The respiratory depressant action of μ-agonistic opioids limits their therapeutic usefulness. So far, opioid-induced respiratory depression has been quantified by comparing the slope and intercept ('apnoeic threshold') of carbon dioxide (CO2)-response curves [1-6], measurement of minute ventilation (V) during room air breathing, challenges such as the addition of CO2 to inspired air [7,8], or the administration of hypoxic mixtures [9-11] and measurement of PaCO2 concentrations before, during and after opioid administration [1,12,13]. It is now well documented that μ-agonistic opioids shift the apnoeic threshold to the right, flatten the slope of the CO2-response curve, blunt the increase of minute ventilation during hypoxia, decrease minute ventilation and increase PaCO2 (PETCO2). Irregular breathing patterns after opioid administration has been described [14-16], but no attempt at quantification and/or exploration as a tool to measure the severity of opioid-induced respiratory depression has been made, presumably because it is more difficult to quantify than measurement of the minute ventilation or changes in PaCO2. Recently, we developed a pharmacokinetic and pharmacodynamic model able to describe the time-course of opioid-induced hypercarbia and to identify the respiratory depressant potency of opioids with vastly different pharmacokinetics and effect-compartment kinetics from non steady-state data [13,17]. During our experiments leading to these models, we observed that irregular breathing invariably accompanied opioid-induced respiratory depression, appeared to increase monotonously with decreasing minute ventilation and disappeared after cessation of the opioid infusion. This observation prompted us to quantify the opioid-induced increase of irregularity in the breathing pattern, observe whether it shows a comparable time-course to hypoventilation and test whether it can be used to identify patients at risk of immediate respiratory arrest.
Approval was obtained from our Hospital Ethics Committee and written informed consent was obtained from patients. We studied 23 ASA I-II patients undergoing major urological surgery. Fourteen patients received alfentanil (the data were previously reported ) and nine patients received pirinitramide under identical experimental conditions.
The unpremedicated patients were studied before induction of anaesthesia. When the patients arrived in the induction room, standard monitoring (noninvasive arterial pressure, electrocardiograph and pulse oximetry) was established; an arterial cannula was inserted into the radial artery of the nondominant hand and two intravenous (i.v.) cannulae were inserted into veins in both forearms. An i.v. infusion of alfentanil (2.3 μg kg−1 min−1) or pirinitramide (17.9 μg kg−1 min−1) was started while the patients breathed oxygen-enriched air (FiO2 = 0.5) via a tightly fitting continuous positive-airway pressure (CPAP) mask. This enabled us to obtain tidal volume, respiratory rate and, therefore, minute ventilation as well as the end-expiratory partial pressure of CO2 (PETCO2) with the standard monitors provided with the anaesthesia workstation with flow/volume autocalibration (Cicero, Dräger, Lübeck, Germany). These parameters were recorded online on a personal computer with a sampling rate of 1 Hz using Dräger proprietary software. The infusion was terminated when either a cumulative dose of 70 μg kg−1 alfentanil or 500 μg kg−1 pirinitramide had been given, PETCO2 > 8.6 kPa, or apnoeic periods lasting >60 s occurred. The upper dose of alfentanil was selected based upon pharmacokinetic modelling of the dose that would yield peak concentrations between 200 and 250 μg L−1, which were identified as allowing adequate spontaneous ventilation on recovery from anaesthesia . Lacking information about the concentration-respiratory depressant effect relationship of pirinitramide, we arbitrarily scaled up the cumulative alfentanil dose by a factor of seven to obtain the cumulative pirinitramide dose, according to our clinical experience with pirinitramide. Further, such scaling was attempted in one patient but it caused prolonged postoperative nausea and vomiting. The study was terminated if verbal stimulation was required to maintain spontaneous ventilation or if the patients felt uncomfortable while breathing through the mask. Otherwise, data were collected for 60 min after the beginning of the infusion. During the study, an observational scale (1: awake, restless; 2: awake, calm; 3: lightly sedated; 4: asleep) and continuous electroencephalographical monitoring were used to assess the state of arousal. Anaesthesia was induced after the study was terminated.
Tidal volume, respiratory rate - and therefore minute ventilation - as well as PETCO2 were measured with an anaesthesia workstation with flow/volume autocalibration (Cicero) and recorded online on a computer with a sampling rate of 1 Hz using Dräger proprietary software.
Data management and statistical analysis
From the available 23 patients (14 alfentanil, 9 pirinitramide), 3 patients who had received alfentanil were excluded from analysis. In our first patient, the sampling rate was erroneously set to 4 min−1, a resolution insufficient to calculate tidal volume variability from breath-to-breath values. In the second patient, we experienced a temporary failure of the flow sensor during the experiment. The third patient was very restless and talkative, leading to major swings in tidal volumes and minute ventilation.
The tidal volume data from the other patients were treated as follows: since the respiratory rates did not exceed 15 breaths min−1 and were decreased during opioid infusion, we first had to avoid multiple recordings of single tidal volumes. Deleting successive identical values, under the assumption that tidal volume differences between two successive breaths >10 mL, the volume resolution of our recordings, achieved this objective. Minute ventilation was calculated as the sum of the tidal volumes for every 1 min.
Tidal volume variability was calculated as the quartile coefficient of 20 successive tidal volumes. The time corresponding to the 10th breath/tidal volume in this sequence was taken as the 'time of measurement' for this respective value of the quartile coefficient: EQUATION
where Qeff20VT is the quartile coefficient of 20 successive tidal volumes and Q25,75 is the 25%, 75% quartile of 20 successive tidal volumes.
The equation was advanced one tidal volume at a time, comparable with the calculation of a sliding average (e.g. after calculation of Qeff20 for tidal volumes 1-20, the next Qeff20 would be calculated from tidal volumes 2-21). For every 1 min, the median of these sliding quartile coefficients was calculated and subsequently used as the quartile coefficient during this 1 min for further analysis.
Minute ventilation and tidal volume variability at the following times were tested for significant differences with a one-way ANOVA for repeated measurements: before, 5 and 10 min after start of infusion, at the cessation of the infusion, and at 5, 10, 20 and 30 min thereafter. Values from patients who had received pirinitramide and alfentanil were tested separately. If the result of the ANOVA yielded significant differences between groups, Dunnett's method was used to test for significant deviation from the baseline. Patients who had received alfentanil were then divided into those who completed the infusion regimen (n = 7) and those in whom alfentanil administration was discontinued for safety reasons (n = 4). Minute ventilation and tidal volume variability between these groups were compared using a t-test. For all statistical procedures, P < 0.05 was considered as significant.
The patient characteristics variables of the population studied are summarized in Table 1. The side-effects experienced and percentage of patients affected are summarized in Table 2.
The infusion had to be terminated prematurely in 6 of 14 patients who received alfentanil because of pronounced respiratory depression (PETCO2 > 8.6 kPa) or apnoeic periods >1 min). All patients who received pirinitramide tolerated the maximal dose of 500 μg kg−1 (700 μg kg−1 in one subject). In six patients, the data collection had to be terminated prematurely. One patient receiving alfentanil needed verbal stimulation to continue breathing after 20 min, four patients (alfentanil, two patients; pirinitramide, two patients) felt uncomfortable or nauseated, or both, while breathing through the mask after 45-50 min, and one patient vomited 50 min after pirinitramide. No patient was more than lightly sedated, as judged from both clinical observation and encephalographic data.
Typical time-courses for minute ventilation and tidal volume variability during and after an alfentanil infusion are shown in Figures 1 and 2. Figure 1 summarizes the data from a patient with pronounced respiratory depression necessitating discontinuation of the alfentanil infusion. Figure 2 shows data from a patient tolerating the scheduled dose of 70 μg kg−1 body weight. Raw data for patients who received pirinitramide resemble those shown in Figure 2 and is therefore not shown. Figure 3 shows the minute ventilation during and after the infusion of alfentanil and pirinitramide at the time points used for statistical analysis. Both alfentanil and pirinitramide caused highly significant changes in minute ventilation (both P < 0.001, ANOVA). Post-testing (multiple comparisons against baseline) identified several groups significantly different from baseline for both drugs (Dunnett's test, P < 0.05). It also revealed a slightly different pattern for alfentanil (earlier onset, shorter duration)- than for pirinitramide (later onset, longer duration)-induced hypoventilation. These findings are replicated when identical procedures are applied to tidal volume variability (Fig. 4). Both drugs lead to a highly significant change of tidal volume variability (both P < 0.001, ANOVA). Again, post-testing reveals several groups differing significantly from baseline (P < 0.05, Dunnett's test) and an earlier onset and shorter duration for alfentanil compared with pirinitramide. The most convincing argument for the validity of tidal volume variability as a measure of respiratory depression is shown in Figure 5. Unlike pirinitramide, alfentanil caused severe respiratory depression necessitating cessation of the infusion in some volunteers. Comparison of both minute ventilation and tidal volume variability between patients who could complete the infusion with those who received only a fraction of the planned dose reveals significant differences in minute ventilation and tidal volume variability (minute ventilation, P = 0.002: tidal volume variability, P = 0.034; t-test).
Opioid-induced respiratory depression has been systematically investigated for approximately 40 yr. During this time, several end-points have been proposed for its quantification (for a review, see [12,19]). Although it has been stated that 'irregular breathing may occur' during the course of opioid-induced respiratory depression, no formal investigation has been undertaken so far. During this study, we invariably observed irregular breathing accompanying a decrease in minute ventilation that subsided after termination of the opioid infusion. We also had the impression that more pronounced respiratory depression was accompanied by more pronounced irregular respiration, and that clinically obvious irregular respiration was a very ominous sign announcing imminent respiratory arrest. This prompted us to investigate (a) whether irregular breathing can be easily quantified, (b) whether it shows a time-course corresponding with that of hypoventilation and (c) whether it differed significantly between patients undergoing the planned infusion of opioid and those in which the infusion had to be terminated prematurely for safety reasons.
A common method for quantification of variability is fast Fourier transformation of the data followed by investigation of their amplitude and frequency patterns for each sampling interval, which has been applied to the analysis of heart rate variability [20,21]. This technique requires a substantial number of measurements per interval. Therefore, the resolution in the time domain of the method has to be limited for relatively infrequent signals. In human beings, the respiratory rate is almost one order of magnitude lower than the heart rate. Obtaining a sample of tidal volumes large enough to undergo analysis, like heart rate variability, would take at least several minutes, rendering the technique unsuitable for the description of non-steady-state respiratory data during and after infusion of a drug (rapidly changing drug concentrations). Furthermore, those methods are rather complex. Therefore, we decided to take 20 successive tidal volumes, corresponding to a period between 1.3 and 4 min for respiratory rates between 15 and 5 min−1, and calculate the so-called quartile coefficient of the tidal volumes, a simple and robust method for the description of variability. The quartile coefficient is calculated by determining the 25% (Q25), 50% (median) and 75% (Q75) quartile of a sample and dividing the difference between Q75 and Q25 by the twice the median. For a symmetrical distribution, the quartile coefficient cannot become <0 or >1. A value of 1 implies that the range between Q25 and Q75 equals the median. For an asymmetrical distribution, it may become >1, which implies that the range between Q25 and Q75 exceeds the median. From our experience, values approaching 1 for the quartile coefficient of 20 successive tidal volumes very likely represent artefacts caused by patients talking. One of these values can be seen in Figure 2. Since quantification of the duration of a respiratory cycle with our sampling rate (1 s−1) would have lead to an error of up to 25% for a respiratory rate of 15 breaths min−1 (duration of a respiratory cycle = 4 s), we decided a priori to analyse the variability of tidal volume size instead (equipment-induced measurement error <8% of true value).
The time-course of respiratory variability follows closely that of minute ventilation. Increased respiratory depression, as judged by minute ventilation, is invariably accompanied by an increase in tidal volume variability (Figs 3 and 4).
Very early in our experiments we felt that irregular breathing heralded respiratory arrest. Figure 5 entirely vindicates this impression. There is a clear distinction between patients able to tolerate the scheduled infusion of alfentanil and those who required its premature cessation lest long-lasting apnoea developed - both for minute ventilation and tidal volume variability. The fact that minute ventilation distinguished better than tidal volume variability - between those patients not able to undergo the prescheduled alfentanil infusion and those who required its premature termination (P = 0.002 versus 0.034) - is at least partly due to the minute ventilation/apnoea information being available online and actually being used as a criterion to stop the infusion. Clinically apparent irregular breathing appears to be a very ominous sign signalling severe respiratory depression. This facet of opioid-induced respiratory depression requires further analysis, especially from physiologists who have access to isolated preparations of the respiratory centres.
In summary, we demonstrated a method to quantify the increase in respiratory variability caused by opioids, and that the time-course of respiratory variability measured as Qeff20 of tidal volume parallels that of minute ventilation and that it correlates with the severity of respiratory depression. Clinically obvious irregular breathing after the administration of opioids is a sign of severe respiratory depression. Opioids apparently not only change the set point for PaCO2, but also impair the function of respiratory centres involved in rhythm generation. Further investigations to elucidate the exact mechanism of opioid-induced respiratory depression are needed.
The authors thank Thomas Falk, Dräger, Lübeck, Germany, for generous support of our research.
1. Bellville W, Seed JC. The effect of drugs on the respiratory response to carbon dioxide. Anesthesiology
2. Heitmann HB, Drechsel U, Herpfer G, Zindler M. Die Wirkung von Piritramid (Dipidolor) auf die Regulation der Atmung und die orthostatische Stabilitat des Kreislaufs. Anaesthesist
3. Romagnoli A, Keats AS. Comparative respiratory depression of tillidine and morphine. Clin Pharmacol Ther
4. Romagnoli A, Keats AS. Ceiling effect for respiratory depression by nalbuphine. Clin Pharmacol Ther
5. Scamman FL, Ghoneim MM, Korttila K. Ventilatory and mental effects of alfentanil and fentanyl. Acta Anaesthesiol Scand
6. Skatrud JB, Begle RL, Busch MA. Ventilatory effects of single, high-dose triazolam in awake human subjects. Clin Pharmacol Ther
7. Bragg P, Zwass MS, Lau M, Fisher DM. Opioid pharmacodynamics in neonatal dogs: differences between morphine and fentanyl. J Appl Physiol
8. Glass PS, Iselin-Chaves IA, Goodman D, Delong E, Hermann DJ. Determination of the potency of remifentanil compared with alfentanil using ventilatory depression as the measure of opioid effect. Anesthesiology
9. Alexander CM, Gross JB. Sedative doses of midazolam depress hypoxic ventilatory responses in humans. Anesth Analg
10. Blouin RT, Seifert HA, Babenco HD, Conard PF, Gross JB. Propofol depresses the hypoxic ventilatory response during conscious sedation and isohypercapnia. Anesthesiology
11. Alexander CM, Seifert HA, Blouin RT, Conard PF, Gross JB. Diphenhydramine enhances the interaction of hypercapnic and hypoxic ventilatory drive. Anesthesiology
12. Jordan C. Assessment of the effects of drugs on respiration. Br J Anaesth
13. Bouillon T, Schmidt C, Garstka G, et al.
Pharmacokinetic-pharmacodynamic modeling of the respiratory depressant effect of alfentanil. Anesthesiology
14. Murphy MR. Opioids. In: Barash PG, Cullen BF, Stoelting RK, eds. Clinical Anesthesia.
Philadelphia, USA: J. B. Lippincott, 1989: 255-279.
15. Bailey PL, Stanley TH. Narcotic intravenous anesthetics. In: Miller RD, ed. Anesthesia.
New York, USA: Churchill Livingstone, 1990: 281-366.
16. Gutstein HB, Akil H. Opioid analgesics. In: Hardman JG, Limbird LE, eds. Goodman & Gilman's The Pharmacological Basis of Therapeutics,
10th edn. New York, USA: McGraw-Hill, 2001: 569-621.
17. Bouillon T, Schmidt C, Garstka G, et al.
Comparison of the respiratory depressant effect of alfentanil and piritramide with a pharmacokinetic/pharmacodynamic model. Anesthesiology
18. Ausems ME, Hug CC Jr, Stanski DR, Burm AG. Plasma concentrations of alfentanil required to supplement nitrous oxide anesthesia for general surgery. Anesthesiology
19. Jennett S. Assessment of respiratory effects of analgesic drugs. Br J Anaesth
20. Jaffe RS, Fung DL. Constructing a heart rate variability analysis system. J Clin Monit
21. Task Force of the European Society and the North American Society of Pacing and Electrophysiology. Heart rate variability. Circulation