Predictors for postoperative nausea and vomiting after xenon-based anaesthesia

Abstract

Postoperative nausea and vomiting (PONV) severely impair patient satisfaction and rank together with pain among the most undesirable outcomes following general anaesthesia.¹ The identification of risk factors for PONV forms the basis for adequate prophylaxis and treatment of patients with a high probability of experiencing PONV. In addition to patient-related risk factors, inhaled anaesthetics induce PONV depending on the duration of exposure.^2,3

As a noble gas, xenon is chemically inert in physiological conditions and therefore free of metabolites. As a result of its low solubility in blood, it is rapidly eliminated from the body during weaning from anaesthesia.^4,5 Furthermore, xenon antagonizes serotonin type 3 (5-HT₃) receptors that mediate nausea and vomiting during chemotherapy and after general anaesthesia.² Taken together, one might assume that the incidence of PONV after xenon-based anaesthesia is rather low, yet Coburn and colleagues⁶ described an incidence of PONV of more than 60%. Thus, identification of risk factors could facilitate titration of prophylactic antiemetics during xenon-based anaesthesia. However, with respect to pharmacokinetic and pharmacodynamic differences compared with other inhaled anaesthetics, it is not known whether risk factors for PONV are valid in the course of xenon-based anaesthesia or if prevalent pharmacological prophylaxis is sufficiently effective in this setting. Therefore, our aim was to determine (i) the risk factors for PONV and (ii) the efficacy of routinely administered antiemetic prophylaxis after xenon-based anaesthesia.

Methods

Here we present data from a previously unpublished prospective multicentre study performed to evaluate the safety and efficacy of xenon-based anaesthesia with subjects recruited between April 2009 and February 2011. After institutional review board approval (Ethik-Kommission der Ärztekammer Berlin, study number ETH-019/08, and Heinrich-Heine Universität Düsseldorf, study number 3386) and registration at the German Federal Institute for Drugs and Medical Devices (BfArM, study number AL-PMS-01/07GER), all patients classified as ASA status I–II (age 18 yr or older) undergoing surgery during xenon-based general anaesthesia were eligible for this study after written informed consent was obtained. Patients with elevated intracranial pressure, pulmonary disease, coronary artery disease, and impaired left ventricular function were excluded. Induction and maintenance of xenon-based anaesthesia was conducted at the discretion of the attending anaesthetists.

Assessment of postoperative nausea and vomiting

Subjects were followed for 24 h after extubation by study physicians. The incidence of nausea, vomiting, or both within the 24 h period was assessed by medical chart inspection followed by a personal patient interview and recorded as a binary variable. The requirement for postoperative antiemetic medication (‘rescue medication’) was assessed from medical charts.

Incidence of postoperative nausea and vomiting and quantification of independent predictors

The observed incidence of PONV was compared with the expected incidence predicted by Apfel Score.⁷ The initial Apfel Score was corrected for the administered postoperative opioids. This comparison was performed only in subjects who did not receive antiemetic prophylaxis to determine the unimpeded emetic activity of xenon-based anaesthesia.

We assessed predictors for PONV or rescue medication after xenon-based anaesthesia by logistic regression analysis. A recent, large meta-regression identified female sex, history of PONV or motion sickness, non-smoking status, younger age, duration of anaesthesia, use of postoperative opioids, and certain types of surgery as independent predictors for PONV after propofol or inhaled anaesthetic-based anaesthesia.⁸ Therefore, we decided a priori to include these variables in the model. The comparison of postoperative opioid consumption was facilitated by calculation of morphine equivalents. In addition, different classes of medical antiemetic prophylaxis and study centres were also included as single binary variables in the model a priori. The aim of the next step was to identify further potential predictors by testing remaining variables (height, weight, body mass index, amount of intraoperative fluids, type of intraoperative opioid, and use of regional anaesthesia) for association with PONV or rescue medication by univariate analysis. As no statistically significant associations were found, we did not include these variables in our logistic regression analysis. Variables within the model were tested for collinearity using the collin extension of Stata (P. Ender, University of California, Los Angeles, CA, USA). Goodness of fit was assessed using the Hosmer–Lemeshow test with 10 groups.

Effectiveness of medical antiemetic prophylaxis

The effect of prophylactic antiemetics was part of the logistic regression analysis. However, we additionally performed a propensity score-matched analysis so that subjects who did or did not receive prophylactic antiemetics were comparable. To this end, a binary variable was generated indicating whether a subject received medical antiemetic prophylaxis or not. Then, a logistic regression model including sex, age, smoking status, history of PONV or motion sickness, regional anaesthesia, duration of anaesthesia, anticipated postoperative opioid use, and study centre was used to calculate the propensity for receiving medical antiemetic prophylaxis. Finally, subjects with prophylaxis were matched on a one-to-one basis with patients without prophylaxis on the logit of the propensity score using callipers of 0.2 sd of the logit.⁹ Univariate analyses were performed to verify that groups were balanced on the variables used for calculation of the propensity score (i.e. that matching was successful). Additionally, standardized differences were calculated to quantify balancing of groups.¹⁰

The sample size estimation of the underlying study was carried out based on the primary end points depth of anaesthesia and incidences of hypertension and anaesthesia. However, when comparing an overall PONV incidence associated with general anaesthesia of 38%¹¹ and a previously reported PONV incidence of 27.5% after xenon-based anaesthesia,¹² a minimal sample size of 364 patients would be required. On this basis, we considered the available number of 500 patients included in this study to be sufficient. In addition, post hoc power analyses of the logistic regression analysis and the propensity score-matched analysis were performed. Results are presented as adjusted odds ratios with the 95% confidence interval (CI), means (sd), median [interquartile range] or absolute numbers (percentage). A two-tailed P-value <0.05 was considered statistically significant. Univariate analyses were performed using Fisher's exact test, Mann–Whitney U-test, and in the case of normally distributed variables, Student's t-test. Calculations were made with Stata IC version 10.0 (StataCorp LP, College Station, TX, USA) and post hoc power analysis with G*Power 3.1.5.¹³

Results

A total of 502 patients undergoing xenon-based general anaesthesia were enrolled in this study. Of those, 488 were included in the final analysis (Fig. 1).

Fig 1

Study flow chart with reasons for exclusion of 14 patients. PONV, postoperative nausea and vomiting.

Open in new tabDownload slide

Management of anaesthesia

Attending physicians chose propofol [cumulative dose 6.4 (3.4) mg kg⁻¹ until start of xenon] for anaesthesia induction and initial maintenance until the end of denitrogenation in all but three subjects, for whom etomidate [0.29 (0.04) mg kg⁻¹] was used as induction agent. Rocuronium [463 subjects, 0.6 (0.2) mg kg⁻¹] was the primarily administered Neuromuscular blocking agent. Systemic analgesia was achieved with remifentanil [292 subjects, 10.8 (3.6) µg kg⁻¹ h⁻¹], fentanyl [118 subjects, 2.1 (1.4) µg kg⁻¹ h⁻¹], or sufentanil [78 subjects, 0.24 (0.17) µg kg⁻¹ h⁻¹]. Finally, 111 subjects received additional regional anaesthesia (Table 1). Xenon wash-in started 26 (12) min after induction of anaesthesia. In three subjects, xenon administration was discontinued before the end of surgery. One of these subjects was a 23-yr-old morbidly obese (body mass index 48 kg m⁻²) female patient undergoing resection of a vulvar dysplasia, who experienced a decrease of oxygen saturation to <90% during xenon wash-in. The second was a 66-yr-old female patient with a body mass index of 32 kg m⁻² undergoing laparoscopic nephrectomy with insufficient elimination of CO₂ as a result of elevated airway pressure at 1 h after xenon wash-in. In the third subject, xenon was discontinued at 45 min after wash-in because of technical problems with the anaesthetic machine. After discontinuation of xenon, anaesthesia was maintained with propofol in the first two subjects and isoflurane in the last subject. These subjects were not excluded after an intention-to-treat analysis.

Incidence of postoperative nausea and vomiting and independent predictors

A total of 136 (28%) subjects suffered from PONV after xenon-based anaesthesia. In subjects without medical antiemetic prophylaxis, PONV occurred in 92 (28%) of subjects. The expected incidence in this subpopulation was 42% (95% CI 42.0–45.0) considering a median Apfel Score of 2 [1;3] (P<0.001 vs observed), and the area under the receiver operating characteristic curve for the Apfel Score was 0.60. The 24 h PONV after xenon-based anaesthesia was predicted by female gender, younger age, and longer duration of anaesthesia (Table 2). Female sex and duration of anaesthesia were additional risk factors for postoperative need for rescue medication (2.09; 1.28–3.43, P=0.003; and 1.24 per hour; 1.06–1.44, P=0.006, respectively). Non-smoking status, history of PONV or motion sickness, and amount of postoperative morphine equivalents were not significant predictors.

Efficacy of medical antiemetic prophylaxis

About one-third of all subjects received antiemetic prophylaxis, with 72% of those subjects receiving two or more classes of antiemetics (Table 1). Subjects receiving prophylaxis were more often female, younger, more likely to be a non-smoker, with a significantly higher Apfel Score, and required less postoperative opioids (left part of Table 3).

In the logistic regression model, none of the antiemetics (neither dexamethasone nor 5-HT₃ antagonists) was associated with a significant reduction in PONV. Furthermore, after propensity score matching of 182 subjects, no statistically significant difference in incidence of PONV or need for rescue medication was detected between the two groups (right part of Table 3).

Discussion

Young age, female gender, and longer duration of anaesthesia were risk factors for PONV after xenon-based anaesthesia.

Incidence of postoperative nausea and vomiting

The incidence of PONV was 28% and thus comparable with previous studies,^12,14 but it was less frequent than predicted by the Apfel Score.⁷ Xenon is a potent antagonist at the 5-HT₃ receptor¹⁵ that mediates nausea and vomiting.¹⁶ Therefore, it has been hypothesized that xenon exerts intrinsic antiemetic properties. In contrast, Coburn and colleagues⁶ demonstrated a higher incidence of PONV after xenon and remifentanil anaesthesia than after propofol and remifentanil (66 vs 27%). However, patients who experienced PONV in the xenon and remifentanil group less often required antiemetic rescue medication than patients suffering from PONV after propofol and remifentanil (55 vs 84%). This observation indicates that although PONV occurred more frequently after xenon and remifentanil, it may have been less severe than after propofol and remifentanil.

Independent predictors

Duration of anaesthesia was a strong predictor for PONV and need for rescue medication; the incidence of PONV was roughly doubled with every additional 2 h of exposure. This relationship has also been known for other inhaled anaesthetics¹⁷ and nitrous oxide,³ and appears to be independent of the targeted receptors. Given that xenon has a very low blood–gas partition coefficient of 0.115¹⁸ and does not accumulate,^4,5 it is questionable whether it can directly trigger PONV several hours after anaesthesia. However, xenon has been traced in human blood and urine up to 24 h after anaesthesia,^19,20 probably because of storage in fatty tissue.²¹ Rather than triggering PONV by direct mechanisms, alterations in gene expression and protein function may contribute to PONV after xenon inhalation. For example, xenon facilitates preconditioning by inducing hypoxia-inducible factor-1α.^22,23 It also upregulates a variety of genes involved in neuronal signalling,²⁴ alters excitability of brain cells by increasing neuronal Ca²⁺ concentration,^25,26 and interferes with neuronal norepinephrine re-uptake.²⁷ However, given that the molecular mechanisms of PONV have yet to be elucidated, it remains highly speculative whether these xenon-induced cellular alterations contribute to PONV.

Not all items of the Apfel Score were predictive in our patient population. The confidence intervals for non-smoking status and history of PONV, motion sickness, or both were too wide to reach statistical significance, and effect sizes for those two factors were smaller than after ‘traditional’ anaesthesia.⁸ It seems reasonable that PONV after anaesthesia with ‘classic’ inhaled anaesthetics or propofol does not necessarily predict PONV after xenon-based anaesthesia because of different mechanisms of action. However, we cannot completely rule out the possibility that in a larger sample, non-smoking status and history of PONV/motion sickness may be detected as significant predictors.

Postoperative nausea and vomiting have been widely attributed to postoperative opioid use.^7,28 In our model, the dose of postoperative opioids was not predictive for PONV. It is known that administration of postoperative opioids is correlated with the duration of anaesthesia.^7,29 When testing for collinearity, we found a positive correlation between postoperative morphine equivalent use and duration of anaesthesia (Pearson's r=0.21, P<0.001). Given that the correlation was weak, we decided a priori to include both variables in the model. However, this may explain why, in our multivariate logistic regression analysis, postoperative opioid use was not predictive for PONV, because the duration of anaesthesia was the stronger predictor. Indeed, when the duration of anaesthesia was excluded, postoperative opioid use was a significant predictor for 24 h PONV (1.03 per milligram morphine equivalent, 1.00–1.06, P=0.024).

Efficacy of antiemetic prophylaxis

We were unable to demonstrate a prophylactic effect of dexamethasone or 5-HT₃ antagonists for preventing PONV after xenon-based anaesthesia. Dexamethasone was administered in doses of 4–8 mg, and 5-HT₃ antagonists were predominantly granisetron (1–3 mg), tropisetron (2–4 mg), or ondansetron (4–8 mg). Therefore, drug doses were adequate and have been shown to prevent PONV after anaesthesia with traditional inhaled anaesthetics.^30–32 The 5-HT₃ antagonists are ineffective as rescue medication in PONV occurring after previous prophylactic treatment with a substance of the same class.^33,34 Given that xenon also exerts antagonism at 5-HT₃ receptors,¹⁵ it seems reasonable that administering a 5-HT₃ antagonist for prophylaxis after xenon might not provide additional benefit. Unfortunately, we did not record data on the efficacy of rescue medication. Droperidol, dimenhydrinate, or metoclopramide were given in a limited number of subjects that was too low to draw conclusions on their prophylactic effects. Based on our findings, we suggest administration of antiemetics targeting receptors other than serotonin antagonists in high-risk patients (i.e. young female patients with longer lasting procedures) undergoing xenon-based anaesthesia. Of note, a prospective randomized controlled trial failed to show a significant beneficial effect of dexamethasone as antiemetic prophylaxis and ondansetron as rescue medication for PONV after xenon and remifentanil anaesthesia (NCT00793663).³⁵

Limitations

The data reported were obtained in a prospective cohort study investigating adverse events during xenon-based anaesthesia so that prophylaxis was not protocolized, but administered at the discretion of the attending anaesthetist. Thus, it would not have been appropriate to compare the raw incidences of PONV in patients who did or did not receive prophylactic antiemetics. This issue was addressed by controlling for other factors in the logistic regression analysis. Furthermore, we conducted a thorough propensity score matching with the objective of creating two groups of patients, with or without antiemetics, which were similar in all relevant factors. After matching was completed, univariate analyses verified that the two groups were similar in all variables except for antiemetics. To confirm these results, we also assessed the quality of the propensity score matching by calculating standardized differences. No generally agreed threshold exists to indicate relevant imbalance between groups, but standardized differences of <0.1 have been considered to represent a negligible difference.^10,36 Although the matched groups were considered to be well balanced, a marginal difference in the distribution of female patients (standardized difference 0.14) and postoperative opioid admission (standardized difference 0.12) between groups cannot be excluded. Accordingly, a minor but most likely clinically irrelevant effect of antiemetic prophylaxis might have been missed.

Postoperative nausea and vomiting were measured as a robust yes–no binary variable by physicians who visited subjects at least once, 24 h after the end of anaesthesia. Thus, underreporting of PONV is possible because subjects might not remember or might have deemed it not severe enough to report.³⁷ However, considering additional documentation of rescue medication, we are sure to have assessed ‘clinically relevant’ PONV that required targeted prophylaxis.

As we were unable to detect a protective effect of routine antiemetic prophylaxis, a possible β error because of lack of power must be considered. Dexamethasone or ondansetron reduces the incidence of PONV by 26% after general anaesthesia with propofol or inhaled anaesthetics.³² Given that a majority of our subjects received two or three antiemetics, we conducted a post hoc power analysis, calculating a 91% power to detect the effect of a combination of two antiemetic medications in our matched cohort of 182 subjects (two-tailed α=0.05). Even when antiemetic prophylaxis was substituted by a binary ‘dummy’ variable, logistic regression analysis did not indicate an effect of any prophylactic medication, although power was high (38 subjects per factor, 83% power, two-tailed α=0.05).^38,39

Conclusions

This is the first study to investigate risk factors for PONV after xenon-based anaesthesia. While incidence of PONV was significantly lower than predicted by Apfel Score, female sex, young age, and long duration of anaesthesia predicted PONV after xenon-based anaesthesia. A history of PONV or non-smoking status failed to reach statistical significance, probably because of lower power associated with a smaller effect size. The lack of efficacy of 5-HT₃ receptor antagonists is reasonable, because xenon antagonizes these receptors itself, yet needs to be interpreted carefully and warrants confirmation by a sufficiently powered randomized controlled study.

Authors' contributions

M.S.S.: data collection, data analysis, and manuscript preparation. C.C.A.: data analysis and manuscript preparation. H.-J.S., R.S., B.B., P.H.T., M.H., and M.R.-H.: study design, patient recruitment, data collection, and manuscript revision. M.N.: patient recruitment, data collection, and manuscript revision. P.K.: patient recruitment, data collection, data analysis, and manuscript preparation.

Declaration of interest

M.S.S., C.C.A., R.S., M.H., and M.N.: none declared. H.-J.S. has received payments and travel funding for introduction of xenon and anaesthetic machines from Air Liquide Medical, Germany. B.B. has received honoraria for consulting and giving lectures from CSL Behring, Abbvie, and Edwards Life Sciences Medical, Germany, and is a member of the Medical Advisory Boards of Air Liquide, MSD Pharma, Pulsion Medical Systems, Orion Pharma, 3M, and Ratiopharm GmbH. P.H.T. has received research funding and travel grants from Air Liquide Medical, Germany. M.R.-H. has received fees for lectures from Air Liquide international. P.K. has received funding for research from Air Liquide Medical, Germany and fees for lectures and travelling from Air Liquide Medical, Germany and Baxter, Germany.

Funding

Air Liquide Deutschland GmbH supported the study financially, but was involved in neither the study design nor the data collection, analysis, interpretation, or reporting of the data. Development of the study design was performed by an external study monitor. Data collection was performed by the study centres themselves, controlled by an external monitor. Analyses, interpretation of the data, and drafting of the manuscript were performed by the investigators as mentioned above.

Acknowledgements

We would like to thank Alexander T. Dilthey for statistical advice.

References

Author notes