Effects of Volatile Anesthetics on Mortality and Postoperative Pulmonary and Other Complications in Patients Undergoing Surgery: A Systematic Review and Meta-analysis

APPROXIMATELY 234 million major surgical procedures are undertaken worldwide every year,¹  and about 7 million patients develop major complications that result in 1 million deaths during surgery or hospital stay.¹  Even if solely anesthesia-related mortality with 34 per million is extremely low, in-hospital postsurgical mortality lies between 0.5 and 4%.^2,3  Both pulmonary and other (nonpulmonary) complications contribute importantly to increase morbidity and mortality in surgery patients in the postoperative period.^3–5 

The type of anesthesia has the potential to affect pulmonary and other complications, since several anesthetic agents may confer organ protection. In experimental studies, the modern volatile anesthetics sevoflurane,⁶  desflurane,⁷  and isoflurane⁸  have been shown to reduce the size of myocardial infarction. Such effect has been ascribed to different mechanisms, including inhibition of mitochondrial permeability transition pores as well as activation of complex signaling pathways in the myocardial cells.^9,10  Also in the lungs, volatile anesthetics may protect against lung injury¹¹  and attenuate inflammation.¹²  However, in contrast to experimental studies, clinical trials have shown inconsistent results regarding organ protection by volatile anesthetics,^13–15  which seems to be restricted to patients undergoing cardiac surgery.^16–18  To our knowledge, the protective effects of volatile anesthetics against postoperative complications in a mixed surgical patient population, consisting of cardiac and noncardiac surgery, have not been addressed so far.

We conducted a systematic review and meta-analysis of randomized controlled trials (RCTs) comparing volatile anesthetics with total IV anesthesia (TIVA) regarding patient outcome. We hypothesized that, compared to TIVA, the use of volatile anesthetics during general anesthesia is associated with reduced mortality and incidence of postoperative pulmonary and other complications in noncardiac and cardiac surgery patient populations.

We conducted a systematic review of RCTs in accordance with a previously registered protocol (International Prospective Register of Systematic Reviews registration no. CRD42014008699). The presented review was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.¹⁹  A complete PRISMA checklist is presented in table 1 of Supplemental Digital Content 1, http://links.lww.com/ALN/B270.

Three databases (EMBASE, MEDLINE, The Cochrane Central Register of Controlled Trials [CENTRAL]) were systematically searched for relevant trials published as journal article from inception until August 23, 2014 without language restriction. Personal files and reference lists of relevant review articles for additional trials were also reviewed. Detailed search strings and reference lists are listed in table 2 of Supplemental Digital Content 1, http://links.lww.com/ALN/B270.

We included only RCTs with patients undergoing cardiac or noncardiac surgical interventions. Studies investigating modern volatile anesthetics (sevoflurane, desflurane, isoflurane) versus TIVA, with respect to mortality or postoperative pulmonary or other complications were considered.

Inclusion criteria with respect to “patients, intervention, comparator, outcomes, study design” (PICOS) criteria²⁰  were as follows: (1) population: adult patients (age more than 18 yr) undergoing general anesthesia for elective or also emergency surgery; (2) intervention: patients receiving anesthesia with volatile anesthetics; (3) comparator: volatile anesthetics versus TIVA, or single volatile anesthetics versus each other; (4) outcomes: mortality (primary endpoint), as the longest reported mortality, or in-hospital mortality, or 30-day mortality. Secondary endpoints were 90-day, 180-day, or 1-yr mortality (according to the definition by the authors of the original article), or postoperative pulmonary complications²¹  (hypoxemia, acute respiratory distress syndrome [ARDS], pulmonary infiltrates, pneumonia, pleural effusions, atelectasis, pneumothorax, bronchospasm, cardiopulmonary edema, aspiration pneumonitis; for detailed description, see table 3 of Supplemental Digital Content 1, http://links.lww.com/ALN/B270), or other postoperative complications (including overall cardiac events, myocardial infarction, acute renal failure, hepatic failure, disseminated intravasal coagulation, extrapulmonary infection, gastrointestinal failure, coma; for detailed description, see table 4 of Supplemental Digital Content 1, http://links.lww.com/ALN/B270), or intensive care unit (ICU) length of stay (LOS, days) or ICU-free days, LOS hospital (days); (5) study design: RCT. Postoperative pulmonary and other complications represented collapsed composites of the respective single complications. A trial was included if the predefined PICOS criteria with respect to population, intervention, comparison study, and at least one primary outcome were reported. RCTs published as abstract or letter only or addressing nitrous oxide were excluded.

The articles for this review were selected by examining titles, abstracts, and the full text if a potentially relevant trial was identified. We translated non-English reports into English, as necessary. Two reviewers (C.U. and T.B.), independently and in duplicate, abstracted data on the a priori defined inclusion PICOS criteria. An attempt was made twice to contact authors in order to request any necessary information not contained in articles.

Risk of bias was assessed as previously described in detail by our group elsewhere.²²  Briefly, in duplicate and independently, two reviewers (C.U. and T.B.) assessed trial methodological quality using the risk of bias tool recommended by the Cochrane Collaboration.²³  For each trial, the risk of bias was reported as “low,” “unclear,” or “high” in the following domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias.²³  We assessed the item “selective outcome reporting” as “unclear” risk if the study protocol was not published or registered previously. For each outcome, we independently and in duplicate rated the overall quality of evidence (confidence in effect estimates) using the Grading of Recommendations, Assessment, Development and Evaluations approach in which trials begin as high-quality evidence, but may be rated down by one or more of five categories of limitations: risk of bias, inconsistency, indirectness, imprecision, and reporting bias.²⁴  Finally, the overall risk of bias for an individual trial was categorized as “low” (if the risk of bias was low in all domains), “unclear” (if the risk of bias was unclear in at least one domain, with no high risk of bias domains), or “high” (if the risk of bias was high in one or more domains). Disagreement was resolved by discussion and consensus.

For direct comparison of volatile anesthetics and TIVA, the pooled odds ratio (OR) and 95% CIs of binary outcomes were calculated using the OR method as described elsewhere.²⁵  For direct and indirect comparison of the specific anesthetic agent (sevoflurane vs. desflurane vs. isoflurane vs. TIVA), a multivariate random effects network meta-analysis was performed.^26,27  These data are presented as network plot, plots of cumulative ranking probability, and forest plot.^28–30  For ICU and hospital LOS, the mean difference (MD) and 95% CI were calculated. Detailed description of statistical methods regarding Peto OR meta-analysis and the multivariate random effects network meta-analysis is given in Supplemental Digital Content 2, http://links.lww.com/ALN/B271.

Statistical heterogeneity was assessed by the I² statistic. Substantial heterogeneity was predefined as I² greater than 50%, which differed from the protocol previously published (International Prospective Register of Systematic Reviews registration no. CRD42014008699) in order to be even more conservative. Publication bias was addressed visually using a funnel plot. The Egger³¹  regression models for publication bias were also used. An a priori defined subgroup analysis of outcomes in cardiac and noncardiac surgical procedures was performed. Heart surgery, regardless of the use of cardiopulmonary bypass, was included in the cardiac surgery group. A posteriori the originally intended analysis of a mixed surgical population combining cardiac and noncardiac surgery patients was discarded because of the artificiality of such a population. Therefore, only the results of the subgroup analysis are reported.

All analyses were performed using STATA (Version MP 11; Stata Corp LP, USA).

The search yielded 5,073 publications, from which the flowchart is depicted in figure 1. Twenty-one articles published in languages other than English and German were translated into English to assess their eligibility. Of 1,076 potentially eligible studies, 46 were not an RCT, 13 studies did not match the population criteria, 80 had a nonselected comparator anesthetic agent (nitrous oxide), 24 compared general anesthesia to regional anesthesia, and 845 studies did not report on the primary outcome investigated herein (mortality); all were excluded. The remaining 68 publications were retained and their data meta-analyzed. Complete references of the included articles are shown in table 5 of Supplemental Digital Content 1, http://links.lww.com/ALN/B270.

The main characteristics of the trials retained for meta-analysis are shown in figure 2 and table 1. Table 6 of Supplemental Digital Content 1, http://links.lww.com/ALN/B270, depicts detailed information of the trials. The 68 meta-analyzed RCTs included data from 7,104 patients published from 1989 to 2014. Sixty-five RCTs (6,716 patients) compared volatile anesthetics (n = 3,506 total, sevoflurane n = 1,895, desflurane n = 708, isoflurane n = 903) to TIVA (n = 3,210). Forty-five RCTs enrolled a total of 4,890 cardiac surgery patients of whom 2,587 received volatile anesthetics (sevoflurane n = 1,077, desflurane n = 673, isoflurane n = 837) and 2,303 received TIVA. Patients from surgical fields other than cardiac surgery were investigated in 20 RCTs including a total of 1,826 subjects allocated to volatile anesthetics in 919 cases (sevoflurane n = 818, desflurane n = 35, isoflurane n = 66) and to TIVA in 907 cases. Three trials (388 patients; two cardiac surgery) compared volatile anesthetics (sevoflurane n = 177, desflurane n = 58, isoflurane n = 153) among each other without a TIVA control group^32–34  and were included in the network meta-analysis only. Detailed information regarding the number of patients enrolled in the articles reporting pulmonary and other complications is given in Supplemental Digital Content 1, http://links.lww.com/ALN/B270.

The risk of bias assessment using the Cochrane tool is summarized in figure 3. Most RCTs had unclear risk of bias. Detailed information regarding risk of bias assessment is provided in table 7 of Supplemental Digital Content 1, http://links.lww.com/ALN/B270. The funnel plot for main outcomes (fig. 1 of Supplemental Digital Content 2, http://links.lww.com/ALN/B271) suggests no statistically significant publication bias.

All included trials reported either in-hospital, 30-day, 180-day, or 1-yr mortality (fig. 2). Based on our evidence synthesis of outcome data from 4,840 patients in 45 trials, we estimate that compared to TIVA, volatile anesthetics reduced the overall mortality in the subgroup of cardiac surgery (OR = 0.55; 95% CI, 0.35 to 0.85; z = 2.71; P = 0.007; I² = 8.5%; fig. 4). Meta-analysis of outcome data of 1,876 patients in 20 trials enrolling patients undergoing noncardiac surgery did not show a statistically significant effect of volatile anesthetics compared to TIVA on either overall mortality (OR = 1.31; 95% CI, 0.83 to 2.05; z = 1.17; P = 0.242; I² = 0.0%) or in-hospital mortality (OR = 1.16; 95% CI, 0.53 to 2.51; z = 0.37; P = 0.711; I² = 0.0%) (fig. 5). Network meta-analysis among the different volatile agents showed no reduction of overall and in-hospital mortality, irrespective of the surgical procedure (figs. 6 and 7). Plots of cumulative ranking probability are depicted in figure 2 of Supplemental Digital Content 2, http://links.lww.com/ALN/B271.

As depicted in figure 8, volatile anesthetics were associated with less postoperative pulmonary complications compared to TIVA in patients who underwent cardiac surgical procedures (OR = 0.71; 95% CI, 0.52 to 0.98; z = 2.07; P = 0.038; I² = 6.0%; fig. 8). In noncardiac surgery, compared to TIVA, volatile anesthetics were not associated with lower incidence of postoperative pulmonary complications (OR = 0.67; 95% CI, 0.42 to 1.05; P = 0.081; I² = 53.5%; fig. 8). Compared to TIVA, none of the single volatile anesthetics reduced pulmonary complications (figs. 6 and 7).

In cardiac surgery patients, compared to TIVA, volatile anesthetics were associated with lower incidence of other postoperative complications (OR = 0.75; 95% CI, 0.58 to 0.96; z = 2.26; P = 0.024; I² = 38.6%; fig. 8), but such association was not observed in noncardiac surgical procedures (OR = 0.70; 95% CI, 0.46 to 1.05; z = 1.78; P = 0.075; I² = 0.0%; fig. 9). Such differences were more pronounced with desflurane and sevoflurane (fig. 6). In noncardiac surgical procedures, compared to TIVA, none of the single volatile anesthetics reduced other complications (fig. 7).

The meta-analysis of ICU and hospital LOS are depicted in figures 4 to 7 of Supplemental Digital Content 2, http://links.lww.com/ALN/B271. In cardiac surgery patients, volatile anesthetics were associated with no statistical significant reduction of ICU and hospital LOS compared to TIVA (MD = −0.09; 95% CI, −0.40 to 0.22; z = 0.57; P = 0.57; I² = 89%; fig. 4 of Supplemental Digital Content 2, http://links.lww.com/ALN/B271; MD = 0.86; 95% CI, −0.84 to 2.55; z = 0.99; P = 0.32; I² = 34%, fig. 6 of Supplemental Digital Content 2, http://links.lww.com/ALN/B271; respectively). In noncardiac surgery patients, volatile anesthetics were associated with no statistical significant reduction of ICU LOS (MD = −0.68; 95% CI, −1.80 to 0.45; z = 1.18; P = 0.24; I² = 72%, fig. 5 of Supplemental Digital Content 2, http://links.lww.com/ALN/B271), but with reduced hospital LOS (MD = −0.80; 95% CI, −1.57 to −0.02; z = 2.02; P = 0.04; I² = 0%; fig. 7 of Supplemental Digital Content 2, http://links.lww.com/ALN/B271) compared to TIVA.

The main results of this systematic review and meta-analysis were that general anesthesia with volatile anesthetics, as compared to TIVA, was associated with reduced mortality and lower risk of pulmonary and other complications in cardiac, but not in noncardiac, surgery.

To our knowledge, this is the first systematic review and meta-analysis addressing the effects of volatile anesthetics on outcome in different surgical patient groups, including 68 trials that investigated 7,104 patients in total. Other strengths of our analysis are the use of PRISMA guidelines¹⁹  and a literature search in three databases without language or time restrictions. Furthermore, we incorporated a network meta-analysis with indirect and direct comparisons to evaluate the contribution of individual volatile anesthetics.

In patients undergoing general anesthesia for cardiac surgery, volatile anesthetics are associated with lower mortality. These findings are in agreement with a previous meta-analysis in the field.¹⁸  This may result from the observation of reduced postoperative complications with volatile anesthetics compared to TIVA, which are likely related to cardioprotective properties of those agents. In the heart, volatile anesthetics induce coronary vasodilation,³⁵  promote early activation of protective enzymes by phosphorylation and translocation of cellular key proteins (early phase of preconditioning),⁶  and result in delayed transcriptional changes of protective and antiprotective proteins (late phase of preconditioning).³⁶  In addition, volatile anesthetics may have decreased inflammation in the lungs, attenuating ventilator-induced lung injury. We did not find any RCTs that have addressed the role of volatile anesthetics in reducing lung inflammation in nonthoracic surgery. In a study in patients undergoing thoracic anesthesia, compared to TIVA, sevoflurane and desflurane reduced the release of proinflammatory cytokines in the ventilated lung.³⁷  Accordingly, other investigators were able to show a reduced inflammatory response of the nonventilated lung during general anesthesia with sevoflurane compared to propofol.³⁸  Those findings are in line with experimental data from lung injury models, where the use of volatile anesthetics reduced lung inflammation.^11,12  Besides the modulation of the inflammatory response, volatile anesthetics also induce bronchodilation, possibly decreasing the mechanical stress on lung units. Volatile anesthetics were found to alleviate bronchoconstriction in patients with chronic obstructive lung disease,³⁹  suggesting a possible role for this mechanism. Furthermore, the antiinflammatory effects of volatile anesthetics^40,41  are not limited to the lungs or heart, but affect also other organs, including brain,^42,43  kidneys,^44,45  and liver.^46–49  In addition, a more recent meta-analysis demonstrated a reduction of acute kidney injury and renal failure after cardiac surgery under general anesthesia with volatile anesthetics compared to TIVA.⁵⁰  Such effects might be related to favorable transcriptional changes in pro- and antiprotective mechanisms, as demonstrated for sevoflurane.⁵¹ 

The fact that, in cardiac surgery patients, lower incidences of pulmonary and other complications were not associated with reduced LOS in ICU or hospital might be due to standard operating procedures determining a minimal period of stay for this type of surgery.

In noncardiac surgery procedures, compared to TIVA, volatile anesthetics were not associated with reduced mortality. There are different possible explanations for this observation. First, since most of the protective effects of volatile anesthetics seem to be related to cardiac preconditioning^9,52  and the surgical population in the analysis also included patients less prone to cardiac complications, beneficial effects may have been diluted in the analysis of this surgical population. Second, the presence of comorbidities that influence the risk of death, for example, cancer, may have acted as confounders. Third, the population investigated may have been more heterogeneous than the cardiac surgery population. In fact, studies retained for analysis included thoracic, vascular, and abdominal surgery, likely reducing the power of the meta-analysis to identify an effect on mortality. The lack of potential benefit of volatile anesthetics in reducing pulmonary and other complications in the noncardiac surgery is possibly due to a high heterogeneity among trials. It is worth noting that studies of noncardiac surgical patients that were retained for analysis showed an association between the use of volatile anesthetics and reduced hospital LOS, compared to TIVA. However, since only a few trials reported hospital LOS, the noncardiac surgery population might have not been adequately represented, and this finding should be interpreted with caution.

The results of this systematic review and meta-analysis suggest that, in patients undergoing cardiac surgery, and provided no contraindication is present, volatile anesthetics should be preferred over TIVA as a strategy to improve postoperative outcome. Nevertheless, it must be kept in mind that coronary artery bypass graft, off-pump or on-pump with cardiopulmonary bypass, and noncoronary artery bypass graft surgery (e.g., valve surgery) have been evaluated together, whereas most patients underwent coronary artery bypass graft surgery.

In noncardiac surgery patients, volatile anesthetics seem not to be associated with a relevant outcome benefit compared to TIVA. However, mortality after noncardiac surgery was relatively high (≈7.2%), suggesting that high-risk procedures may have been selected, precluding extrapolation of the results to low- and mid-risk surgery.

Given the importance of organ protection during surgery, large RCTs investigating the potential of volatile anesthetics to reduce pulmonary and other complications also in noncardiac surgery patients are highly needed. The current results might be valuable for estimation of effect sizes and sample size calculations when designing such studies. Such a trial should be powered for mortality and assess pulmonary and other complications systematically. Currently, a large trial comparing the effects of sevoflurane, desflurane, and isoflurane with TIVA in cardiac surgery is being performed (clinicaltrials.gov identifier NCT02105610).

The present systematic review and meta-analysis has several limitations. First, the quality of the trials is heterogeneous, with relatively large overall risk of bias. Second, most trials were relatively small and either did not report any, or reported only a few events, which jeopardizes the generalizability of the findings, especially with regard to mortality. The internal validity of those trials could be compromised by underreporting, and the external validity may be affected if the study population has different properties compared to the general population. However, we made adjustments for the effect size with respect to trials with small sample size and used the Peto OR method,^25,53  which offers a more precise statistical analysis for rare events than a fixed or random effects model as proposed by Mantel and Haenszel⁵⁴  or DerSimonian and Liard.⁵⁵  Furthermore, a network meta-analysis was performed, which included all trials regardless of the number of events. Third, publications within a time period of three decades were included, and we cannot exclude the possibility that other time-dependent factors, for example, advances in postoperative care, influenced the results. Fourth, we cannot rule out that publication bias did impair our analysis, since negative results are more likely not to be published. Fifth, pulmonary and other complications were analyzed as collapsed composite outcomes from events with different degrees of severity. For example, ARDS and atelectasis were counted in the same way. In addition, the assessment and definition of those complications differed across the trials retained in the analysis, increasing the risk of bias. Sixth, postoperative complications are importantly influenced by the type of surgery,^3,56,57  which may overwhelm the role of anesthetic agents. Nevertheless, we conducted a separate analysis of cardiac and noncardiac surgery. Furthermore, in cardiac surgery, coronary artery bypass graft, off-pump or on-pump with cardiopulmonary bypass, and noncoronary artery bypass graft surgery (e.g., valve surgery) have been evaluated together, whereas most patients underwent coronary artery bypass graft surgery. Thus, generalization of the findings for all types of cardiac surgery may be inappropriate. Seventh, mortality rates were relatively low in cardiac surgery, likely due to sampling bias of small trials and use of narrow inclusion criteria that might have resulted in inclusion of mainly low- and medium-risk procedures. Eighth, we did not address the use of opioids, which might have potential cardiac protective effects.^58,59 

In cardiac, but not in noncardiac, surgery, compared to TIVA, general anesthesia with volatile anesthetics was associated with major benefits in outcome, including reduced mortality, as well as lower incidence of pulmonary and other complications. Further studies are warranted to address the impact of volatile anesthetics on outcome in noncardiac surgery.

The authors express their gratitude to Silke Thrun, LIB, Library of the Departments of Surgery and Anesthesiology and Intensive Care Medicine, University Hospital Dresden, Dresden, Germany, and Susanne Henninger Abreu, B.Pd., R.N., Department of Anesthesiology and Intensive Care Medicine, University Hospital Dresden, Technische Universität Dresden, Dresden, Germany, for providing articles; Roman Rodionov, M.D., Ph.D., University Center for Vascular Medicine and Department of Medicine III, Section Angiology, University Hospital Dresden, Technische Universität Dresden; Martin Böke, Ph.D., Institute of Ethnology, University of Cologne, Cologne, Germany; Anja Eckardt, M.D., Department of Gynecology and Obstetrics, Oberlausitz-Kliniken Bautzen, Bautzen, Germany; and Lorenzo Ball, M.D., Department of Surgical Sciences and Integrated Diagnostics, IRCCS San Martino IST, University of Genoa, Genoa, Italy, for their assistance with article translation. The authors are indebted to Michael Andreae, M.D., Department of Anesthesiology, Montefiore Medical Center, Albert Einstein College of Medicine, Bronx, New York, for revising the article.

The authors declare no competing interests.