Study Design and Ethics

The study was designed and performed as a mono-center, factorial, randomized controlled clinical trial according to the Declaration of Helsinki. The factorial design consisted of several distinct study arms [21–23], of which the occurrence and prophylaxis of nausea following xenon anesthesia is subjected here. The study subjects were blinded to the randomized anesthetic, and the prophylactic antiemetic treatment was administered in a double-blinded, placebo-controlled manner.

This study was designed to test nausea incidence following volatile anesthesia in PONV-sensitive subjects (Apfel score ≥ 2). Thus, the design inherited a possibly higher risk for participants to develop nausea and/or vomiting. All subjects were explicitly informed about this possibly elevated risk of nausea and/or vomiting during their study participation before obtaining written informed consent. Furthermore, it was determined in the study protocol that randomized study subjects who dropped out of the study procedure were not replaced to avoid any possible risk of repeating cases with severe nausea. Ethical approval for this study was provided by the ethics committee of the Medical Faculty of the University Hospital RWTH Aachen in Germany on 24 October 2008 (Chairperson: Professor G. Schmalzing, MD; reference number: EK 110/08). This design was furthermore approved by the German official authority for supervision and approval of pharmaceuticals (Bundesinstitut fuer Arzneimittel und Medizinprodukte). The trial was registered at the EMA (EudraCT No.: 2008-004132-20) and at ClinicalTrials.gov (NCT No.: 00793663) (http://clinicaltrials.gov).

Subjects

Enrolment was performed between November 2008 and March 2011. Study physicians screened all patients scheduled for elective abdominal surgery (i.e. gynecological, urological or abdominal surgery, with either open or laparoscopic approach) with a planned duration ≥ 60 min and a planned admission to the ward after post anesthesia care unit (PACU) stay. This procedure led to a screening rate of approx. 5 patients per day and an approx. total of 3000 patients over the recruitment period. A screening log was not implemented. After written informed consent, 220 patients aged 18–75 years, ASA status I-II and Apfel’s simplified PONV predicting score of 2–4 (i.e. a PONV likelihood of about 40–80%) [3;24] were enrolled in the trial. Enrolment was performed by the study physicians. Among exclusion criteria were severe cardiac, respiratory, liver or kidney function disorders, known or suspected allergic reaction to one of the study medications, suspicion of malignant hyperthermia, pregnancy/lactation period, legal incapacity to give informed consent and refusal to participate.

Randomization

Two independent permuted-block randomizations were performed; the first for type of anesthesia (Xenon vs. Sevoflurane; block size 10; allocation ratio 1:1), and the second for type of prophylactic (dexamethasone vs. placebo; block size 10; allocation ratio 1:1). The randomization lists were generated by computer allocation using a randomization-software (RandList version 1.2, DatInf) by a member of the clinic staff independent from the study team and sealed in an opaque envelope and locked in an office cupboard only accessible for the head investigator in case of emergency unblinding due to a serious adverse event. Randomization codes for the type of anesthesia were sealed in opaque envelopes parameterized with the corresponding study number on the outside, and were opened by the study physician on the day of surgery shortly before the start of intervention. Anesthesia technique was blinded to the participants throughout their trial participation. Randomization codes for the type of prophylactic treatment were consigned to the hospital pharmacy, who received an email with the according study number on the eve of surgery for each individual participant. The hospital pharmacy then provided a blinded syringe containing the randomized prophylactic treatment in equal amount of liquid, marked with the individual study number. Participants, study physicians, and outcome assessors were blinded to the type of prophylactic treatment.

Trial Procedure

A sketch of the study procedure is displayed in Fig 1. Following premedication with oral midazolam 7.5 mg 45 min before induction, general anesthesia was induced by a bolus of propofol (2.0 mg¹kg^-1 initially, repeating dose if considered necessary 0.5–1.0 mg¹kg^-1) and 0.5 μg¹kg^-1min^-1 remifentanil infusion for 60s. Endotracheal intubation was facilitated by 0.6 mg¹kg^-1 rocuronium. Xenon or sevoflurane wash-in was started with a target end-tidal concentration of 60±5 vol.% xenon or 2±0.2 vol.% sevoflurane in min. 30% oxygen. Within the first hour patients received a single injection of randomized prophylactic treatment (dexamethasone 4 mg or placebo NaCl 0.9%, provided by the hospital pharmacy in a blinded syringe, same liquid volume with the concordant study number). General anesthesia was maintained through inhalation of xenon or sevoflurane and supported by remifentanil infusion titrated to clinical needs. Standard monitoring included pulse oximetry, three channel ECG, non-invasive arterial blood pressure measurement, control of the tube cuff pressure, neuromuscular monitoring, core temperature, assessment of end-tidal oxygen, end-tidal carbon dioxide and end-tidal anesthetic gas concentration. Vital parameters (i.e. heart rate, systolic and diastolic blood pressure (AS/3 monitor, GE Datex-Ohmeda, Helsinki, Finland), end-tidal O₂, CO₂) were automatically recorded throughout anesthesia at 10s intervals and logged every five min. End-tidal concentrations of anesthetics and infusion rate of remifentanil were logged at the same intervals. Depth of anesthesia was assessed by standard physiological criteria [21]. Normoxia, a physiological carbon dioxide concentration and normal body temperature were maintained. Medical quality xenon in steel cylinders was provided by Air Liquide Santé International (Paris, France). Sevoflurane was provided by Abbott (Wiesbaden, Germany). Both anesthetic agents were administered using a closed circuit respirator (Felix Dual^®, Taema, France) with concordant software allowing the use of xenon only under closed circuit conditions [25].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Study procedure.

A simplified sketch of the study procedure is shown in Fig 1. After obtaining written informed consent and baseline examination, patients were first randomized to receive xenon or sevoflurane anesthesia for their individually scheduled surgery, then randomized to receive dexamethasone or placebo for prophylaxis of post-operative nausea. Within the first hour of anesthesia, the randomized prophylactic treatment was applied. Patients were monitored for 24h following anesthesia in fixed intervals (during PACU stay at 0, 5, 15, 30, 45, 60, 75, 105, 120 min until discharge as well as 2h, 6h, and 24h post anesthesia). The study-specific survey was completed 24h post anesthesia.

https://doi.org/10.1371/journal.pone.0153807.g001

Standard treatment of blood loss, fluid replacement and hemodynamic support were applied when indicated. Twenty minutes ahead of the estimated end of surgery piritramide 0.05 mg¹kg^-1, and either paracetamol 10 mg¹kg^-1 or metamizole 15 mg¹kg^-1 were administered for post-anesthetic pain management. Remifentanil was stopped when all painful interventions were terminated. Anesthesia was discontinued by washing out of the inhalational anesthetic via high-flow 100% oxygen when all surgical interventions including bandaging were terminated and complete recovery from neuromuscular block was assured. Extubation was accomplished when patients’ upper airway reflexes were recovered, breathing was regular and sufficient (breathing frequency >8 min^-1; breathing volume >6 ml¹kg^-1), and patients were able to react on demand (eye opening, swallowing, breathing). Patients were observed in the operation room until full recovery, and then transported to the post anesthetic care unit (PACU) for further observation. Pain medication (piritramide 0.02–0.05 mg¹kg^-1) was titrated to clinical needs.

Nausea was assessed asking subjects about their individual nausea levels rated with an eleven-point numeric rating scale (NRS) with ‘0’ indicating no nausea and ‘10’ indicating worst nausea ever experienced. Patients were measured at fixed time-points upon admission to the PACU, after five and fifteen minutes and thereafter in fifteen-minute intervals. The primary outcome parameter was the development of subjective nausea, defined as NRS ≥ 2, at any time-point in the 0-2h PACU interval. Episodes of vomiting were recorded at the same time points.

Follow-up examinations were performed at 6h, and 24h after anesthesia and patients were questioned about the incidence and their specific levels of nausea during these time periods (2-6h and 6-24h post anesthesia) with an eleven-point NRS as described above. Occurrence of vomiting and number of vomiting episodes were assessed in the same intervals. At the end of their study participation (24h post anesthesia) patients were asked to rate the individual quality of their anesthesia with a six-point rating scale according to the German school grade system, with 1 as best and 6 as worst.

Statistics

The primary outcome parameter was the development of early-onset nausea defined as NRS ≥ 2 at any time-point in the 0-2h PACU interval. The parameter nausea was selected as the most sensitive (in general nausea as a symptom of PONV appears first) and most reliable (subjects do not vomit without nausea) determinant characterizing PONV. Secondary outcome parameters were the occurrence and intensity of early-onset nausea, and the influence of dexamethasone on early onset nausea. The incidence of vomiting during the first two hours following anesthesia and the occurrence of late-onset nausea and vomiting during the 24h post-anesthesia survey were evaluated as well.

The a priory calculation of the quantity of subjects and subsequent power of the trial was performed for the total of several independent study arms as described in the study design section [21–23], using nQuery Advisor^® Version 7.0 (Statistical Solutions, Saugus, Massachusetts, USA). Based on an expected mean Apfel score of 3, a PONV -and thereby nausea- incidence of about 60% in the sevoflurane group and a similar incidence in the xenon group were hypothesized. A reduction of nausea by dexamethasone of about a third (i.e. from 60% to 40%) was assumed to be relevant. The a priory power was calculated for the H0 hypothesis that dexamethasone prophylaxis did not cause a different level of reduction of nausea with xenon anesthesia compared to sevoflurane anesthesia. With an alpha error of 0.05 (i.e.) and a beta error of 0.2 (i.e. 80% power), we calculated a need of 107 patients per group. A total of 110 patients per group was planned for inclusion to account for drop-outs.

When planning for the trial and a priory calculation were implemented, we didn’t take into account that there might be a difference in nausea incidence between xenon and sevoflurane anesthesia, which was noted when the statistical analysis was performed. Due to this important observation, a high drop-out rate entailing a lower participant number for analysis than a priory expected plus the fact that the trial protocol was lacking a total statistical analysis plan, we decided to change the analysis parameters and conduct a post-hoc analysis with the primary outcome parameter incidence of post-operative nausea defined as above and the H0 hypothesis that xenon anesthesia had no different incidence of nausea than sevoflurane anesthesia. We performed an intention to treat-analysis, including those patients who had completed the randomized anesthesia and had withdrawn preterm because of severe nausea, but did not account for the dropped-out participants due to other reasons (see Fig 1) that had not completed the randomized anesthesia (i.e. no implementation of missing values to obtain the full analysis set). The primary outcome parameter was evaluated with a risk analysis via binary logistic regression (influence of xenon, dexamethasone, and the combination of dexamethasone and xenon on the risk to develop significant nausea), followed by a direct comparison via one-way ANOVA. The secondary outcome parameters, including the isolated incidences of significant nausea following xenon and sevoflurane anesthesia, were directly compared with bivariate analyses. Logistic regression was performed as the primary efficacy analysis; all further p-values were understood merely descriptive. Statistical analysis was performed using IBM SPSS statistics software version 20.0 (SPSS Inc., Chicago, Illinois, USA). Figures were generated with GraphPad PRISM^® (GraphPad Software Inc., La Jolla, California, USA).

Limitations

Several factors in the design and realization of this study might limit our findings. We designed this study as a multifactorial study with several study arms which assessed independent questions within one setting. Even if independency of the study arms were ensured, there might have been hidden interactions or influences that have biased our study observations. Additionally, due to the non-replacement of subjects in case of drop-out (see methods‘ section, study design) and a high study drop-out rate due to other reasons (see Fig 1) we did not achieve the initially estimated patient number. This might have further influenced our obtained results, especially in the case of the secondary analysis performed for dexamethasone prophylaxis. However, the median Apfel score achieved in this study was 2 and thereby lower than presupposed for the calculated number of participants. In consequence this fact subsequently led to a lower nausea rate in the study population than would have been eligible especially for secondary analyses.

The statistical planning of the study also holds several limitations: The a priory calculation didn’t take into account the observed difference in nausea incidence between xenon and sevoflurane anesthesia. Since this difference entailed an impact on the primary endpoint definition and thus had a direct influence on all further estimations, a change in statistical analysis from the primarily planned calculation towards a post-hoc evaluation was necessary. The post hoc analysis within this existing sample might have led to a bigger type one error than expected just by looking at the p-value. Post-hoc power calculations are only marginally able to prevent this bias. Moreover, the study is likely to be underpowered with respect to at least one of the study arms in this linked design. A further lack of total statistical analysis plan in the design and the exclusion of missing values handling in the protocol and thereby loss of the full analysis set for statistical calculation have further weakening influences on the conclusions form our results. An interim analysis of the primary outcome parameter would have been very helpful to detect the tendencies concerning different incidences of nausea in both study groups, but was unfortunately neither planned nor performed for the primary endpoint initially planned for this study arm.

Within our study set we had a trend towards failure to receive allocated interventions (see Fig 1), with a maximum in the xenon group. Since reasons for this failure were seemingly manifold and incoherent (withdrawn consent, safety reasons), a missing at random assumption and treatment of missing values might have helped to identify further factors that might have strengthened the outcome of the statistical analysis. However, the handling of missing values was neither planned a priory nor implemented into the post-hoc analysis.