carbon intermediates; one of four types containing carbon in a different valence state (the others are trivalent: carbonium ions, carbanions, and free radicals). [17]By analogy, we tested the hypothesis that anesthetic degradation to CO proceeds via base-catalyzed proton abstraction and a dihalocarbene intermediate and examined several component steps in the putative pathway.

Experiments tested the hypothesis that specific structural features are required for anesthetic degradation to CO. Previous reports [4,9]suggested that CO formation results from an unspecified absorbent action on the difluoromethoxy carbon of the methylethyl ethers desflurane, enflurane, and isoflurane, because this moiety is not present on halothane or sevoflurane (which form minimal or undetectable amounts of CO). [5,9]Nevertheless, halothane (a simple alkane) and sevoflurane (a methyl-isopropyl ether) also differ in other ways from these methyl-ethyl ethers. A methyl-ethyl ether without difluoromethyl substituents was required to test the hypothesis. Methoxyflurane, with an ethyl group similar to that of enflurane but a nonfluorinated methyl carbon, was ideally suited for this purpose. Methoxyflurane degradation to CO, either by soda lime or barium hydroxide lime, was undetectable. Carbon monoxide formation from sevoflurane and halothane was also undetectable, confirming previous reports. [5]Thus, a difluoromethyl ether function (or possibly any dihalomethyl ether group) does appear to be required for halogenated anesthetic degradation to CO by carbon dioxide absorbents.

Experiments also tested the hypothesis that specific structural features are required for proton abstraction. Neither potassium nor sodium hydroxide catalyzed any measurable proton abstraction from methoxy (methoxyflurane) or monofluoromethoxy (sevoflurane) groups. Thus, difluoro substitution (or possibly any dihalo substitution) appears to be required for methoxy proton abstraction and CO formation.

The more specific hypothesis that base-catalyzed difluoromethoxy proton abstraction is an initial step in CO formation and that differences in abstraction between anesthetics may influence CO formation was also tested. Results are consistent with this hypothesis. There were excellent correlations between absorbent-catalyzed CO formation and methoxy proton abstraction by the strong bases that they contain. This relationship, however, applied only to methoxy protons. The presence of any abstractable proton per se was not sufficient for CO formation. Halothane, sevoflurane, and methoxyflurane contain a highly abstractable (ethyl or isopropyl) proton yet formed no detectable CO. Also, there no correlation between CO formation and ethyl proton abstraction from desflurane, enflurane, and isoflurane, although such a correlation might not occur if carbanion decay to difluorocarbene is rapid relative to the rate of initial proton abstraction (vide infra).

Additional evidence for initial proton abstraction derives from experiments with deuterium-substituted anesthetics. The stronger carbon-deuterium bond is more resistant to cleavage than the carbon-hydrogen bond. Deuterium substitution on the difluoromethoxy carbon of enflurane and desflurane significantly diminished CO formation, and isotope effect magnitude (1.6-1.8 for enflurane) was similar to that for base-catalyzed proton abstraction from other halomethanes that are degraded to CO. [19-21]This supports a role for difluoromethoxy proton abstraction in CO formation. However, an isotope effect was also observed for ethyl carbon deuterium substitution. One interpretation of this result is that [Greek small letter alpha]- or [Greek small letter beta]-ethyl carbon proton abstraction also leads to CO formation. However, not all results support such a mechanism (vide infra). An alternate explanation for an ethyl carbon deuterium isotope effect is that vinyl ethers also undergo difluoromethoxy proton abstraction and CO formation (indeed, because of resonance effects they appear more susceptible than their parent compounds). [22]Ethyl carbon deuterium substitution would diminish vinyl ether formation, and, hence, vinyl ether-dependent CO production. This speculation awaits haloalkene synthesis and experimental confirmation.

The strongest evidence for dihalocarbenes in chemical reactions derives from experiments in which the intermediate is trapped, usually by olefins. [17,23]Such convincing evidence for difluorocarbene formation from difluoromethoxy ether anesthetics was obtained by trapping, with [Greek small letter alpha]-methylstyrene, the putative difluorocarbene intermediate formed from enflurane and desflurane by barium hydroxide lime. Difluorocarbene formation from other halogenated compounds has been demonstrated previously, [21,24,25]to which difluorocarbene generation from fluoromethyl anesthetics may be similar.

Stoichiometry between anesthetic degradation and CO formation was not pursued in these experiments. Strong bases degrade anesthetics both to CO and to haloalkenes, [26,27]thus anesthetic degradation will not be reflected stoichiometrically by CO formation. Moreover, the ratio of products (CHF³, CO, formate) varies greatly under different conditions, [4]in part because CO may react further. [18]This additionally complicates interpretation of stoichiometry.

By analogy to CO formation from halomethanes, [15,16,18]the present results are consistent with a mechanism of CO formation from desflurane, enflurane, and isoflurane (Figure 10). These difluoromethyl ethers undergo rapidly reversible base-catalyzed difluoromethyl proton abstraction to yield a difluoromethyl ethyl ether carbanion.[double vertical bar] This carbanion, in the presence of sufficient water, would simply reprotonate to regenerate the original anesthetic. In the absence of sufficient water, the carbanion could eliminate a halogen anion from the ethyl moiety [Greek small letter alpha] to the oxygen and decompose to difluorocarbene and the corresponding aldehyde. The difluorocarbene can subsequently react with hydroxide and/or residual water# in the absorbent to form CO, via difluoromethanol and formyl fluoride intermediates. [28-30]Difluorocarbene could also react with CO²or silica in the absorbent to form CO. [30,31]Formate may arise from formyl fluoride reaction with water or by subsequent reactions of CO. [18]For desflurane and isoflurane, fluoride and chloride elimination, respectively, from the [Greek small letter alpha]-ethyl carbon would yield trifluoroacetaldehyde. Trifluoroacetaldehyde, in the presence of strong base, may further decompose to formate and trifluoromethane. [32]Trifluoroacetaldehyde may also decompose to CO and trifluoromethane, but this seems unlikely. [33]Additional reactions and alternative mechanisms are also possible, however, these are more speculative. Difluorocarbene, in the absence of water, may combine with itself to form tetrafluoroethylene. [23,34]A concerted reaction may occur, with deprotonation and direct difluorocarbene formation occurring simultaneously, rather than initial carbanion formation and subsequent decomposition. [20,24]This has been seen with difluoromethanes [20,24,35]and dichloromethyl methyl ether [36]but not all difluorocarbene-forming compounds. [25]A concerted mechanism, however, is not supported by the ability (Figure 5) to incorporate deuterium. [20] 

The difluorocarbene mechanism permits rationalizing the structural features of anesthetic degradation to CO. Why is a difluoromethyl ether a common structural element in CO formation? First, initial base-catalyzed abstraction is dictated by methoxy proton acidity, and these are substantially increased by two fluorine substituents. Second, dihalocarbanion decomposition to a dihalocarbene is favored in the absence of water, and fluorine is best at stabilizing difluorocarbenes (F >or= to Cl > Br > I). [17,37]Thus, two fluorine substituents on the methoxy carbon increases both initial proton abstraction and subsequent carbanion decomposition to a dihalocarbene. [17,37]The above mechanism can also account for the observations that formate is found in soda lime after anesthetic degradations to CO,** and that trifluoromethane is formed concomitantly with CO from desflurane and isoflurane but not enflurane. [4,38] 

In addition to methoxy halogen substituents, ethyl halogen substituents may also influence anesthetic degradation to CO, because desflurane, enflurane, and isoflurane possess identical difluoromethoxy groups yet differ in extent of CO formation. This difference is attributable in part to [Greek small letter alpha]- and [Greek small letter beta]-ethyl halogen substituent influence on difluoromethoxy proton susceptibility to base-catalyzed abstraction. [39]Thus, substitution of the [Greek small letter alpha]-ethyl carbon with two fluorines renders the difluoromethoxy hydrogen of enflurane most acidic, compared with one [Greek small letter alpha]-ethyl fluorine on desflurane and one less electronegative [Greek small letter alpha]-ethyl chlorine on isoflurane. Differences in methoxy proton acidity (enflurane > desflurane > isoflurane) were reflected in proton abstraction and CO formation (Figure 3and Figure 5). Ethyl halogen substituents, specifically fluorine, can also influence CO formation by double bond-no bond resonance stabilization. [22,39]Thus, the relative reactivities for proton abstraction and fluorocarbanion stability are 3 [degree sign] > 2 [degree sign] > 1 [degree sign] hydrofluorocarbons, and fluorine substitution on the carbon [Greek small letter beta] to the carbanion is most important. [39]Consistent with this consideration, difluoromethyl isopropyl ether (lacking any fluorine substitution except that on the methoxy group) does not undergo base-catalyzed deprotonation and difluorocarbene formation. [21] 

The influence of water in carbon dioxide absorbents on CO formation merits mention. Fully hydrated or rehydrated absorbents do not degrade anesthetics to CO, and CO formation is inversely proportional to absorbent water content. [5,6,9,10]By analogy to base-catalyzed proton abstraction from haloforms, which is rapid and reversible (by water) compared to subsequent carbanion decomposition and dihalocarbene formation, [16,17]base-catalyzed methoxy proton abstraction from desflurane, enflurane, or isoflurane also appears rapid and reversible. Anesthetic carbanion reprotonation by water to regenerate parent anesthetic would compete with disfluorocarbene formation. Thus, as absorbent water content diminishes, this competition would shift in favor of difluorocarbene formation. The exact role of absorbent water in difluorocarbene decomposition to CO is less certain. Experiments with H²¹⁸O clearly showed absorbent¹⁸O as a source of the oxygen in CO (Figure 9). Because, however, desiccated absorbents could contain both OH^-and some residual water, the added¹⁸O water could react as H²¹⁸O^or18OH^-, or possibly as¹⁸O silicates, to form C¹⁸O; the current results do not differentiate between these sources of oxygen. Dichlorocarbene reacts exclusively, or almost exclusively, with water and not hydroxide ions, [17,18]and difluorocarbene formation under anhydrous conditions does not give CO. [24]Although this supports a role for water in CO formation, oxide,[dagger, dagger] hydroxide, or silicates might also form CO. Any residual water in desiccated absorbents, although sufficient to hydrolyze difluorocarbene and form CO, appears insufficient to shift equilibrium toward carbanion reprotonation and thereby prevent difluorocarbene formation.

An alternative mechanism for CO formation involves proton abstraction from the ethyl, rather than the methyl, carbon of desflurane, enflurane, and isoflurane. For desflurane, the resulting carbanion could (1)[Greek small letter alpha]-eliminate difluoromethanol (yielding CO) to give fluorotrifluoromethylmethylene (hydrolyzing to trifluoroacetaldehyde), or (2)[Greek small letter alpha]-eliminate fluoride to give difluoromethoxytrifluoromethylmethylene (hydrolyzing to trifluoroacetaldehyde and difluoromethanol, yielding CO). An analogous scheme for isoflurane could be proposed. For enflurane, [Greek small letter beta]-elimination of This alternative, or coexistent mechanism, is consistent with the deuterium isotope data (Figure 4). In contrast, this mechanism is not supported by (1) inconsistencies between ethyl proton abstraction (enflurane = isoflurane >or= to desflurane) and CO formation (enflurane >or= to desflurane > isoflurane), (2) the formation of a difluorocarbene intermediate from enflurane and desflurane, and (3) the inability to observe chlorotrifluoroethylene formation specifically from enflurane (results not shown). A variation of the above mechanism, for desflurane, proposed nucleophilic attack by OH^-at the [Greek small letter alpha]-ethyl carbon, leading to a difluoromethyl anion and trifluoroacetyl fluoride, further sequential nucleophilic attacks at [Greek small letter alpha]-chloride (sic), and [Greek small letter alpha]-fluoride sites to form trifluoroacetic acid then an [Greek small letter alpha]-dihydroxyl compound, and further nucleophilic attack at the terminal hydroxyl to form CO. [40]This mechanism however is also not supported by difluorocarbene formation and difluoromethoxy requirements for CO formation and does not account for enflurane degradation to CO, absence of CO formation from methoxyflurane, or trifluoromethane formation from desflurane and isoflurane. It is also energetically unfavorable:[Greek small letter alpha]-ethyl deprotonation would more likely lead to dehydrofluorination than to a difluoromethyl anion, hydroxylation (initial and even more so subsequent) of a trifluoromethyl group requires vigorous conditions not occurring in absorbents (whereas base-catalyzed decarboxylation of trifluoroacetic acid is more likely [41]); there is no [Greek small letter alpha]-chloride in desflurane. At present, both methoxy- and ethyl-proton abstraction mechanisms are plausible, but more data support the former mechanism.

Previous investigations have shown that the magnitude of CO production by carbon dioxide absorbents from equiMAC anesthetic concentrations is desflurane > enflurane > isoflurane, whereas halothane and isoflurane produced minimal or no CO, both in vitro [4,9,11]and in anesthesia machines. [5]The current results corroborate these findings. Furthermore, they indicate that greater CO formation from desflurane than enflurane at equiMAC concentrations results from a difference in anesthetic potency. Although at equimole concentrations enflurane produces more CO, the lower potency of desflurane (6% MAC) versus enflurane (1.7%) necessitates much higher clinical desflurane concentrations, and hence, greater CO formation ensues.

Carbon monoxide formation from desflurane, enflurane, and isoflurane was greater with barium hydroxide lime than with soda lime, both in vitro [4,9]and in anesthesia machines. [4-6]The current results confirm these findings and provide an explanation: base-catalyzed difluoromethoxy proton abstraction was greater with potassium than sodium hydroxide. Barium hydroxide lime contains 4.6% potassium hydroxide, whereas soda lime contains only 2.5% potassium hydroxide and 1.5% sodium hydroxide, hence greater initial proton abstraction with barium hydroxide lime. Although barium hydroxide is contained in barium hydroxide lime but not soda lime, this appears to be less relevant because barium hydroxide does not catalyze difluoromethoxy proton abstraction.

Anesthetic development and the search for an “ideal” anesthetic agent continue. A previous investigation identified structural features that predispose halogenated compounds to metabolism and metabolism-based toxicity and proposed methods for designing safer chemicals by avoiding these elements. [42]Similarly, the current investigation identifies structural features that predispose halogenated anesthetics to CO²absorbent-catalyzed degradation to CO, and a possible mechanism of CO formation. Safer anesthetics can be designed by avoiding structural features that facilitate CO formation.

The authors thank Dr. Jeff Jones, University of Rochester, for the helpful discussions regarding difluorocarbene chemistry and Dr. William Trager, University of Washington, for the discussions regarding deuterium isotope effects and their mechanistic implications. We also thank Drs. Trager and Rob Palmer for critical review of the manuscript.

[section]Bedford RF: Note from the FDA. Anesthesiology 1995; 83(4):33A

[double vertical bar]Formation of a similar carbanion, difluoromethyleneisopropylether (isopropoxyfluoromethylene), has been shown. [21] 

#Defined as any water remaining after the absorbent is dried to constant weight. Desiccated absorbents may contain some residual water, as judged by their ability to scavenge CO². [10] 

**Moon RE, Ingram C, Brunner EA, Meyer AF: Spontaneous generation of carbon monoxide within anesthetic circuits. Anesthesiology 1991; 75:A873.

[dagger, dagger]Potassium superoxide did catayze formation of CO from desflurane, enflurane and isoflurane, although the relative rates for the three compounds were much closer than those observed with soda lime or barium hydroxide lime (results not shown).