Respiratory Sites of Action of Propofol: Absence of Depression of Peripheral Chemoreflex Loop by Low-dose Propofol

Ten healthy volunteers aged 18–25 yr (7 men and 3 women) participated in the protocols after approval was obtained from the local Human Ethics Committee (Leiden, The Netherlands). The subjects were healthy and did not have a history of tobacco or illicit drug abuse.

After arrival at the laboratory, two intravenous catheters were inserted in the left and right cubital veins (one for propofol administration and one for blood sampling). Subsequently, electrodes for electroencephalographic measurement (BisSensor; Aspect Medical Systems, Newton, MA) were placed on the head as specified by the manufacturer, and the subjects rested for 20–30 min. The subjects breathed through a face mask (Vital Signs, Totowa, NJ). Gas flows were measured with a pneumotachograph connected to a pressure transducer and were electronically integrated to yield a volume signal. Corrections were made for the changes in gas viscosity caused by changes in oxygen concentration of the inhaled gas mixtures. The inspired gas mixture was set via  a system of mass-flow controllers (Bronkhorst High-Tec, Veenendaal, The Netherlands), which received control signals from a personal computer. This allows the forcing of end-tidal partial pressure of oxygen (PETo²) and PETco²according to a specified pattern in time, independent of the ventilatory response. Inspired and expired O²and CO²concentrations and arterial hemoglobin–O²saturation (Spo²) were measured with a Datex Multicap gas monitor (near the mouth) and Datex Satellite Plus pulse oximeter, respectively (Datex-Engstrom, Helsinki, Finland). PETo², PETco², tidal volume, respiratory frequency, V̇ⁱ, and Spo²were collected and stored on disc for further analysis.

The electroencephalogram was recorded using an Aspect A-2000 EEG monitor (Aspect Medical Systems; software version 3.3). The BIS values were averaged over 1-min intervals.

End-tidal partial pressure of carbon dioxide was varied according to an MFBS that involved 13 steps into and 13 steps out of fixed PETco²levels (low and high CO²: 2 mmHg and 12 mmHg above the subjects’ normal air breathing values for PETco²), altogether lasting 1,408 s (23 min 28 s). Figure 1shows a schematic diagram of the PETco²input function. The MFBS experiments were performed at a background of moderate hypoxia (PETo²= 70 mmHg) and started 20 min after the initiation of hypoxia. This was done to allow time for hypoxic ventilatory decline to develop before investigating the response to CO².

A Psion palm-top computer (London, England) programmed with a three-compartment propofol pharmacokinetic data set was used to control a Becton Dickinson infusion pump (St. Etienne, France) for intravenous administration of propofol. 13Each subject performed two control MFBS experiments, two during low-dose propofol infusion (P^low; target plasma concentration, 0.75 μg/ml) and two during high-dose propofol infusion (P^high; target plasma concentration, 1.5 μg/ml). Control studies preceded propofol studies, and low-dose propofol experiments preceded high-dose propofol experiments. MFBS studies started 15 min after plasma target concentrations had been reached. The duration of propofol infusion was 150 min (75 min for P^lowand 75 min for P^high).

Six venous propofol samples were obtained before and after each of the MFBS experiments. The samples were collected in syringes containing potassium oxalate, and propofol concentrations were determined by reverse-phase high-performance liquid chromatography.

The data were analyzed by fitting the breath-to-breath ventilatory responses to a two-compartment model, as described previously. 9–11,14The steady-state relation of V̇ⁱto PETco²at constant PETo²in humans is described by:

where G^P= the carbon dioxide sensitivity of the peripheral chemoreflex loop, G^C= the carbon dioxide sensitivity of the central chemoreflex loop, and B = the apneic threshold or extrapolated PETco²of the steady-state ventilatory response to carbon dioxide at 0 V̇ⁱ. The sum of G^Pand G^Cis the total carbon dioxide sensitivity (G^TOT). To describe the delay in effect and dynamics of the peripheral and central ventilatory responses to CO², time delays (T) and time constants (τ) are incorporated in the model. The deterministic model parameters are as follows: B; G^C; G^P; time constant of the peripheral chemoreflex loop (τ^P); time constant of on-response of the central chemoreflex loop, i.e. , at high PETco²(τ^ON); time constant of the off-response of the central chemoreflex loop, i.e. , at low PETco²(τ^OFF); time delays of the central and peripheral chemoreflex loops (T^Cand T^P); and a linear trend term. 10The noise corrupting the data was modeled through an external pathway with first-order dynamics. 10Estimation of the parameters was performed with a one-step prediction error method. 15 

We performed an a posteriori  sensitivity analysis. Sensitivity analysis enabled us to determine whether the parameter values could be estimated with finite precision from the actual data. 16–18The analysis was performed by fixing one parameter (i.e. , by not allowing it to be estimated) at a time to a series of values (−100% to +100%) around the “optimum” value (in terms of the “cost” function or residual sum of squares of the difference between measured and estimated ventilation). The other parameters were estimated by minimizing the residual sum of squares. The shape of the relation between parameter and residual sum of squares informed us of whether parameters were estimable using the specific PETco²and PETo²inputs. 16–18Furthermore, because we performed the sensitivity analysis on actual data (and not on simulated data), we were informed of whether local minima existed.

The estimated parameters of control and propofol experiments were subjected to a one-way analysis of variance and post hoc  least significant differences tests. P  values less than 0.05 were considered to be significant. All values reported are mean ± SD.

All subjects terminated the protocol without side effects. Because of propofol, the arousal state of the subjects decreased with Bispectral Index values of 84 ± 8 at low-dose propofol infusion and 67 ± 14 at high-dose propofol infusion. The concentration of propofol remained constant over time during the two infusion schemes (table 1). Examples of a control and a propofol MFBS experiment (propofol target = 1.5 μg/ml) and model fits of one subject are given in figure 2. Only the deterministic part of the model is shown. It shows a large effect of propofol on the output of the central chemoreflex loop (a 55% reduction of G^C) with only a minor effect on the output of the peripheral chemoreflex loop (a 12% reduction of G^P).

The averaged model parameters are collected in table 1. At all three treatment levels, the estimated model parameters did not differ between the first and second CO²response. Propofol reduced the total CO²sensitivity (G^TOT) by approximately 20% at a propofol target of 0.75 μg/ml and by approximately 34% at a target of 1.5 μg/ml. At both propofol concentrations, the reduction of G^TOTwas caused by a reduction of the output of the central chemoreflex loop by 20% and 40% at low- and high-dose propofol, respectively, without affecting the output of the peripheral chemoreflex loop. As a consequence, the ratio G^Pto G^Cis increased relative to the control state (fig. 3). The apneic threshold (B) showed a small but significant reduction during propofol infusion (table 1). The time constants and time delays of both chemoreflex loops remained unaffected by propofol (data not shown).

Figure 4shows the results of the sensitivity analysis of the model parameters in one subject. A well-defined minimum of the residual sum of squares was observed for all parameters, indicating that they could be identified with acceptable accuracy (including parameters T^Pand T^P, not shown). The most accurately estimated parameters were B and G^P, as shown by the steepness of the increase in residual sum of squares at parameter values above and below the optimum. The shape of the curves for G^C, τ^P, τ^ON, and τ^OFFare markedly asymmetric, indicating that the estimation may be less accurate at values higher than the optimum. As expected, the steepness of the increase in the residual sum of squares of G^Pis less when this parameter is estimated from the single-step CO²input function (broken line in fig. 4). This indicates that G^Pis estimated with greater accuracy from an MFBS input function relative to a single-step input. G^Cis well-estimated from an MFBS and step input. However, the analysis indicates somewhat greater accuracy using a single CO²step for values above its optimum and an MFBS input for values below its optimum.

We used a multifrequency binary sequence in PETco²to quantify the effect of propofol on ventilatory control. The MFBS was designed by Pedersen et al.  12to spread its power over the frequency range of interest for identification of both peripheral and central chemoreflex responses and to optimize identification of the peripheral chemoreflex response. Using a single step, the peripheral response is determined from only a limited portion of the data (2 min of the 15–20 min of a CO²study). Using an MFBS, this increases significantly (19.5 min of a 24-min experiment). Consequently, the precision of estimation parameters related to the peripheral chemoreflex loop is greater when derived from MFBS compared with single steps. Indeed, our sensitivity analysis indicates the improvement of the estimation of the peripheral CO²sensitivity compared with a step PETco²function, without compromising the accuracy of estimation of central CO²sensitivity (fig. 4).

We used a target-controlled infusion system to administer propofol. Fifteen minutes after target plasma concentrations of propofol were attained, the respiratory studies started. Estimation of the effect-site propofol concentration indicated that this time was ample for equilibrium between blood and effect site. We measured venous propofol concentrations, which may not reflect arterial or effect-site concentrations. However, we observed no time effect on venous propofol concentrations or on parameter estimates at P^lowor P^high. This indicates stable arterial and effect-site propofol concentrations and suggests a small gradient between venous and arterial propofol concentrations.

With respect to the control of breathing, propofol may have an effect at the central or peripheral chemoreceptors, at the respiratory centers in the brainstem, at the neuromechanical link between brainstem and ventilation (V̇ⁱ), or at sites in the central nervous system involved in behavioral state control. The exact location of the central chemoreceptors is unknown, but they are probably located in the dorsomedial medulla, the rostroventrolateral medulla, or both. 19The peripheral chemoreceptors are located in the carotid bodies, which are strategically situated at the bifurcation of the common carotid arteries and have an important role in oxygen delivery to the brain. The peripheral chemoreceptors respond to changes in arterial oxygen tension (Pao²) and arterial carbon dioxide tension (Paco²). 20,21 

We observed that propofol, at doses causing a decrease in Bispectral Index to approximately 70, has an important effect on the control of breathing. Specifically, propofol reduced G^Cbut had little influence on G^P. This indicates the absence of a (selective) effect of low-dose propofol on the peripheral chemoreflex loop. In this respect, propofol stands in sharp contrast to the modern volatile halogenated anesthetics. 5–8Our findings are in agreement with studies from the literature. Dow and Goodman 22showed in humans that during propofol anesthesia, the carotid bodies retain their ability to respond to hyperoxia. In anesthetized cats, propofol displayed an inhibitory effect on areas of the dorsomedial and ventrolateral medulla, which possibly contain the central chemoreceptors. 19,23Evidently, our data do not preclude some depressant effect of higher doses of propofol than used by us on the carotid bodies or its afferent pathways. For example, animal data show that high-dose propofol infusion, 18–35 mg · kg⁻¹· h⁻¹, causes the cessation of carotid body chemoreceptor activity. 24 

Our finding of the absence of an effect of propofol up to plasma concentrations of 1.25 μg/ml on the peripheral CO²response seems in disagreement with our previous observation of a 50% depression of the acute hypoxic V̇ⁱresponse by 0.6 μg/ml propofol. 2Because the acute hypoxic ventilatory response originates at the peripheral chemoreceptors of the carotid bodies, 25some depression of the peripheral CO²response was anticipated from our earlier results. Apart from the possibility that O²and CO²sensing at the carotid bodies are differentially affected by propofol, there are three conceivable explanations for this discrepancy.

First, carotid body depression by anesthetics is dependent on stimulus intensity. For example, Ponte and Sadler 26showed that relative to a mild hypoxic stimulus, a more intense stimulus (partial pressure of oxygen [Po²] < 40 mmHg) is able to overcome volatile anesthetic-induced depression of the carotid bodies. In analogy, the stimulus in this study (a hypercapnic–hypoxic stimulus of a PETco²of 13 mmHg above resting and an Spo²of 88–90%) may offset depression of the carotid bodies by propofol as observed previously using a less-intense hypoxic stimulus (a hypercapnic–hypoxic stimulus of a PETco²of 5 mmHg above resting and an Spo²of ∼87%). Interestingly, when assessing the effect of low-dose volatile anesthetics on ventilatory control, we observed depression of carotid body–mediated responses even when intense stimuli, such as used in this study, were applied. 6,8This suggests a difference in stimulus intensity needed to overcome carotid body depression caused by propofol and volatile anesthetics.

Second, in cats, Berkenbosch et al.  21studied the peripheral V̇ⁱresponse dynamics to hypoxic stimulation while the Pao²of the medulla oblongata was kept constant using the technique of artificial brainstem perfusion. Mathematically, the responses were best described by two components: a fast component with a time constant of approximately 2 s and a slow component with a time constant of approximately 73 s. The fast component was considered to originate at the carotid bodies, whereas it was argued that the slow component was due to central modulation of the carotid body response (i.e. , neuronal dynamics). 21Interestingly, in the same animal preparation, the response of the peripheral chemoreflex pathway to changes in PETco²does not show a slow component. 21This indicates that although peripheral hypoxic stimulation activates central neuronal dynamics, peripheral hypercapnic stimulation does not. Also, in humans, the hyperventilatory response to hypoxia is well-described by a fast and a slow component. 27We previously studied the effect of 0.6 μg/ml propofol on the ventilatory responses to 3 min hypoxic pulses 2and reanalyzed the data using the two-component model as described by Berkenbosch et al.  21All control curves were best described by two components as judged by the Akaike criterion, 28with time constants of 3 and 100 s for the fast and slow components, respectively. Propofol did not affect the gain (i.e. , hypoxic sensitivity) of the fast component (G^propofol/G^control= 0.95), but caused a significant reduction of the gain of the slow component (G^propofol/G^control= 0.45;P < 0.05). If we assume that the fast response reflects the carotid body response to hypoxia and the slow component central neuronal dynamics, 21,29these results suggest that propofol affects central neuronal dynamics but has little effect on the carotid bodies or their output, and thus does not reduce G^p. This is in contrast to the effect of inhalational anesthetics. We previously studied the effect of sevoflurane, 0.25% end-tidal (∼0.15 MAC), on the ventilatory responses to 3-min hypoxic pulses 8and reanalyzed the data using the two-component model as described above. Sevoflurane reduced the fast and slow component by 25 and 60%, respectively (P < 0.05), an indication for an effect of sevoflurane on the carotid bodies and on central neuronal dynamics.

Third, apart from a stimulatory effect at the carotid bodies, hypoxia causes depression of ventilation via  central mechanisms, i.e. , within the central nervous system. 30The central effect of hypoxia on V̇ⁱis already apparent after 1 min of hypoxic exposure 29; therefore, any measured hypoxic V̇ⁱresponse is the mixture of carotid body and central effects on V̇ⁱ. Because propofol enhances the magnitude of the central depressant effects of hypoxia in humans, 2greater depression by propofol of the measured ventilatory response to hypoxia relative to the measured peripheral CO²response is expected.

These three mechanisms should be taken into account when comparing our current results on the effect of propofol on the peripheral and central chemoreflex loops with studies from the literature on the effect of propofol on the ventilatory response to acute 2,31and subacute hypoxia. 32 

The discrepant effects of propofol and sevoflurane on the carotid body response to CO²is striking and may be explained by differences in molecular sites of action of intravenous and halogenated inhalational anesthetics. We believe that propofol, like inhalational anesthetics, affects breathing through enhancement of γ-aminobutyric acid–mediated transmission and reduction of glutamatergic activity in the brainstem. 2,33This may have induced the depression of G^Cin our study. Furthermore, inhalational anesthetics activate background K⁺channels in the peripheral and central nervous system. 34–36These channels are involved in tonic inhibition of cellular excitability, and activation by volatile anesthetics may be the cause of some major side effects, such as depression of cardiac function and respiratory depression. Buckler et al.  36recently showed the existence of an oxygen-, acid-, and inhalational anesthetic (halothane)–sensitive background K⁺channel in the carotid body chemoreceptor cells, which possibly is an important link in the cascade leading to CO²and O²sensing in the carotid bodies and the selective site of inhalational anesthetic depression of carotid body function. Further studies are needed to show how anesthetics (including propofol) modulate the pH–Po²sensitivity of these background channels.