Comparisons between Desflurane and Isoflurane or Propofol on Time to Following Commands and Time to Discharge: A Metaanalysis

Searching computer databases identified all of the studies included in the metaanalysis. The Medline database was searched to identify articles recorded between 1966 and November 1994. The Reference Update database was searched to find articles recorded between week 48, 1993, and week 47, 1994. Natural language terms (i.e., key words) in the title and/or abstract gave us the highest sensitivity for relevant manuscripts. Controlled vocabulary terms (i.e., subject headings) were not used, because they identified many irrelevant manuscripts. We first read the titles of all papers identified by combining "desflurane" with "interval," "time," "duration," or "recovery." We excluded articles with clearly irrelevant titles. The abstract, and if necessary the text, was read to establish whether the article was relevant for this metaanalysis. Supplementing our search for published manuscripts using other information sources identified no additional articles.

Studies were included in the metaanalysis only if they met the following criteria: patients were randomly assigned to receive desflurane or another maintenance anesthetic, extubation was planned at the end of surgery, and general anesthesia was induced with an intravenous agent. Never was there ambiguity about whether a study satisfied the criteria. All the studies satisfied the first criterion, because they were all randomized. We originally planned to give greater statistical weight to blinded studies. However, we found no studies that were blinded. The second criterion had the effect of excluding studies involving coronary artery bypass graft surgery. The third criterion was included because we evaluated desflurane as a maintenance anesthetic, not as an induction agent.

We originally hoped to include studies comparing desflurane to halothane. However, the perioperative conditions of the two studies we identified [1,2]were different. Therefore, we did not consider pooling the results to be justifiable and did not include halothane in the metaanalysis.

Data were extracted from each manuscript. The following outcomes were identified as important before the literature search was started: (1) time from discontinuation of anesthesia until patients followed simple (i.e., hand-squeezing) commands, (2) time from discontinuation of anesthesia until patients left the first stage (i.e., higher acuity level) of recovery, (3) time from discontinuation of anesthesia until ambulatory surgery patients were discharged from the hospital, (4) group sample size, (5) duration of anesthesia, and (6) dose of volatile anesthetic. Criteria for transfer from the first to the second stage of recovery were inconsistent among studies. Therefore, times to leaving first stage were eliminated from the analysis. However, several papers comparing desflurane to propofol included time from discontinuation of anesthesia until patients sat. We therefore added this outcome to the study but did not analyze it statistically. For the comparison of desflurane to isoflurane, only one study measured time to discharge to home. [3]Therefore, this variable was not included for the comparison of desflurane to isoflurane. In one study comparing desflurane to isoflurane, there were three groups: desflurane, nitrous oxide/desflurane, and nitrous oxide/isoflurane. [4]To be consistent with most of the other desflurane versus isoflurane studies, we compared the nitrous oxide/desflurane and nitrous oxide/isoflurane groups. Dose of volatile anesthetic was reported as the last end-tidal multiple of the minimum alveolar concentration before the agent was turned off. [5-7].

Studies were statistically evaluated individually and collectively. Confidence intervals for differences between mean durations were calculated for each study using Student's t distribution and pooled standard deviations.

Metaanalysis was used to calculate confidence intervals for pooled mean differences. [8]Pooled mean differences were considered statistically significant if the 95% confidence interval did not include zero. Homogeneity tests examine whether studies can be described as sharing a common mean difference. However, tests of homogeneity can have low statistical power for rejecting the null hypothesis of homogeneity. Therefore, tests of homogeneity among the mean differences were not done before pooling. Instead, we took the conservative approach to finding the pooled mean's standard error. We used random-effects metaanalysis (unweighted variance method) to combine individual study data into a pooled result. [8]Random effects metaanalysis applies regardless whether the mean differences from different studies are homogeneous. The random-effects model permits differences among observed outcomes in studies not only because of sampling error but also because there may be true, underlying population differences.

We also qualitatively examined dependence of the mean differences on characteristics of the subjects in the studies. We hypothesized before review of the literature that duration of anesthesia and/or relative equality of doses of anesthetics might affect results. However, we did not test these relationships statistically, because they were not the focus of the study. We were concerned that statistically analyzing the relationships would sufficiently increase the number of statistical comparisons that the overall error rate for this metaanalysis would be unacceptably high.

We considered studies that used propofol for induction and compared desflurane to propofol for maintenance (Table 1). All studies used 60% N²O, except one that did not use nitrous oxide in the desflurane group. [9]For all studies, doses of the maintenance agent were adjusted as needed to maintain acceptable hemodynamic parameters.

There was no statistically significant difference in time to following commands between desflurane and propofol (mean difference 0.7 min (propofol minus desflurane), 95% confidence interval - 0.2 to 1.7 min; Figure 1). Patients who received propofol were discharged to home a mean of 17 min (4-30 min) more quickly than patients who received desflurane (Figure 2). There was a trend for a positive correlation between duration of anesthesia and the time to following commands (Table 1). There also was a trend for patients who received propofol to be able to sit up a mean of 10 min earlier than patients who received desflurane.

The studies comparing desflurane to isoflurane differed in several ways (Table 2). Most studies [10,11]used 60% N²O for maintenance. All studies included young to middle-aged adults, except one that studied elderly patients. [12].

Being aware of these differences among the studies, we pooled the results. Patients who received desflurane followed commands a mean of 4.4 min (3.3-5.4 min) earlier than patients who received isoflurane (Figure 3). We did not find a dependence of the mean differences on the duration of anesthesia (Table 2). There was a tendency for studies in which end-tidal concentration was kept constant and equal to have larger mean differences in time to following commands. Pooling these studies, patients who received desflurane followed commands a mean of 5.3 min earlier than patients who received isoflurane.

We used research synthesis (i.e., metaanalysis) to compare differences between drugs both systematically and quantitatively. The metaanalyses show that, when deciding whether to use desflurane instead of propofol or isoflurane, differences in times to following command may not be clinically important. We did not find a difference in the time to following simple commands between desflurane and propofol (Figure 1). The 95% confidence interval was narrow (-0.2 to 1.7 min) and was much smaller than confidence intervals from each of the original studies. Studying more subjects would not be likely to change the clinical relevance of the result. In contrast, there was a slight advantage to using desflurane instead of isoflurane (Figure 3). Patients awakened a mean of 4.4 min (95% confidence interval 3.3-5.4 min) more quickly after desflurane than after isoflurane. Whether this difference is clinically important is a matter for the individual anesthesiologist. Pooling data from several studies does not decrease the importance of critically examining each study for weaknesses. Nevertheless, we believe the value of this metaanalysis is that, by pooling data, the analysis generated narrow confidence intervals for the mean differences. Such narrow confidence intervals should allow clinicians to know more precisely what can be expected by using desflurane rather than propofol or isoflurane with regard to awakening and following commands.

Propofol may offer an advantage over desflurane with respect to discharge to home after ambulatory surgery. Patients who received propofol were discharged to home a mean of 17 min (4-30 min) earlier than patients who received desflurane (Figure 2). We do not have data as to whether this was due to a difference in the rate of nausea between agents.

The metaanalysis suggests that the differences between desflurane and isoflurane or propofol are small, at least with respect to rate of recovery from anesthesia. In fact, we may have overestimated the benefits of desflurane, because we used only published studies in our metaanalysis. Publication bias may have prevented studies showing smaller differences between drugs from being published. Nevertheless, in the studies that compared desflurane to propofol, the drugs were titrated to effect. Therefore, we expected differences between desflurane and propofol in times to following command to be small. However, we had expected greater differences between desflurane and isoflurane, based on animal studies. [13]The lower tissue/gas and tissue/blood partition coefficients of desflurane versus isoflurane [14]did not translate into a much more rapid recovery from anesthesia. Two explanations can account for the species differences. First, values for recovery in rats may not quantitatively apply to the time to follow commands in humans. [13]Second, humans received other drugs that may have tended to equalize differences between groups.* We believe that the latter explanation is important. Several studies that compared desflurane to isoflurane administered equal doses of the two drugs (Table 2). These studies tended to have larger (5.3 vs. 4.4 min) mean differences between desflurane and isoflurane in time to following commands.

A weakness of our research synthesis is the absence of other relevant time data from the studies we reviewed. For example, we could not reliably analyze times to postanesthesia care unit discharge between desflurane and isoflurane. These times are important for clinicians who must decide which of these drugs to use.

In conclusion, we used the methods of metaanalysis to estimate the mean decrease in times to following commands when desflurane was being used instead of isoflurane or propofol. There was no difference in the time to following commands after discontinuation of desflurane versus propofol. However, patients who received propofol were discharged to home more quickly (mean 17 min) than patients who received desflurane. There was a slight (mean 4.4 min) advantage to using desflurane over isoflurane with respect to the rate of awakening from general anesthesia.

* Eger EI II: Discussion (S53) of White PF: Studies of desflurane in outpatient anesthesia. Anesth Analg 75:S47-S53, 1992.