THE sedative, analgesic, and euphoric effects of opioids have been known since antiquity. First described by the Sumerians some 6,000 yr ago, the euphoric and analgesic properties of the seed-pod exudate of papaver somniferum are described in Homer's Iliad and were well known to physicians by the time of Hippocrates (460–377 B.C.E.). However, since the time of Pliny the Elder (23–79 C.E.), it has also been known that opioids may produce life-threatening respiratory depression, which limits both their utility and their safety. This issue of Anesthesiology includes two reports that enhance our understanding of these respiratory depressant effects. Bouillon et al. at Stanford provide new insights into the acute effect of rapidly acting opioids on ventilatory control through the use of a new modeling technique, which permits non–steady state estimates of the interaction between the respiratory depressant effect of remifentanil and the respiratory stimulating effect of retained carbon dioxide. 1In contrast, Romberg et al. at Leiden studied the pharmacodynamics of morphine-6-glucuronide (M6G) 2; this is of particular interest because of the possibility that M6G may have greater affinity for μ1opioid receptors (responsible for analgesia) than for μ2receptors (responsible for respiratory depression), offering the promise of systemic analgesia with an increased margin of safety. 3
There are several ways to study the effects of sedative, analgesic, and anesthetic drugs on the ventilatory response to carbon dioxide. Perhaps the simplest is to follow changes in respiration (rate, tidal volume, minute ventilation) and carbon dioxide tension (arterial, end-tidal) during room air breathing. This type of closed-loop design (so called because carbon dioxide tensions are regulated by the body's intrinsic negative feedback system) most closely mimics the clinical situation where patients breathe spontaneously after receiving sedative medications. However, measurements based on resting ventilation have significant disadvantages that make it difficult to draw meaningful pharmacodynamic conclusions. First, resting ventilation and carbon dioxide tension are determined by the intersection of the carbon dioxide ventilatory response curve (fig. 1, curve A ) and the carbon dioxide excretion (metabolic) hyperbola (fig. 1, curve C ). A drug that causes a 50% decrease in the slope of the carbon dioxide response (fig. 1, curve B , would cause Paco2to increase by less than 10%, with a similarly modest decrease in resting ventilation. 4Furthermore, when ventilation is unstimulated by exogenous carbon dioxide, conscious influences such as anxiety or merely thinking about one's breathing tend to perturb the measurements more than when ventilation is stimulated by an imposed increase in Paco2. Finally, especially with rapidly acting drugs, it is not uncommon for apnea to occur; under these circumstances it becomes impossible to quantitate the ventilatory response until the carbon dioxide tension rises sufficiently to restimulate ventilation (fig. 1, line B ′).
This Editorial View accompanies the following articles: Bouillon T, Bruhn J, Radu-Radulescu L, Andresen C, Cohane C, Shafer SL: A model of the ventilatory depressant potency of remifentanil in the non–steady state. Anesthesiology 2003; 99:779–87; Romberg R, Olofsen E, Sarton E, Teppema L, Dahan A: The pharmacodynamic effect of morphine-6-glucuronide versus morphine on hypoxic and hypercapnic breathing in healthy volunteers. Anesthesiology 2003; 99:788–98.
To overcome these obstacles, a variety of study designs have been used. These involve measuring the effects of medications on VEmeasured at two or more imposed levels of hypercarbia. Such methods are termed open loop because the values of Paco2are predefined and thus are unaffected by drug-induced changes in VE. For long-acting drugs (or continuous infusions of shorter-acting drugs), both steady-state and rebreathing techniques can be used to determine the ventilatory response to artificially imposed hypercarbia. Rebreathing techniques involve simultaneous measurement of VEand Paco2while the subject's Pco2increases as a result of rebreathing from a closed system without carbon dioxide absorption. For steady-state measurements, VEis measured after the subject equilibrates for 6–8 min at each of two or more elevated levels of Paco2(typically between 46 and 58 mmHg). Romberg et al. used just such a steady-state technique in their comparison of the long-acting drugs morphine sulfate (MS) and M6G. 2For single injections of short-acting drugs such as propofol or remifentanil, isohypercapnic techniques have been described; ventilation is continuously measured during and after drug administration, whereas end-tidal pressure of carbon dioxide is held constant. 5This method allows minute-to-minute determination of the carbon dioxide ventilatory response. If the pharmacokinetics of the drug are known, the effect compartment rate constant (keo, which indicates how quickly the drug gets from the bloodstream to the site in the central nervous system where it causes respiratory depression) and potency (EC50, which indicates the plasma concentration of the drug that causes a 50% decrease in ventilatory drive) of the drug can also be estimated. However, none of these techniques directly predicts what will happen to an actual patient, whose ventilation is not stimulated by exogenous carbon dioxide, when a rapidly acting ventilatory depressant is administered.
Bouillon et al. have overcome some of the shortcomings of previous closed-loop determinations of ventilatory drive by developing a mathematical model incorporating the pharmacodynamics of both carbon dioxide and remifentanil. 1Their model is designed to predict both the magnitude and time course of changes in VEand Paco2that would be expected to follow a remifentanil-induced perturbation of the ventilatory response to carbon dioxide. To determine the pharmacodynamics of the carbon dioxide response, the investigators used a standard rebreathing technique. As indicated above, this method only provides information about that part of the carbon dioxide response curve which lies above the carbon dioxide excretion hyperbola (because ventilation is being stimulated by exogenous carbon dioxide). However, administration of a sedative drug in the absence of exogenous carbon dioxide will cause ventilation to drop below the carbon dioxide excretion hyperbola, as the carbon dioxide response curve acutely shifts down and to the right. To estimate the relationship between VEand Paco2below the hyperbola, it was necessary to extrapolate. The authors demonstrated that their model fit the data best when the relationship between ventilation (VE) and Pco2was expressed as a power function:
where V0and P0are the baseline values for Pco2and VE, respectively. The exponent F is an indicator of the “strength” of the carbon dioxide response before remifentanil administration (analogous to the slope of curve A in the current fig. 1). As shown in Bouillon's figure 2, 1the carbon dioxide response lines predicted by this exponential model are curvilinear, resembling the “hockey stick” appearance of the awake carbon dioxide response.
To model the effect of remifentanil on the carbon dioxide response, the authors assumed a sigmoidal relationship:
which implies that at any given Paco2, alveolar ventilation (Valv) decreases from V0(its baseline value in the absence of remifentanil) to asymptotically approach 0 at high plasma concentrations (Cp) of remifentanil; C50is the concentration at which ventilation is depressed to half its baseline value at any given Paco2, whereas γ determines the “steepness” of the decline in ventilation with increasing Cp.
The second novel part of the model involves incorporating the pharmacodynamics and kinetics of carbon dioxide itself. Under the reasonable assumption that carbon dioxide production remains constant, the rate at which Paco2increases as a function of VEcan be easily calculated; this is a generalization of the observation that during apnea Paco2increases by 3–6 mmHg · min−1to situations in which ventilation is depressed but not zero. After incorporating an equilibration delay constant for carbon dioxide (to account for the ventilatory response to changes in Paco2being not instantaneous), the authors closed the loop by substituting the sigmoidal prediction of VE(their equation 11 into the equation predicting the rate of increase of Paco2(their equation 9, to get their final model (their equation 13. With model in hand, the authors administered target-controlled infusions of remifentanil, designed to achieve stepwise increases in remifentanil plasma concentration. Based on frequent measurements of plasma remifentanil concentration and Paco2, they determined values for the parameters of their pharmacodynamic model that best fit the observed data.
The conclusions are fascinating and consistent with our observations in clinical practice. The plasma concentration of remifentanil predicted to cause a 50% decrease in alveolar ventilation at any given Paco2(EC50) was 0.92 ng · ml−1; this corresponds closely with previously published values of the concentration required to decrease ventilation by 50% during imposed hypercapnia. 5,6More important, however, the authors demonstrate that the observed ventilatory effect depends on how quickly a given blood concentration is achieved. As shown in their figure 4, following a rapid 0.5 μg/kg dose of remifentanil (peak plasma concentration ≈ 5 ng · ml−1) the model predicts that ventilation will rapidly decrease to about 10% of its baseline value *; within 10 min ventilation will recover and actually exceed its baseline value as the remifentanil concentration drops in the presence of an elevated Paco2. In contrast, if a similar blood level is gradually achieved by intravenous infusion (≈ 0.2 μg · kg−1· min−1for 10 min), ventilation gradually decreases to a nadir of about 35% of its baseline value, before stabilizing at 60% of baseline. The reason for the discrepancy is that slow administration of the remifentanil allows Paco2to increase (to a predicted value in excess of 60 mmHg), partially offsetting the ventilatory depressant effect of the remifentanil itself.
What are the clinical implications? For a given level of sedation, “bolus” administration of a sedative or opioid drug is more likely to cause more severe respiratory depression than gradual administration. During deep sedation or general anesthesia, administration of 50 μg of fentanyl to a typical adult almost always causes apnea, whereas administration of an equipotent dose (5 mg) of morphine typically causes modest respiratory slowing. The discrepancy can be explained by the slower onset of morphine, which allows Paco2to gradually increase and stimulate ventilation. The distinction is important because gradual respiratory slowing minimally affects arterial oxygenation, whereas opioid-induced apnea is likely to produce hypoxemia unless the patient has been breathing an oxygen-enriched mixture.
Romberg et al. used more routine methodology to compare the ventilatory effects of MS and M6G; their thesis, based on ligand binding studies, was that M6G would produce less respiratory depression than an equi-analgesic dose of MS. To prove their thesis, of course, it is first necessary to determine the relative analgesic potencies of MS and M6G. Previously published studies provide a wide range of potency ratios, depending on the species studied and route of administration. For example, whereas Person et al. found that 0.05 mg/kg of intravenous M6G was almost as effective as 0.15 mg of intravenous MS in relieving experimental ischemic pain, 7Lötsch et al. found that 0.045 mg/kg of intravenous M6G (followed by an infusion to maintain steady-state plasma concentrations) was ineffective in blunting the pain associated with nasal insufflation of 60% CO2. 8In the present study, Romberg chose a M6G dose of 0.2 mg/kg, based on unpublished data which suggested that it “caused potent and long-lasting analgesia,” although no data are provided to establish equal analgesic efficacy with their 0.13 mg/kg dose of MS. Interestingly, owing to its higher molecular weight, the molar dose of M6G (0.43 μm/kg) was essentially the same as that of MS (0.46 μm/kg).
As shown in their figure 2, both M6G and MS significantly depressed the carbon dioxide ventilatory response; although there was no statistical difference between the time courses of the two drugs, it seems that the effect of M6G dissipated more rapidly than that of MS. Similarly, the peak effect on acute hypoxic response was similar between the two drugs, although the effect of M6G diminished more quickly than that of MS. Based on previously published data for the pharmacokinetics of MS and M6G, the authors then used pharmacokinetic/pharmacodynamic modeling techniques to estimate the EC25(effect-site concentration causing 25% depression) and T1/2keo(the time required for the drug concentration within the central nervous system to reach half of its level in the plasma).
The authors’ observation that the EC25of M6G is 20–50 times greater than that of MS does not, by itself, imply that respiratory depression is less likely to occur with M6G than with MS. Rather, it is a consequence of differences in the pharmacokinetics of the two drugs. The volumes of distribution of M6G are appreciably smaller than those of MS. Thus, for any given dose, the plasma and effect site concentrations of M6G are necessarily higher than those of MS. However, these higher concentrations are also required to produce the desired, analgesic effect. Thus, Lötsch et al. found that plasma M6G concentrations of 400 nm/l, similar to those estimated by the present study to cause a 25% decrease in the ventilatory response to carbon dioxide, were completely ineffective in blunting experimental pain. 8
Does the mechanism of ventilatory depression by M6G differ from that associated with MS? Romberg et al. found that for MS, the EC25for the carbon dioxide response was greater than that for the hypoxic response (28.0 vs . 16.5 nm), whereas for M6G, the reverse was true (528 vs . 873 nm). However, neither confidence limits nor the results of statistical tests are provided, so it is impossible to determine whether this discrepancy implies that the two drugs act at different sites in the ventilatory control mechanism, or whether it is merely an example of type I error.
The other interesting finding of the Romberg study is that the effect-site equilibration of M6G (T1/2keo= 2.1 h) seemed to be more rapid than that of MS (T1/2keo= 3.8 h). This seems to be inconsistent with M6G being a more polar molecule than MS and, hence, crossing the blood-brain barrier more slowly. 9The authors suggest that this inconsistency may be related to differences in the transport of the two substances within the central nervous system. However, when pupil size rather than ventilation was used as a pharmacodynamic indicator, Lötsch et al. found that the T1/2keoM6G was more than double that of MS. 10A possible explanation for the discrepancy is that the volunteers developed tolerance to the respiratory depressant effects of M6G, artificially lowering the calculated T1/2keofor M6G.
The key role of the blood-brain barrier in modulating the effects of M6G as compared with MS has been established by studies in which the substances have been injected directly into the subarachnoid space. In rodents, the analgesic potency of subarachnoid M6G is about 100 times greater than that of subarachnoid MS. 3Grace and Fee found that the analgesic potency of subarachnoid M6G was about five times that of subarachnoid MS; however, respiratory depression was more likely in patients receiving M6G, again suggesting that use of M6G does not reduce the risk of respiratory depression. 11
Thus, although Romberg et al.’ s findings suggest that the respiratory depressant effects of M6G are similar to those of an equi-analgesic dose of MS, the data are not conclusive. Ideally, analgesic and respiratory depressant potency would be measured in the same subjects at a series of times after drug injection; pharmacodynamic modeling could then be used to determine if the respiratory margin of safety of M6G exceeds that of MS. Until such a comparison is undertaken, the utility of M6G as a respiration-sparing opioid analgesic remains a matter of speculation rather than reality.
Development of an opioid agonist devoid of ventilatory depressant effects remains the holy grail of analgesic pharmacology. Unfortunately, at least with regard to M6G, the cup remains empty, because Romberg et al. were unable to demonstrate an increased margin of safety for this drug compared to nonselective μ-opioid agonists such as morphine. However, Bouillon et al.’ s study suggests that the risks of currently available nonselective opioids can be reduced through a more thorough understanding of the pharmacodynamics of ventilatory control. Their data demonstrate that by avoiding rapid changes in the concentration of depressant drugs in the respiratory control centers of the central nervous system, we can keep our patients inspiring , so they do not expire.