Relative Amnesic Potency of Five Inhalational Anesthetics Follows the Meyer-Overton Rule

After obtaining institutional animal care and use committee (University of California, Irvine, California) approval, 330 male Sprague-Dawley rats (weight, 250–280 g on arrival) were obtained from Charles River Laboratories (Wilmington, MA). They were housed individually in a temperature-controlled (22°C) colony room, with food and water available ad libitum . Animals were maintained on a 12-h light–12-h dark cycle (7:00 am–7:00 pm, lights on). Rats were maintained in the animal colony for 1 week before IA training. Within each study, rats were randomly assigned to receive either no anesthesia (air control) or a specific subanesthetic target dose of a particular agent during IA training. Each animal was used in only one determination with the IA technique. Subsequently, some animals were further used for dose–response determinations of pain sensitivity or the loss of righting reflex endpoint.

Experiments were performed in the following sequential order, based on the anesthetics being tested: (1) halothane, target doses of 0, 0.1, 0.15, 0.2, and 0.3% inspired, n = 90; (2) isoflurane, target doses of 0, 0.05, 0.1, 0.2, and 0.3% inspired, n = 60; (3) sevoflurane, target doses of 0, 0.1, 0.2, 0.3, and 0.4% inspired, n = 90; (4) desflurane, target doses of 0, 0.5, 1.0, and 2.0% inspired, n = 50; and (5) nitrous oxide, target doses of 0, 10, 20, 40, and 60% inspired, n = 40. The large number of animals used reflects the inherent variability of the IA technique, which typically requires approximately 8–15 animals per data point to establish a dependable measurement. The highest dose studied with each agent was selected empirically after parameter experiments on a few animals, which established the dose for which the animals could no longer reliably perform the IA task. Animals retained at least some mobility at 0.4% halothane, 0.4% isoflurane, 0.6% sevoflurane, 3% desflurane, and 80% nitrous oxide but could not reliably learn the IA task at these doses.

On the training day, the rats were taken from their home cages, weighed, and then placed into small (i.e. , 3.2 l) anesthetizing chambers that were filled with the targeted dose of the selected agent in air. Anesthesia was delivered in air through standard vaporizers at 0.5 l/min. Nitrous oxide was delivered with oxygen. Chamber and apparatus agent concentrations were monitored continuously during the experiment using a Datex-Ohmeda Ultima Capnomac (Helsinki, Finland) and verified with gas chromatography (model 80123B; SRI Instruments, Redondo Beach, CA). The animals remained in the anesthetizing chamber for at least 45 min. They were then quickly (i.e. , < 4 s) removed from the chamber and placed into the “safe” compartment of the IA apparatus, which had also been filled with the targeted dose of agent in air. Table 1shows the comparison between the targeted dose selected and the actual dose delivered to the IA apparatus during training, as assessed with chromatography. The gas chromatograph was calibrated against known standard calibration gases and by measuring gas concentrations after injection of a known amount of drug into a known calibrated volume.

The IA apparatus was airtight, and experiments were conducted in a large fume hood. The apparatus consisted of a V-trough–shaped alley (91 cm long, 15 cm deep, 20 cm wide at the top, and 6.4 cm wide at the floor) that was divided into two compartments separated by a manually controlled sliding door that opened by retracting into the floor. The starting compartment (31 cm long) was colored white and illuminated, whereas the shock compartment (60 cm long) was dark colored and not illuminated. Animals sat for 3 min in the safe compartment of the apparatus before the beginning of training to allow for the small fluctuations in anesthetic concentrations associated with the transfer to stabilize. We separately determined that this rapid transfer process did not appreciably change the agent concentration in the IA apparatus. Control rats were treated identical except they were only exposed to air.

For each agent, the dose causing a loss of righting reflex was also determined using four animals on a separate day after completion of their memory retention testing. Animals were again placed into the anesthetizing chambers that were initially filled with the highest dose of agent used in the memory assessment. The dose was incrementally increased (0.05 MAC) every 30 min until the animals could no longer right themselves after being placed on their backs. This dose was used in determining the zero reference point for calculation of ED⁵⁰values for amnesia. The rationale for this endpoint is twofold. First, the IA paradigm assesses whether animals retain a contextual link between the dark part of the training apparatus and the delivered shock experience; if the animals do not have contextual information about their environment, because they are unconscious, then they cannot make the necessary learning association. Second, Dutton et al.  8used a fear-potentiated startle paradigm to investigate learning and memory during isoflurane anesthesia at doses that crossed the loss of righting reflex threshold. In their study, contextual learning approached zero at a dose approximating the loss of righting reflex dose.

A continuous multiple-trial IA training procedure was used for assessment of learning and memory. For training, each animal was placed into the light–safe compartment of the IA apparatus facing away from the door. After 3 min, the door was lowered out of the way to reveal the dark–shock compartment. The rat instinctively prefers a dark environment. As the rat stepped into the dark–shock compartment with all four paws, a foot shock (0.3 mA; Master Shocker-Model 82400, Lafayette Instrument Co., Lafayette, IN) was delivered until the animal escaped back into the starting light–safe compartment. The door to the dark compartment remained open, and the animals could choose to either stay in the light-safe compartment or again cross into the dark–shock compartment. Animals crossing back into the dark–shock compartment were again given another foot shock and allowed to escape back to the light–safe compartment. Learning was considered to have occurred when animals avoided the dark–shock compartment for greater than 100 consecutive seconds. After the animals attained the 100-s learning criterion, they were removed from the apparatus and returned to their home cages.

The number of trials required (i.e. , the number of shocks delivered) for each animal to learn the task was taken as an index of how difficult the task was for a particular group of animals to learn it. It is important to note, however, that even though the exposure to the anesthetic condition might be expected to make the task more difficult for some animals to learn, the use of the multitrial training technique assures that all animals eventually did acquire the task information (i.e. , they all learned to stay out of the dark chamber to avoid a shock).

Memory retention was tested 24 h after the training session. No shock or drug was delivered during the memory testing. Each rat was placed back into the starting light–safe side of the apparatus, and the time taken (600 s maximum) for each rat to again cross into the dark–shock side was recorded. Longer latencies to cross into the dark side were interpreted as indicating better retention of the training experience.

It is well established that retention time with the IA paradigm depends in part on the intensity of the shock delivered to an animal.22,23Higher foot shock intensity equates with longer retention times (i.e. , a greater memory of the shock) and a lower number of trials needed to learn the task (i.e. , they learn it the first time if the shock is sufficiently intense). We established with parameter experiments a control performance in the range of 300 s by using a foot shock of 0.3 mA. This midrange control retention performance allowed for both memory-impairing and memory-enhancing effects of these anesthetic agents to be demonstrated as either relative increases or decreases of retention performance versus  the controls. Furthermore, this level of foot shock provided for some range in task acquisition performance of the control animals and thus allowed for the quantification of anesthetic-induced changes in this correlate of learning behavior to be documented.

Given the influence subanesthetic doses of volatile agents have on pain processing, where doses around 0.1 MAC cause hyperalgesia24and greater doses cause analgesia, it is likely that anesthetic-induced changes of pain processing at the time of encoding ultimately have at least some influence on subsequent retention performance. We quantified the magnitude of this potential influence in two ways. First, in animals that had already completed the IA training and testing, we separately determined (1–7 days later) flinch threshold for pain at the various doses and with the different types of anesthetic agents studied. This allowed us to compare analgesic potency differences between agents and understand how such analgesic potency differences between agents might influence subsequent retention of the training experience. In essence, given equivalent amnesic doses of two agents, the agent with the more potent analgesic properties will likely show a greater reduction of retention performance because of reduced encoding of the training experience than would be expected solely on the basis of its true amnesic potential. Therefore, the clearest interpretation of an amnesic effect is seen with those agents for which a particular dose is found that causes a robust decrease in retention performance without also causing a noticeable analgesic effect.

Second, to rule out the possibility that the anesthetics only seemed to be amnesic simply because the animals were not experiencing the shock as painful during encoding, we directly examined with one presumed “amnesic” dose of sevoflurane (i.e. , 0.4%) the relation between retention latency and increasing levels of foot shock delivered at current intensities clearly above the pain threshold (such that all rats jumped when shocked and many vocalized). We looked at this issue with sevoflurane because it seemed to have the strongest analgesic effect of the volatiles examined. If the dose-related changes in retention performance caused by exposure to the anesthetic during training are only related to the anesthetic-induced decrements in encoding intensity caused by an analgesic effect, levels of foot shock clearly above the pain intensity threshold should likely eliminate any apparent amnesic effect.

For flinch sensitivity determination, rats were placed into the shock compartment of the IA apparatus, which had been filled with a targeted dose of the inhalational agent being studied. Rats were exposed to the target dose of anesthesia for 45 min before testing. Rats were then given a foot shock starting at 0.2 mA. The current was then slowly increased at the rate of 0.025 mA/min until the rats made a noticeable flinch response to the shock as defined by paw withdrawal, vocalization, or sudden movement. After a clear response, the current was stopped, and the animals were allowed to rest for 3 min before the process was repeated. At least three separate determinations were made for each animal. The flinch threshold was taken as the mean value of shock current intensities that just demonstrated a response. Each data point at a particular dose and type of anesthesia represents the mean of the values obtained from 9–17 animals per measurement.

In the case where the behavioral data were not normally distributed, such as during training or retention testing, we used a nonparametric analysis approach. The Kruskal-Wallis test was used to assess group effects, and post hoc  comparisons were made using the Mann–Whitney U test. For normally distributed variables, such as pain sensitivity, a parametric analysis approach was used, including analysis of variance for group effects and post hoc t  tests. A probability level of P < 0.05 was considered significant, after Bonferroni/Dunn correction for multiple comparisons, where appropriate. Control rats were run within each agent-specific experiment. However, because there was no significant difference in control rat learning or retention behavior across groups, the control data were pooled to help minimize the total number of rats used.

Nonlinear logistic regression was used to calculate ED⁵⁰values for the amnesic threshold. Dose–response data for inhibition of memory were fitted using GraphPad Prism software version 4 (GraphPad Software, Inc., San Diego, CA). The data were fit to a sigmoidal (four-parameter logistic) dose–response curve with variable slope bound by upper and lower constraints with the data normalized on a scale ranging from 0 to 100. The lower constraint, where retention was set to zero, was the dose of agent that caused a loss of righting reflex. The upper constraint was set at the value of mean retention performance for the control animals. Differences between the ED⁵⁰and Hill slope values for the various modeled curves were subsequently determined with P < 0.05 considered significant. Correlations were investigated with Pearson product moment correlation analysis and Fischer r  to z  conversion.

Eleven animals were not included in the final analyses secondary to procedural difficulties during training or testing. Such difficulties included rare occasions of an animal’s tail or foot getting caught in the door or lid of the apparatus, an animal jumping out of the IA box or anesthesia chamber during transfer, failure of the shock to be properly delivered, and others. Exclusions were random across the various conditions, and all exclusions were made blind to the retention test data.

Dose–response effects of halothane on task acquisition, memory retention latency, and flinch threshold are shown in figures 1A, B, and C, respectively. Halothane began to interfere significantly with task acquisition at the relatively low target dose of 0.15% inspired. Learning became progressively more difficult with increasing doses of anesthesia (fig. 1A). Psychomotor effects of stumbling, unsteady gait, and uncoordinated movements were clearly evident at the 0.2% level for many rats and for virtually all rats at 0.3%. In essence, doses of halothane above 0.15% caused animals to need approximately 1 trial more to acquire the task than controls. However, it is important to note that there was no significant difference in learning difficulty between the 0.2 and 0.3% inspired conditions (P < 0.46). Loss of righting reflex occurred at 0.68 ± 0.03% (mean ± SD) inspired halothane.

Unexpectedly, halothane at the 0.1% inspired target concentration significantly enhanced 24-h memory retention latency (fig. 1B). Retention performance returned to near control values at 0.15%, and although the median was slightly decreased from control values at 0.2%, the 0.2% group was not significantly different from control. Retention performance, however, was significantly lower than control values at 0.3%. The 0.3% performance was also significantly (P < 0.005) lower than that found at 0.2%, indicating that the amnesic effect caused by halothane becomes apparent in this model over a relatively narrow dosage window. This change in memory behavior over the 0.2–0.3% window sharply contrasts with the relatively equivalent effect of this dosage range on task acquisition (fig. 1A).

The effect of such doses of halothane on the flinch threshold or the sensitivity of the animals to the foot shock is shown in figure 1C. At the 0.1% target dose, there was a significant 15% decrease in the mean flinch threshold compared with control values. Thus, at approximately 0.1 MAC halothane, there was a hyperalgesic response to electrical foot shock. This 15% change in foot shock sensitivity occurred coincident with the 38% enhancement of median retention performance (figs. 1B and C). A significant analgesic effect of halothane relative to control did not occur until the 0.3% dose. However, the foot shock sensitivity at 0.3% was not significantly different from that found at 0.2%. This suggests that the analgesic effect of halothane only marginally, if at all, contributed to its apparent amnesic effect, because both the analgesic response and the effects on task acquisition did not change much over the same dosage window where memory performance changed the most.

Dose–response effects of isoflurane on task acquisition, memory retention latency, and flinch threshold are shown in figures 2A, B, and C, respectively. Isoflurane did not cause a significant increase in the trials to learning criterion at any dose (fig. 2A), although psychomotor effects were clearly evident at the 0.2% dose for most animals. Loss of righting reflex occurred at 0.82 ± 0.05% inspired isoflurane. There was no apparent memory-enhancing effect at any dose (fig. 2B). Memory retention latency was significantly decreased, consistent with amnesia, at 0.2% and even more markedly decreased at 0.3%. A significant increase in flinch threshold versus  control was found at 0.2%, consistent with an analgesic response, and the flinch threshold had a marked further increased at 0.3%.

Dose–response effects of sevoflurane on task acquisition, memory retention latency, and flinch threshold are shown in figures 3A, B, and C, respectively. There was a significant increase in the trials to criteria at the 0.3% dose. There was some qualitative difference in behavior during training between the 0.3% and 0.4% doses such that the 0.4% dose did not necessitate a significantly increased number of trials. Although both groups clearly showed signs of being affected by the anesthetic, the rats at 0.3% seemed more excited and impulsive, whereas rats at 0.4% seemed more sedate and showed less exploratory behavior. Loss of righting reflex occurred at 0.76 ± 0.01% inspired sevoflurane. Similar to halothane and differing from isoflurane, the lowest target dose of sevoflurane (0.1%) suggested a memory-enhancing effect, but this was not statistically significant (P < 0.16). Retention performance then sharply decreased over a narrow dosage range, and the 0.2% dose approached an amnesic effect, but this dose was not significantly different from control (P < 0.06). Retention performance was significantly reduced at both the 0.3% and the 0.4% doses. In contrast with both halothane and isoflurane, there was no apparent hyperalgesic response or tendency toward hyperalgesia with the lower dose of sevoflurane. An analgesic response was found relative to control at the lowest 0.1% dose, and the analgesic response continued to dose-dependently increase up to the 0.4% level.

Dose–response effects of desflurane on task acquisition, memory retention latency, and flinch threshold are shown in figures 4A, B, and C, respectively. Desflurane, like halothane, showed significant inhibition of task acquisition (fig. 4A). At 1%, animals required approximately one additional trial to acquire the task information, a relative dose-response–related effect that was similar to that seen with halothane. At 2%, animals required approximately two additional trials to acquire the task information. This was the most inhibition of task acquisition seen with all of the agents studied. Loss of righting reflex occurred at 3.51 ± 0.01% inspired desflurane. Desflurane significantly inhibited retention latency at the lowest targeted dose studied of 0.5% or 0.44% measured (i.e. , 0.06 MAC) and progressively further reduced retention at increasing doses of 1% and 2% (fig. 4B). Desflurane, like isoflurane and unlike halothane or sevoflurane, did not demonstrate any memory-enhancing effects at any of the doses studied. Desflurane was significantly analgesic versus  control only at the highest (i.e. , 2%) dose studied (fig. 4C). Desflurane was similar to sevoflurane and unlike halothane or isoflurane in that it did not show any evidence of a hyperalgesic response at any dose. The amnesic effect of desflurane seems to be one of its most potent properties, because the lowest target dose of 0.5% (i.e. , 0.44% measured) caused a significant reduction in retention latency without also significantly affecting the trials to learning criteria or the flinch threshold.

Dose–response effects of nitrous oxide on task acquisition, memory retention latency, and flinch threshold are shown in figures 5A, B, and C, respectively. Nitrous oxide did not demonstrate a significant effect on task acquisition at any dose studied (fig. 5A). This apparent lack of an effect on learning with nitrous oxide was unlike that seen with any other agent. Animals were clearly affected by the 60% dose of the drug because they showed noticeable psychomotor difficulty, but after the single initial shock, the animals ran back to the safe compartment and essentially froze until they were removed from the apparatus when the time criterion expired. Loss of righting reflex did not occur at the highest dose of nitrous oxide studied, which was 85% inspired. An estimated loss of righting reflex endpoint of 95% was used in the ED⁵⁰calculations based on the animal’s near loss of this reflex at the 85% dose that was studied. The 10% nitrous oxide exposure group seemed to have a lower retention latency than the controls, but the difference was not significant (P < 0.09). A robust amnesic response was found with the 20% nitrous oxide dose (fig. 5B), which was even more significantly different from controls at the 40% dose. Nitrous oxide was also the most analgesic of the agents studied (fig. 5C) with the 60% dose increasing the mean flinch threshold by 23% versus  control. Nitrous oxide at 10% seemed to be hyperalgesic, but this response was not significantly different from control (P < 0.07).

The relative amnesic effects of the different agents and doses are shown in figure 6. MAC values (table 1) of halothane (0.9% atm),24isoflurane (1.49% atm),24,25desflurane (7.8% atm),26–28sevoflurane (2.2% atm) and nitrous oxide (220% atm)27,29were taken from the literature for age-appropriate Sprague-Dawley rats. The ED⁵⁰values for amnesia with 95% confidence intervals are shown in figure 6and listed in table 2. As expected from visual inspection of the figure, agents with overlapping 95% confidence intervals were not significantly different from each other. Nitrous oxide was the most potent amnesic, and halothane was the least potent. Figure 6and table 2show that the relative potency for causing an amnesic response in this model of learning and memory follows the order nitrous oxide ≥ desflurane ≥ sevoflurane ≥ isoflurane >> halothane. Note that this order is opposite to relative MAC potency such that the weakest agent for producing the MAC response (i.e. , nitrous oxide) is the most potent at causing amnesia as a percentage of its MAC value. In addition, it is noteworthy that the dose–response curves for nitrous oxide and desflurane have a different shape with a significantly lower slope as compared with the other agents. This raises the possibility that differential mechanisms are involved with mediating anesthetic-induced amnesia.

The relative potency order between agents also highly correlates with the agent’s respective oil:gas partition coefficients (r =−0.956, P < 0.007).30,31This relation is shown graphically in figure 7. This correlation suggests that an agent’s hydrophobicity plays a role in determining its amnesic potency just as it does in determining anesthetic potency.32,33The well-established correlation between MAC and oil:gas partition coefficients is also shown graphically in figure 7for comparison using the same five agents that were studied (r =−0.991, P < 0.0002).34,35 

Sevoflurane at 0.4% caused an increase in the flinch threshold of 15% from baseline (fig. 3C). This suggests that one factor contributing to the apparent amnesic effect of sevoflurane (for example) could have been an analgesic-induced reduction in the aversive quality of the training shock. That is, it is possible that the animals did not remember the training experience simply because the anesthetic exposure itself reduced the perceived intensity of the shock to a level below that which the animals found aversive (i.e. , a reduction in encoding). If so, then increasing the shock during training to a value that the animals would clearly find aversive might serve to eliminate such an amnesic response. We tested this idea for an “amnesic” dose of sevoflurane by systematically increasing the delivered shock intensity during task acquisition for additional control and sevoflurane (0.4%)–exposed animals. The results are shown in figure 8. Note that increasing the shock intensity to a value twice that found for the baseline flinch threshold did not eliminate the amnesic effect found with sevoflurane at 0.4% (fig. 8). Both control and anesthetic memory retention latencies did proportionally increase with increasing shock intensities, as expected from much work demonstrating the role of the amygdala in memory modulation. However, if the 15% reduction in pain sensitivity caused by 0.4% sevoflurane were the only reason for the measured decrement in retention performance caused by this dose of anesthesia, the 100% increase in delivered shock intensity during training with 0.4% sevoflurane exposure should have restored retention to near baseline control levels, which it did not.

The effects of various subanesthetic doses of common volatile anesthetic agents and nitrous oxide on learning, memory, and pain sensitivity were assessed in rats using the well-established IA paradigm. A number of conclusions can be drawn from these data. (1) For all agents studied, which included halothane, isoflurane, desflurane, sevoflurane, and nitrous oxide, an amnesic effect was found. The amnesic doses were clearly subanesthetic, and they occurred in the 0.06–0.3 MAC range. These threshold doses seem to be in line with the literature or slightly lower than expected from the literature.36(2) Amnesic potency, based on relative MAC dosing and ED⁵⁰values, was significantly different among the agents and followed the relative order of nitrous oxide ≥ desflurane ≥ sevoflurane ≥ isoflurane >> halothane. This order correlates with the oil:gas partition coefficients for these agents and suggests that a hydrophobic site plays a role in mediating the amnesia of anesthesia. (3) Despite the correlation with hydrophobicity, the relation between amnesic potency and analgesic potency was not consistent between agents. The doses causing amnesia were somewhat higher (i.e. , sevoflurane), roughly equivalent to (i.e. , halothane, isoflurane, nitrous oxide), or somewhat lower (i.e. , desflurane) than those producing analgesia. (4) The agents also differed in their ability to inhibit initial acquisition of the IA task. Desflurane dose-dependently affected task acquisition the most, whereas nitrous oxide and isoflurane did not significantly increase the number of trials needed to learn the task at any dose studied. (5) A significant hyperalgesic response to electrical foot shock pain was only found for halothane at the approximate 0.1 MAC inspired dose. The other agents were not associated with a significant hyperalgesic response.

This study did not identify a common or prototypical response profile for how an “inhalational” agent would be expected to affect learning, memory, and pain sensitivity; rather, these data reveal that these various endpoints of anesthesia represent dissociable agent-specific properties such that a potent action on one neurobiologic system does not necessarily determine a potent action on another. Moreover, even within the same system, such as the “memory” system, the differential potencies found between these agents imply that agent specific mechanisms may be at work and that anesthesia studies of memory should be interpreted relative to the exact agent being investigated. A differential activity profile for these agents is likely mechanistically related to how they differentially interact with various specific receptors or target systems.30,37These different profiles of action also strongly argue against the idea that a unitary mechanism of anesthesia might exist. The unitary mechanism hypothesis posits that all agents should be roughly interchangeable and equivalent in their behavioral effects. Clearly, relative to the behavioral endpoints examined here, they are not.

One fundamental observation, however, stands in stark contrast to the above argument against a unitary theory. Amnesic potency highly correlated with oil:gas partition coefficients in a manner identical to that expected from the Meyer-Overton rule (fig. 7). This suggests that a common mechanism may be involved in mediating both amnesia and anesthesia, or at least the site mediating amnesia is likely to be a hydrophobic site. This correlation also suggests that the process that causes anesthesia may be the same process that causes anesthetic-induced amnesia, with the amnesia component simply being manifest at a proportionately lower dose than that required to cause anesthesia. In addition, this correlation suggests that a line of inquiry into the mechanisms of anesthetic-induced amnesia using substances that deviate from the Meyer-Overton rule might be useful, in a manner similar to the utility of this approach in mechanistic studies of the MAC response.38 

The agents studied did have two behavioral properties in common. For each agent studied, a dose was found that caused an amnesic response and a dose was found that caused an analgesic response. Given that the IA paradigm uses a painful stimulus as the source of memory, it seems logical that an analgesic effect of an anesthetic might itself diminish the encoding of the painful stimulus and lead to a reduction in retention performance. The magnitude of this potential confound is hard to establish for nitrous oxide, isoflurane, and sevoflurane because these agents caused relatively more dose-related analgesia in a pattern that inversely mirrored their apparent memory reducing effects. In contrast, this potential confound does not seem to apply to desflurane and halothane. Desflurane was significantly amnesic at a very low dose (i.e. , 0.44%) that was well below the dose needed to demonstrate any analgesic effect (i.e. , 2%). In the case of halothane, there was no significant difference in the pain threshold between 0.2 and 0.3% inspired concentrations, but a robust amnesia was apparent at 0.3% and not at 0.2%.

It could be further argued that the analgesic component of the drug exposure caused a failure of encoding because the animals did not even feel the shocks or find them aversive. A few points argue against a failure to encode as being the sole factor in determining retention performance. First, all animals reacted to the shock at each dose and each type of anesthesia studied, and they actively avoided it. Second, all animals learned the IA task to the selected criterion. Had they not found the shocks aversive, they would not have learned the task in the relatively low number of trials needed for each agent. Third, as a control experiment addressing the analgesia issue, we found that increasing shock intensity to a value well above flinch threshold somewhat enhanced retention performance but did not eliminate the amnesic effect of sevoflurane (fig. 8). Fourth, at least one agent, desflurane, demonstrated an amnesic response at a dose that was much lower than that which caused an analgesic effect or any noticeable psychomotor or behavioral effects. Taken together, these points suggest that the decrements in 24-h retention performance were primarily due to a strong amnesic effect of the anesthetics studied and not just due to a weaker encoding state caused by the anesthetic exposure.

In animal studies of learning and memory, it is essential to attempt to dissociate amnesic effects of a drug from other behavioral effects of the drug present during learning. Such behavioral effects include anesthetic-induced changes in pain thresholds, motor activity, sedation, or mood effects. In this study, we quantified the effects on pain processing by measuring flinch thresholds, and we characterized each agent’s effects on task acquisition using the continuous multitrial training approach as a correlate for the other behavioral effects. However, this correlate of learning should be interpreted with some caution because an increase in training trials can be due to any one of or any combination of the effects mentioned above. Furthermore, a lack of increase in training trials does not mean that a particular agent is without effect on the learning process at the dose studied. For example, one might misinterpret the lack of increased trials with nitrous oxide to imply that nitrous oxide at the doses studied does not affect learning, but there is good evidence in both animals and humans that doses of nitrous oxide around 30% do affect learning.9,39Therefore, the training trial data should most appropriately be viewed only as a rough indicator of an agent’s potential to interfere with task acquisition and as an indicator of which agents might deserve further detailed study as inhibitors of learning.

Are the amnesic doses found here reasonable and relevant to human studies of amnesia? The data from various reports using various memory paradigms and examining a number of the agents do seem consistent with the current results. For halothane, Cook et al.  3found amnesia for word pairs but not pictures on a short-term memory test at a 0.2% inspired dose. This compares with our demonstration that 0.3% halothane was needed to cause amnesia at 24 h for an aversive stimulus. For isoflurane, Dwyer et al.  4found the drug reduced retention for answers to multiple-choice questions that had been presented during anesthetic exposure with an ED⁵⁰of 0.26% and 95% confidence intervals of 0.19–0.32%. We found comparable results with our paradigm in that isoflurane had a somewhat lower ED⁵⁰value of 0.20%, but with 95% confidence intervals of 0.18–0.23%, which do overlap the bottom range of the findings of Dwyer et al.  Our animal work is also comparable with the animal work of Dutton et al. ,8who used a fear conditioning paradigm in rats. They found that isoflurane-induced amnesia to contextual information occurred with a slightly higher ED⁵⁰value of 0.375%, with 95% confidence intervals of 0.33–0.42%. For nitrous oxide, the dose causing amnesia for contextual information seems to be in the range of 30%,40which is in line with or a bit higher than our 20% dose for suppression of aversive memory. However, Dwyer et al.  4also studied nitrous oxide and found a much higher dose needed for amnesia with an ED⁵⁰of 52%. For desflurane, Gonsowski et al.  41found complete suppression for learning of multiple-choice questions during a 0.6-MAC dose. This complete suppression dose is consistent with our finding a loss of righting reflex occurring at 0.5 MAC desflurane in the rat, which we used as our zero memory reference point in modeling the dose–response profile of desflurane. Taken together with the current data, these reports suggest that the IA technique represents a good model system of human explicit memory.

These current findings also suggest that a distinct amnesic threshold dose may exist for some agents, as related to the shape of their dose–response profiles. For example, halothane did not cause a significant reduction in retention performance at 0.2%, but it did at 0.3%, despite similar effects at both doses on pain processing, qualitative motor performance, and task acquisition.

The lowest dose of halothane studied was associated with a significant increase in retention performance. The ability of halothane to enhance retention performance raises the possibility that patients recovering from halothane (and possibly sevoflurane) anesthesia likely pass through a dose stage where there is not only increased sensitivity to pain,24but also an increased propensity toward remembering such an experience. This low dose–related memory enhancing effect might at times contribute to cases of intraoperative awareness, because clinically, the dose of a volatile agent is often reduced when a lighter level of anesthesia is desired, such as during obstetric anesthesia or as might occur with trauma and cardiac anesthesia.12 

Based on the work of Zhang et al. ,24who tested hind paw withdrawal latency to a heat stimulus during exposure to various doses and types of anesthesia in the rat, it was expected that a hyperalgesic effect would also be found with all of the agents studied here at around the 0.1-MAC dose level. However, a hyperalgesic effect was found only for halothane. This failure to confirm a hyperalgesic effect for the other agents in our study is likely due to the different nature of the pain stimulus used (i.e. , electrical vs.  heat), because others have documented that behavioral results are influenced by the type of pain stimulus used when studying anesthetic related drugs.42,43 

Much of the work directed toward understanding how inhalational agents work and the receptor mechanisms mediating the MAC response is highly relevant to the effects of anesthetics on learning and memory.30,44Of those systems investigated as a basis for MAC, a number emerge as particularly relevant for amnesia, including but not limited to anesthetic effects on γ-aminobutyric acid, glycine, glutamate, acetylcholine, and serotonin systems.45–54Within the amygdala, γ-aminobutyric acid systems are known to play a role in mediating memory for IA learning, and the fact that the amnesic effect of both the benzodiazepines and propofol,19,55both γ-aminobutyric acid agonists, depends on the intact functioning of the basolateral amygdala strongly implicates this neurobiologic system as a key mediator for anesthetic amnesic effects.

Also of interest is the fact that the slope of the dose–response curve is lower for nitrous oxide and desflurane (fig. 6). This suggests that these two agents may invoke slightly different mechanisms for causing amnesia than the other agents. A low slope on a dose–response graph can sometimes implicate that a second messenger system is involved in the response. A likely possibility here would be a G-protein–coupled receptor system, such as would be found with a number of targets relevant to anesthesia mechanisms such as muscarinic and adrenergic receptors.56,57The fact that nitrous oxide seemed to have one of the most potent affects on memory retention in this model is intriguing when considered in light of its strong in vitro  effects on N -methyl-d-aspartate receptors.58The N -methyl-d-aspartate receptor has long been known to be involved in learning and memory.59Given that other volatile agents also are known to have effects on the N -methyl-d-aspartate receptor,60–62this receptor system also seems to offer promise for further study.

In summary, the relative amnesic potency of various inhalational agents was determined in the rat using the IA model of learning and memory. On a relative MAC basis, task acquisition (i.e. , a correlate of learning) was most inhibited by desflurane, whereas 24-h retention (i.e. , amnesic potency) followed the order nitrous oxide ≥ desflurane ≥ sevoflurane ≥ isoflurane >> halothane, a potency order that obeys the Meyer-Overton rule.

The authors thank James L. McGaugh, Ph.D. (Professor, Department of Neurobiology and Behavior, University of California-Irvine, Irvine, California), for his continued support and Larry Cahill, Ph.D. (Assistant Professor, Department of Neurobiology and Behavior, University of California-Irvine), for helpful discussions.