Nitrous oxide is commonly used during noncardiac surgery, and in view of the large number of patients exposed worldwide every year, good evidence for its beneficial or harmful effects is desirable. However, the available large randomized trial1 and observational studies2–4 report conflicting results, which limit reliable conclusions about the value of nitrous oxide.
Nitrous oxide may reduce potent hypnotic and opioid requirements intraoperatively and improve acute and chronic pain outcomes postoperatively.5 However, the effect of nitrous oxide on perioperative cardiac, pulmonary, and thrombotic outcomes is unclear.1–4 Nitrous oxide use may increase the risk of myocardial infarction (MI) via increased plasma homocysteine concentrations and endothelial dysfunction after surgery.3 Investigating the effectiveness of anesthetic management using randomized trials is challenging6,7; therefore, analysis of large prospectively collected data sets is warranted.4
The Perioperative Ischemic Evaluation (POISE) trial randomized 8351 patients with or at risk of ischemic heart disease having noncardiac surgery to 30 days of extended-release metoprolol succinate or placebo.8 Metoprolol was associated with a decreased risk of nonfatal MI (4.2% vs 5.7%; P = 0.002) but with an increased risk of stroke (1.0% vs 0.5%; P = 0.005) and death (3.1% vs 2.1%; P = 0.03) compared with placebo. The POISE trial required complete prospective collection of many relevant indices throughout the 30-day study period. With the significant adverse event rates and the substantial use of nitrous oxide in the trial, this data set provides an opportunity to further explore the effects of nitrous oxide on major outcomes, although this was not the original aim of POISE. The primary aim of this post hoc subanalysis, therefore, was to determine the effects of nitrous oxide on the incidence of a composite outcome of cardiovascular death, nonfatal MI, and nonfatal cardiac arrest occurring within 30 days after randomization in the POISE trial patients.
METHODS
The POISE trial protocol has been described in detail elsewhere and was registered with ClinicalTrials.gov (NCT00182039).8,9 In summary, patients were eligible for this multicenter, blinded, randomized controlled trial if they were aged 45 years, were undergoing noncardiac surgery with an expected hospital length of stay ≥24 hours, and fulfilled at least 1 of the following criteria: history of coronary artery disease, peripheral vascular disease, stroke, hospitalization for congestive heart failure, undergoing major vascular surgery, or any 3 of 7 criteria (intrathoracic or intraperitoneal surgery, history of congestive heart failure, transient ischemic attack, diabetes, serum creatinine ≥175 µmol/L, age ≥70 years, or undergoing emergency surgery). Patients were excluded if they met the following criteria: heart rate <50 beats per minute, second-degree or third-degree heart block, asthma, already receiving a β-blocker, coronary artery bypass surgery within 5 years and no cardiac ischemia since, low-risk surgical procedures, receiving verapamil, or previous randomization into the trial. All centers obtained IRB approval, and all patients provided written informed consent for the original trial. Patients were only recruited once, and data relate to the index surgery and not any other surgery. Patients who were related to another participant were not excluded. IRB pendency was not maintained, and approval to add investigators who assisted with these analyses was not sought (these investigators did not have access to identifiable data nor individual case report forms).
A total of 8351 patients from 190 hospitals in 23 countries were randomly assigned to extended-release metoprolol succinate or placebo, starting 2 to 4 hours before surgery and continuing for 30 days. Study medication was only administered if the heart rate was ≤50 beats per minute and the systolic blood pressure was ≥100 mm Hg. Patients were monitored with cardiac biomarker assays and electrocardiographs during the 30-day follow-up period, and cardiovascular outcomes were adjudicated centrally by a blinded committee. The dosage and monitoring regimens were described in detail previously.8,9
The primary outcome was a composite of cardiovascular death, nonfatal MI, and nonfatal cardiac arrest occurring within 30 days of randomization. Secondary outcomes included MI and stroke. Clinically significant hypotension was defined as a systolic blood pressure of <90 mm Hg requiring fluid resuscitation, an inotropic drug or study drug discontinuation, or intraaortic balloon pump.9 Testing the effect of nitrous oxide on these outcomes was not the primary purpose for which the POISE trial was designed.
Nitrous oxide anesthesia was defined as coadministration of nitrous oxide in patients receiving general anesthesia, with or without additional neuraxial blockade or nerve blockade.
Data Analysis
Baseline characteristics were summarized as number (%) for categorical variables and mean (SD) for continuous variables and were compared between intervention groups using χ2 tests and analysis of variance, respectively.
Use of nitrous oxide was left to the discretion of the attending anesthesiologists; that is, such use was not randomly assigned. Patient characteristics were therefore imbalanced between intervention groups, so we used a propensity score technique to account for potential confounding. The propensity score is the probability of receiving the intervention, modeled as a function of observed variables, and can be used in various ways to adjust for confounding because of observed characteristics.10,11 We adopted an inverse probability weighted approach, which uses the propensity score to create a “pseudopopulation” in which all measured characteristics are balanced between the intervention groups, hence removing any confounding by these characteristics (but not necessarily by unmeasured or unknown confounders).12 This is akin to the balance that is expected to be observed among measured variables in a randomized trial; however, it is not akin to the balance one would expect with unknown prognostic variables in a large randomized trial. A key assumption in the use of such methods is that every patient must have a nonzero probability of receiving each studied intervention. For nitrous oxide, patients not receiving a general anesthetic were excluded.
The propensity score for nitrous oxide use was estimated using a logistic regression model in which the outcome was the intervention group. An iterative procedure was used to select independent variables to include in the model, initially including main effects for all characteristics listed in Tables 1 and 2 and adding interaction terms until no further imbalance could be removed. All analyses were repeated using a more comprehensive propensity score model, including all region-by-covariate interactions, to investigate and account for geographic differences in intervention allocation.
;)
Table 1: Nitrous Oxide Baseline Characteristics (n = 5104)
Table 2: Estimated Associations with Outcome for Nitrous Oxide
We then imposed the “common support” condition,11 in which we excluded any patients in the intervention group who had an estimated propensity score higher than that of any patient in the no-intervention group, and any patients in the no-intervention group with a propensity score lower than that of any patient receiving the intervention. This removed subjects for whom there was no comparable subject in the other intervention group, because the effect of the intervention cannot be estimated for such patients.
Each subject was inversely weighted by the probability of that subject receiving the intervention that they did receive (calculated using the propensity score). Within the weighted sample (the pseudopopulation), measured patient prognostic characteristics should be balanced between the intervention groups. This was assessed using standardized differences (the difference in the percentage [or mean] of each characteristic between intervention groups, divided by an estimate of the standard deviation and expressed as a percentage)13 calculated both for the original sample and the weighted sample. It has been suggested that a standardized difference of ≥10% is a meaningful imbalance.14
Odds ratios (OR) and 95% confidence intervals (CIs) for the effect of nitrous oxide on the primary composite outcome, MI, and stroke were estimated using logistic regression models for each outcome, including the intervention group as the sole independent variable, by applying a weighted analysis as described earlier. Characteristics that remained imbalanced in the weighted sample were additionally included as independent variables in the logistic regression model to remove any residual confounding by these variables.
To assess between-region variability in effect, an interaction between intervention and geographic region was additionally included in the weighted logistic regression model. Estimated associations with the primary outcome for nitrous oxide by region were also calculated.
To assess the sensitivity of results to a few individuals with high weights, we excluded 1% of subjects with the highest weights and repeated all the analyses. Subjects with missing data for surgery type or baseline heart rate or systolic blood pressure were excluded from all analyses. Analyses were conducted using Stata version 11.1 (Stata Corp., College Station, TX). P < 0.05 was considered statistically significant.
RESULTS
For nitrous oxide, 5104 patients remained after exclusions were made for patients who did not receive general anesthesia (n = 3139), for those with missing data (n = 30), and for the common support condition (n = 78).
Patients receiving nitrous oxide were more likely to have ischemic heart disease and were less likely to be having intra-abdominal or emergency surgery than those not receiving nitrous oxide (Tables 1). There was also significant regional variation in the administration of nitrous oxide, ranging from 5% in Central/South America to 80% in India. Imbalances between the intervention groups in the initial sample were reduced to minimal levels by the propensity score weighting (Tables 1). In particular, the groups were well balanced for the randomized β-blocker treatment.
There was no evidence of association between nitrous oxide and the risk of the primary outcome (OR, 1.08; 95% CI, 0.82–1.44; 99% CI, 0.75–1.57; P = 0.58), MI (OR, 0.99; 95% CI, 0.75–1.31; 99% CI, 0.69–1.42; P = 0.94) or stroke (OR, 0.85; 95% CI, 0.26–2.82; 99% CI, 0.17–4.11; P = 0.79), death (OR, 1.04; 95% CI, 0.6–1.81; 99% CI, 0.51–2.15; P = 0.88), or clinically significant hypotension (OR, 0.92; 95% CI, 0.74–1.15; 99% CI, 0.70–1.23; P = 0.48; Tables 2). Trimming the weights modified the ORs slightly (1.14, 1.13, 0.63, 0.88, and 0.94, respectively); however, all P values remained > 0.27, thereby retaining the conclusions from the untrimmed data. There was no evidence to support an effect of nitrous oxide on the primary outcome based on region (Table 3).
Table 3: Estimated Associations with the Primary Outcome for Nitrous Oxide by Region
DISCUSSION
We found no evidence that nitrous oxide had an effect on the primary outcome (cardiovascular death, nonfatal MI, and nonfatal cardiac arrest), MI, stroke, or clinically significant hypotension in the POISE trial patients. We did, however, find marked geographic differences in the rates of nitrous oxide administration, suggesting differences in regional preferences and ongoing uncertainty of the benefits and risks of nitrous oxide. Our results should, however, be considered in light of the study’s limitations, outlined as follows.
This result is possibly because of a lack of power but, nevertheless, is consistent with previous literature on this subject,1,3,4 despite the plausible pathophysiologic rationale3 and clinical trial support15 for increased myocardial ischemia in nitrous oxide patients. The trend toward an increased rate of death and MI at 30 days was not statistically significant in the Evaluation of Nitrous Oxide in the Gas Mixture for Anaesthesia-I trial,1 although a long-term follow-up revealed a marginal increase in the risk of MI.3 In contrast, Shiba et al.4 reported no effect on 30-day mortality and a decrease in hospital morbidity in nitrous oxide patients when analyzing a large administrative data set. We thus have uncertainty regarding this important clinical issue for anesthesiologists. Consequently, we are conducting a randomized trial of nitrous oxide–based versus nitrous oxide–free anesthesia in 7000 noncardiac surgery patients who have or are at risk of ischemic heart disease (the Evaluation of Nitrous Oxide in the Gas Mixture for Anaesthesia–II trial; www.enigma2.org.au).6 The contrasting patterns of use around the world reported in the present study illustrate the uncertainty about the risk–benefit ratio for nitrous oxide and further support the need for compelling evidence. In the meantime, anesthesiologists who omit nitrous oxide from their general anesthesia plan in high-risk patients should not do so on the basis that the risk of important morbidity will definitely be reduced, because these results and others are not definitive.
The present analyses have several limitations. Firstly, the POISE trial was not designed with the primary purpose of testing the effects of nitrous oxide on the primary or secondary outcomes. The patients in the POISE trial were not randomized to nitrous oxide, and the use of this drug was at the discretion of the attending anesthesiologist. Although we included a propensity analysis to account for the decision to use nitrous oxide, we were limited by the data available. In addition, the original protocol was not developed with this analysis in mind. There are likely to be unmeasured and unknown factors that influenced the choice of nitrous oxide that may provide an alternative explanation for our findings.
Interest in and experience with nitrous oxide use varies among anesthesiologists, and there are certainly institutional and regional differences. Although nitrous oxide use is not commonly discussed with patients, surgeons may be concerned about the effects of nitrous oxide during bowel or middle ear surgery.16 In the present study, the only datapoints that captured these factors were the regional variations in the use of these techniques, and we included these in our propensity score.
Secondly, we did not record the details of nitrous oxide use when it was administered (i.e., timing of initiation, inspired nitrous oxide concentration, and duration of administration). Randomized trials on this intervention need to control or record these factors. The dose of nitrous oxide determines extent of hyperhomocysteinemia and its possible consequences.17 Lastly, we did not collect data on other aspects of anesthesia care, such as the use of volatile-based or propofol-based maintenance of anesthesia, opioid use, or postoperative care, on the actual operation undertaken, or on the success of surgery (particularly in relation to vascular graft patency). Nitrous oxide was the only component of the general anesthetic technique that was recorded.
In conclusion, we found no evidence that nitrous oxide was associated with any of these outcomes. These results have important implications for future research.
DISCLOSURES
Name: Kate Leslie, MBBS, MD, MEpi, FANZCA.
Contribution: This author helped design and conduct the study, and write the manuscript.
Attestation: Kate Leslie reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.
Name: Paul Myles, MBBS, MD, MPH, FANZCA, FCARSCI, FRCA.
Contribution: This author helped design and conduct the study, and write the manuscript.
Attestation: Paul Myles reviewed the analysis of the data and approved the final manuscript.
Name: Philip J. Devereaux, MD, PhD.
Contribution: This author helped design and conduct the study, and write the manuscript.
Attestation: Philip J. Devereaux has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Andrew Forbes, MSc, PhD.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Andrew Forbes has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Purnima Rao-Melancini, MSc.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Purnima Rao-Melancini has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Elizabeth Williamson, PhD.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Elizabeth Williamson has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Shouchun Xu, MD.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: Shouchun Xu approved the final manuscript.
Name: Pierre Foex, MD, DPhil, FRCA, FANZCA, FCA(SA), FmedSci.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: Pierre Foex approved the final manuscript.
Name: Janice Pogue, MSc.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Janice Pogue has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Maribel Arrieta, MD.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: Maribel Arrieta approved the final manuscript.
Name: Gregory L. Bryson, MD, FRCPC, MSc.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: Gregory L. Bryson approved the final manuscript.
Name: James Paul, MSc, MD, FRCPC.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: James Paul approved the final manuscript.
Name: Michael J. Paech, MBBS, DM, DRCOG, FRCA, FANZCA, FFPMANZCA, FRANZCOG (Hon).
Contribution: This author helped conduct the study and write the manuscript.
Attestation: Michael J. Paech approved the final manuscript.
Name: Richard N. Merchant, MD, FRCPC.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: Richard N. Merchant approved the final manuscript.
Name: Peter T. Choi, MD, MSc, FRCPC.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: Peter T. Choi approved the final manuscript
Name: Neal Badner, MD, FRCPC.
Contribution: This author helped conduct the study and write the manuscript
Attestation: Neal Badner approved the final manuscript.
Name: Philip Peyton, MBBS, MD, FANZCA.
Contribution: This author helped conduct the study and write the manuscript
Attestation: Philip Peyton approved the final manuscript
Name: John W. Sear, MA, BSc, MBBS, PhD, FFARCS, FANZCA.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: John W. Sear approved the final manuscript.
Name: Homer Yang, MD, CCFP, FRCPC, CCPE.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: Homer Yang approved the final manuscript.
This manuscript was handled by: Sorin J. Brull, MD, FCARCSI (Hon).
REFERENCES
1. Myles PS, Leslie K, Chan MT, Forbes A, Paech MJ, Peyton P, Silbert BS, Pascoe EENIGMA Trial Group. . Avoidance of nitrous oxide for patients undergoing major surgery: a randomized controlled trial. Anesthesiology. 2007;107:221–31
2. Leslie K, Myles PS, Forbes A, Chan MT. The effect of bispectral index monitoring on long-term survival in the B-aware trial. Anesth Analg. 2010;110:816–22
3. Leslie K, Myles PS, Chan MT, Forbes A, Paech MJ, Peyton P, Silbert BS, Williamson E. Nitrous oxide and long-term morbidity and mortality in the ENIGMA trial. Anesth Analg. 2011;112:387–93
4. Shiba S, You J, Saager L, Sessler D, Turan ANitrous oxide and postoperative outcomes after non-cardiac surgery.. American Society of Anesthesiologists Annual Meeting. San Diego. 2010
5. Stiglitz DK, Amaratunge LN, Konstantatos AH, Lindholm DE. Intraoperative nitrous oxide as a preventive analgesic. Anaesth Intensive Care. 2010;38:890–3
6. Myles PS, Leslie K, Peyton P, Paech M, Forbes A, Chan MT, Sessler D, Devereaux PJ, Silbert BS, Jamrozik K, Beattie S, Badner N, Tomlinson J, Wallace SANZCA Trials Group. . Nitrous oxide and perioperative cardiac morbidity (ENIGMA-II) Trial: rationale and design. Am Heart J. 2009;157:488–494.e1
7. McPeek B. Inference, generalizability, and a major change in anesthetic practice. Anesthesiology. 1987;66:723–4
8. POISE Study Group. . Effects of extended-release metoprolol succinate in patients undergoing non-cardiac surgery (POISE trial): a randomised controlled trial. Lancet. 2008;371:1839–47
9. POISE Trial Investigators. . Rationale, design, and organization of the PeriOperative ISchemic Evaluation (POISE) Trial: a randomized controlled trial of metoprolol versus placebo in patients undergoing noncardiac surgery. Am J Heart. 2006;152:223–230
10. Rosenbaum P, Rubin D. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41–55
11. Williamson E, Morley R, Lucas A, Carpenter J. Propensity scores: from naive enthusiasm to intuitive understanding. Stat Methods Med Res. 2012;21:273–93
12. Hernán MA, Robins JM. Estimating causal effects from epidemiological data. J Epidemiol Community Health. 2006;60:578–86
13. Austin PC, Grootendorst P, Anderson GM. A comparison of the ability of different propensity score models to balance measured variables between treated and untreated subjects: a Monte Carlo study. Stat Med. 2007;26:734–53
14. Sharon-Lise T, Normand S-L T, Landruma M, Guadagnolia E, Ayaniana J, Ryand T, Clearya P, McNeil B. Validating recommendations for coronary angiography following acute myocardial infarction in the elderly: a matched analysis using propensity scores. J Clin Epidemiol. 2001;54:387–98
15. Badner NH, Beattie WS, Freeman D, Spence JD. Nitrous oxide-induced increased homocysteine concentrations are associated with increased postoperative myocardial ischemia in patients undergoing carotid endarterectomy. Anesth Analg. 2000;91:1073–9
16. Myles PS, Leslie K, Silbert B, Paech MJ, Peyton P. A review of the risks and benefits of nitrous oxide in current anaesthetic practice. Anaesth Intensive Care. 2004;32:165–72
17. Myles PS, Chan MT, Leslie K, Peyton P, Paech M, Forbes A. Effect of nitrous oxide on plasma homocysteine and folate in patients undergoing major surgery. Br J Anaesth. 2008;100:780–6
