Speed of Onset and Offset and Mechanisms of Ventilatory Depression from Sevoflurane : An Experimental Study in the Cat

Experiments were performed in 12 adult cats of either sex (body weight, 2.5-3.6 kg). The Ethical Committee for Animal Experiments of the Leiden University Medical Center approved the use of the animals. Experiments were performed with the animals during light anesthesia. Anesthesia was induced with 10 mg/kg ketamine given intramuscularly, followed by sevoflurane inhalation. Subsequently, the right femoral vein was fitted with a cannula and 20 mg/kg [Greek small letter alpha]-chloralose and 100 mg/kg urethane were slowly administered intravenously, and the volatile anesthetic was withdrawn. Thereafter, a continuous infusion of 1 or 2 mg [middle dot] kg^-1[middle dot] h^-1[Greek small letter alpha]-chloralose 5-10 mg [middle dot] kg^-1[middle dot] h^-1urethane was started. With this regimen, the level of anesthesia is sufficient to suppress the pain withdrawal reflex but light enough to preserve the corneal reflex. The anesthetic regimen has little effect on the ventilatory response to hypercapnia compared with the awake state and does not yield systemic changes in respiratory variables over time (>6 h). [16-19]The right femoral artery was fitted with a cannula to allow continuous measurement of blood pressure.

The trachea was fitted with a cannula at the mid cervical level and connected to a Fleisch number 0 flow transducer (Lausanne, Switzerland) to measure inspiratory and expiratory flow. The flow transducer was connected to a T piece of which one arm received a continuous fresh gas flow of 5 l/min. Three computer-controlled mass flow controllers (High-Tec, Veenendaal, The Netherlands) composed desired inspiratory gas mixtures of oxygen, carbon dioxide, and nitrogen. A sevoflurane vaporizer (Vapor 19.3; Drager, Lubeck, Germany) was used to administer the volatile anesthetic. Sevoflurane, oxygen, and carbon dioxide concentrations were measured using a Capnomac Ultima monitor (Datex, Helsinki, Finland). Temperature was controlled within 1 [degree sign]C in each cat and ranged among the cats between 38 and 39 [degree sign]C.

All signals were recorded (sample frequency, 100 Hz) and processed using a PDP minicomputer (Digital Equipment Co., Maynard, MA). Ventilation, end-tidal carbon dioxide pressure (PET^CO(2)), end-tidal oxygen pressure (PET^O(2)), and PET^SEVOwere determined and collected on a breath-to-breath basis.

PET^O(2) and PET^CO(2) were controlled using the “dynamic end-tidal forcing” technique. [14,15]Studies started when the cat showed a regular breathing pattern. This was generally reached 1.5 h after the start of the continuous [Greek small letter alpha]-chloralose infusion.

Protocol 1. Stepwise changes in PET^SEVOwere performed with constant levels of PET^CO(2)(clamped at 5 mmHg above resting pressure) and PET^O(2)(clamped at 110 mmHg) in seven cats. The PET^SEVOpattern applied consisted of (1) 5-10 min at 0 mmHg, (2) a stepwise increase to a specified target level, (3) 10-20 min at the target level, and (4) 20-30 min at 0 mmHg. In four cats, a single study was performed. To obtain information on intrasubject variability, three to seven studies were performed in the three remaining cats. Target PET^SEVOranged among the cats from 3.8 mmHg ([approximate] 0.5%) to 7.5 ([approximate] 1%), 11.3 ([approximate] 1.5%), and 15 mmHg ([approximate] 2%).

The V^i-PETSEVOdata were analyzed using an empiric model consisting of two parts. The first part (part 1) postulates a hypothetical effect compartment and describes the time lag between end-tidal and effect-site partial pressures. The second part (part 2) consists of an inhibitory sigmoid Emax model that translates effect site partial pressure into ventilation (Figure 1). [9-13] 

The effect-site concentration is described by part 1 of the model:Equation 1where k is a first-order rate constant, PET^SEVO(t) is the imposed end-tidal sevoflurane tension at time t (i.e., the input function), and Pec(t) is the effect-site partial pressure at time t. We report the effect-site equilibration half-life (t^1/2), which equals ln2/k. Note that parameter k describes the hysteries, which may result from more than pharmacokinetics (see Discussion). Part 2 of the model is described by Equation 2where E(t)= Vⁱ(inspired ventilation) at time t, minimal effect E^min= estimated baseline Vⁱor V^BL, and E^max= V (ⁱ) at maximal respiratory depression from sevoflurane or V^DEP. PEC (⁵⁰) is the partial pressure that results in 50% of maximal effect and gamma is a shape parameter. Substitution results in Equation 5Because inhalational anesthetics can cause apnea, V^DEPis set at zero and thus Equation 3Parameters of the E^maxmodel were estimated using non-linear regression based on the actual PET^SEVOwaveform. On and off responses were analyzed simultaneously. We did not observe systematic differences in parameter estimates in the same cat among the different sevoflurane target groups. Therefore, we pooled that data is cats 005, 006, and 007 by averaging the mean values of the data obtained at identical PET (^SEVO) targets.

Protocol 2. In eight cats, the dynamic ventilatory responses to stepwise changes in PET^CO(2) were determined before and during two constant end-tidal sevoflurane concentrations (0.5 and 1%). The PET^O(2) was maintained at 110 mmHg during the studies. Three of these cats participated also in protocol 1. After a period of steady state Vⁱwhen PET^CO(2) was increased slightly above resting values, PET^CO(2) was increased by 7-10 mmHg in a stepwise manner and held constant for 7 or 8 min. Thereafter, the PET^CO(2) was decreased to its original level and kept constant for another 7 or 8 min. Four to six control studies were performed in each cat. Subsequently, three to four steps at 0.5% end-tidal sevoflurane and three to four steps at 1% end-tidal sevoflurane were performed. The order of sevoflurane studies was random. They started after approximately 30 min of steady state end-tidal sevoflurane concentration.

The steady state relation between Vⁱand PET^CO(2) at constant PET^O(2) in the cat is linear down to apnea and described by Equation 4where G^cand G^pare the central and peripheral ventilatory carbon dioxide sensitivities and B represents the apneic threshold or extrapolated PET^CO(2) at zero Vⁱ. To estimate G^p, G^c, and B, we fitted the ventilatory responses to a two-compartment model, as described previously. [14,15]We report all parameters of the model: time constant of the peripheral chemoreflex loop, [Greek small letter tau]^p, time constants of the on and off responses of the central chemoreflex loop ([Greek small letter tau]^onand [Greek small letter tau]^off), time delays of the peripheral and central chemoreflex loops (T (^p) and T^c), G^c, G^p, and B. Parameter estimates of the two-compartment model were obtained using a least-squares method.

To compare the control and sevoflurane data, a two-way analysis of variance was performed on the estimated parameters using a fixed model. Post hoc comparisons were made using the Newman-Keuls test. The level of significance was set at 0.02. All values reported are the mean +/− SD, unless otherwise stated.

A total of 20 studies was obtained. In all studies we observed rapid changes in Vⁱon the introduction and withdrawal of sevoflurane, with an evident time lag between Vⁱand PET^SEVO(see Figure 1, bottom left where the arrows represent the time course). Inspection of the individual data fits showed that the model adequately describes the ventilatory dynamics of sevoflurane at constant PET^CO(2) and PET^O(2). An example of data fitted to the model is given in Figure 1(bottom right). Table 1shows parameter mean values of individual cat means. Overall, parameter t^1/2averaged 3.6 to 1.0 min. A sensitivity analysis of the empiric model is presented in Appendix 1.

Ninety-eight studies were performed: 44 control, 27 at 0.5% end-tidal sevoflurane, and 27 at 1% end-tidal sevoflurane. To get an appreciation of the quality of the data, we plotted typical runs obtained at 0, 0.5, and 1% end-tidal sevoflurane in one cat in Figure 2. Table 2shows the mean values of all estimated parameters. Figure 3plots the individual parameter values. The relations of PET^SEVOversus G^pand G^cwere not linear (Figure 3and Figure 4). At 0.5%, sevoflurane reduced G^pand G^cto the same extent, causing no change in the ratio of G^pto G^c. At 1%, in contrast to G^c, G^pdecreased further, resulting in a decrease in the G^p:G^cratio (Figure 4).

The time constants of the peripheral and central chemoreflex pathways decreased significantly at 1% end-tidal sevoflurane (Table 2).

We studied the influence of anesthetic concentrations of sevoflurane on breathing in an experimental cat model. We had two end points. First, we intended to learn about the dynamics of onset and offset of respiratory depression; and second, we wanted to identify the mechanisms or sites of action responsible for the depressant effects. The findings of the first study indicate that a steady state in Vⁱis reached after approximately 18 min (five times the value of the effect-site equilibration half-life) on a change in PET^SEVO. The second study shows the steeper end-tidal sevoflurane-peripheral chemoreflex sensitivity (G^p) relation compared with the end-tidal sevoflurane-central chemoreflex sensitivity (G (^c)) relation.

To study the temporal profile of respiratory changes caused by sevoflurane, we used an empiric model. This model has been applied previously in studies of the influence of morphine, fentanyl, and remifentanil on ventilation in animals and humans [10,11]and many other effects of pharmacologic agents (e.g., the effect of opioids on electroencephalographic-derived parameters). [9,12,13] 

The speed of onset and offset of the respiratory effects of sevoflurane, the t^1/2, was 3.6 min. In humans, we recently measured the dynamics of the 95% spectral edge frequency of the electroencephalogram resulting from changes in the end-tidal sevoflurane concentration. [20]The mean value of t^1/2obtained in 12 patients averaged 3 min. When we extrapolate the results of this study to patients after sevoflurane anesthesia, the time course of sevoflurane-induced ventilatory dynamics would be of the same order as the time course of electroencephalograph dynamics. However, inhalational anesthetics significantly affect ventilatory control at values less than the concentration at which patients “wake up”(the so-called minimum alveolar concentration [MAC]-awake). For example, in humans, isoflurane, 0.13%, reduces the hypoxic ventilatory response by approximately 50%. [21]This is well below the MAC-awake of 0.36%. [22]The latter indicates that respiratory depression from inhalational anesthetics persists for some time after patients awaken from general anesthesia.

The time lag or hysteresis between the PET^SEVOand ventilation depends on several factors. These include (1) the end-tidal to arterial sevoflurane gradient;(2) cardiac output-dependent delivery of sevoflurane to sites relevant to ventilatory control, such as carotid bodies, brain, medulla, respiratory muscles, and lung;(3) changes in blood pressure at the carotid sinus baroreceptors causing (fast) changes in ventilation (independent of arterial P^O(2) and P^CO(2))[23,24];(4) the wash-in and wash-out of sevoflurane into and out of the brain compartment (this factor depends on cerebral blood flow, the blood-brain partition coefficient, and the partial pressure of sevoflurane in arterial blood);(5) changes in central respiratory drive caused by alterations in brain blood flow (an increase in brain blood flow causes the washout of carbon dioxide and acid metabolites, and, consequently, central respiratory drive will diminish)[25]; and (6) central neuronal dynamics. The rate constant k combines the dynamics of all of these factors. We simulated the washout of sevoflurane from the brain (factor 4) using elimination kinetic data obtained from rats. [26]The estimated brain elimination half-life (t^1/2elim) was a factor of two faster than the t^1/2observed in protocol 1. This implies that in addition to the brain or effect-site partial pressures, factors 2, 3, 5, and 6 play an appreciable role in the respiratory pharmacokinetics and pharmacodynamics of sevoflurane.

Despite the presence of background anesthesia, the cats did not become apneic during the administration of sevoflurane, even at the highest PET^SEVO(15 mmHg or approximately 2%). We cannot draw conclusions about the effects of sevoflurane at partial pressures more than 15 mmHg. At high sevoflurane concentrations, nonlinear threshold values may cause the respiratory oscillator to stop abruptly and cause irregular or cyclic breathing and apnea (see references 1 and 8).

We used a mathematic two-compartment model of the respiratory controller incorporating gains (ventilatory carbon dioxide sensitivities), time constants, and time delays to estimate the relative contributions and dynamics of the fast peripheral and slow central chemoreflex loops. [14,15]This model has been validated extensively in the anesthetized cat. For example, DeGoede et al. [18]showed that the steady state contributions of the peripheral and central chemoreflex loops obtained with the technique of artificial brain stem perfusion closely reflect values derived from the two-compartment model in intact cats. Furthermore, studies using artificial brain stem perfusion to measure the dynamic ventilatory responses to changes in PET^CO(2) at constant central (i.e., brain stem) P^CO(2) and to changes in PET^CO(2) obtained values for time constants and delays that corresponded closely with the control values observed in this study. [27,28] 

The decrease in G^cand G^pat anesthetic levels of sevoflurane indicates an appreciable depressant effect on the ventilatory control system. However, even at the highest sevoflurane concentration studies, a peripheral response to carbon dioxide was observed. Although small, G^pcould be identified in seven of eight cats. This observation corresponds with recent findings of Stuth et al. [29]They studied the influence of the anesthetic concentrations of halothane (0.45-1.8% end-tidal)[parallel] on the canine peripheral chemoreceptor-mediated phrenic nerve response to acute hypoxia. They observed a dose-dependent depression, but not abolishment of the phrenic nerve response to hypoxia for the range studied. The authors calculated that at halothane, 1.8%, the phrenic nerve response to hypoxia was approximately 20% of the awake value. Stuth et al. [30]obtained similar results using carotid body carbon dioxide stimulation. Sevoflurane (in cats anesthetized with [Greek small letter alpha]-chloralose-urethane) is a more potent respiratory depressant than halothane (in vagotomized dogs) in terms of the end-tidal concentration. How this translates into equal MAC conditions remains unknown, because the MAC value of sevoflurane in our feline preparation (with background anesthesia) was not determined.

At 0.5% sevoflurane, the values of G^cand G^pwere 43 +/− 20% and 37 +/− 18% of control, respectively. At sevoflurane, 1%, the respective values of G^cand G^pwere 37 +/− 225 and 12 +/− 13% of control (Table 2, Figure 4), indicating that only G^pdecreased further. We interpret these findings as a steeper dose-response relation for G^pversus PET^SEVOthan for G^cversus PET^SEVO. # This suggests that in addition to an effect of the anesthetic concentrations of sevoflurane at pathways common to both chemoreflexes (causing relatively equal reductions in G^pand G^c), there is an additional effect on the peripheral chemoreflex loop (which causes a further reduction in G^p). The pathways common to both chemoreflexes are the respiratory integrating centers in the brain stem and the neuromechanical link between the brain stem and ventilation. The latter consists of efferent nerves, spinal respiratory motoneurons, neuromuscular junctions, the diaphragm, intercostal muscles, and the lungs and airways.

Animal studies of the influence of anesthetics on respiratory control are equivocal with respect to sites of action. Although all investigators agree that the ventilatory control system is depressed at pathways common to the peripheral and central chemoreflex loops, some have found an additional selective effect within the peripheral chemoreflex loop, [4,30-32]some have not, [19,29,33,34]and others have shown a dependency of a selective effect on the (hypoxic) stimulus intensity. [35]Comparisons among these studies are difficult to make because of (1) evident species differences, (2) methodologic differences with respect to anesthesia (for example, the absence [4,29,30,32,34]or presence [19,33,35]** of background anesthesia or decerebration [31]), (3) differences in preparation (for example, artificial brain stem perfusion in tracheostomized animals [19,33]versus phrenic nerve reading in vagotomized dogs [29,30]), (4) absence of dose-response curves in some studies, and (5) possible masking of depression of peripheral responses by systemic hypotension. Furthermore, similar to other investigators, [29,30,35]we suggest that differences in results are caused in part by variations in stimulus intensity. For example, at deep hypoxic stimulation, the carotid bodies remain undepressed by anesthetics, whereas, at mild hypoxic stimulation, depression is present (for example, see the data from Ponte and Sadler [35]).

The reduction in time constants at sevoflurane, 1%, may be related to several mechanisms: an increase in baseline brain blood flow, an increase in the reactivity of brain blood flow to carbon dioxide, and inactivation or suppression of central neuronal ventilatory dynamics, or short-term potentiation. [5]The latter refers to respiratory activity at a higher level than expected just from the dynamics of the peripheral and central chemoreceptors. It is thought to originate at the pontomedullary region of the brain stem and exhibits slow dynamics. [5]Human studies indicate that inhalational anesthetics cause the reduction of time constants as a result of the inactivation of short-term potentiation. [5]More studies are needed to clarify the underlying mechanisms and the relation between the absence of short-term potentiation and respiratory instability during inhalation anesthesia.

When we compare our current results with those from human studies of the influence of sevoflurane and halothane on ventilatory control, [5,36-38]we observe quantitative rather than qualitative differences. In humans already at 0.1 MAC sevoflurane and halothane, investigators have observed a selective peripheral effect (30-50% depression of the peripheral chemoreflex loop)[5,36-38]and, at 0.1 MAC halothane, a reduction of the central time constants. [5]The MAC fraction of sevoflurane in our cats anesthetized with [Greek small letter alpha]-chloralose-urethane is not known. Because of the interaction that exists between anesthetics in reducing the MAC value, we argue that a MAC fraction much larger than 0.1 produced a degree of depression of G^pand a reduction of central time constants in our cats of similar magnitude as in humans at 0.1 MAC. The cause of these quantitative differences may simply be the species difference in anesthetic potency. Comparison of the MAC for skin incision in humans, cats, and dogs indicates that sevoflurane, halothane, isoflurane, and enflurane are more potent in humans. [39]There is no reason to doubt that this would not also apply to the effects of these anesthetics on the peripheral chemoreflex loop (G^p).

We determined the temporal profile of sevoflurane-induced respiratory depression in a feline model. More studies are required to compare our data with those from other anesthetics and from humans. Our results indicate that depression of minute ventilation is appreciable, which may persist postoperatively beyond the time that wakefulness has been regained. We further determined the “respiratory” sites of action of sevoflurane and observed an effect at the pathways common to both chemoreflex loops and a selective effect within the peripheral chemoreflex loop. These latter findings indicate that in addition to the species difference in anesthetic potency, similar modes of action of sevoflurane exist in humans and cats.

A sensitivity analysis of a proposed model (i.e., the empiric model of protocol 1) is important to determine whether, in practice, the parameter values of the model can be estimated with finite precision from the measured data. Parameters may not be identifiable for various reasons: because of the model structure, dependence on other parameters, or the specific input function chosen. The model is fitted to the data by choosing those parameter values that minimize the “cost” function, which is the sum of the squares of the difference between the measured ventilation data and the model prediction, on a breath-to-breath basis. We performed an a posteriori sensitivity analysis of the data shown in Figure 1. The sensitivity analysis was performed by fixing one parameter (i.e., by not allowing it to be estimated) at a time to a series of values (+/− 20%) around the “best” value (in the sense of minimal residual sum of the squares). The other parameters were estimated by minimizing the residual sum of the squares. [40]This method will show whether any of the parameters are or are not estimable using the specific PET^SEVOinput function. If not, the curve of the residual sum of the squares versus the fixed parameter (i.e., the cost surface) will be flat. Because we performed an analysis of actual data, which evidently incorporates some breath-to-breath noise, we will also identify the existence of so-called local minima.

(Figure 5) shows the results of the sensitivity analysis for each of the parameters of the empiric model: t^1/2, V^BL, PEC⁵⁰, and [Greek small letter gamma]. The cost function shows that all four parameters could be identified and that the estimation accuracy was acceptable for all parameters. The most accurately estimated parameter was V (^BL), and the least sensitive parameter was [Greek small letter gamma]. The latter is not surprising considering the single-step input function that we used (Figure 1). Using a different input function shape may increase the estimation accuracy of [Greek small letter gamma]. [40]Correlations between the estimated values of PEC⁵⁰and [Greek small letter gamma](correlation coefficient =-0.45) and [Greek small letter gamma] and t^1/2(correlation coefficient = 0.36) were low. This indicates that we were able to estimate these parameters independently.

[parallel] Canine and feline MAC values are 2.4% for sevoflurane and 0.9% for halothane.

# The slope of the linear relation of logG^pversus PET^SEVO=-0.91 against -0.44 for logG^c.

** This study.