Canadian Journal of Anesthesia/Journal canadien d'anesthésie

, Volume 58, Issue 8, pp 714–724

Meta-analysis of desflurane and propofol average times and variability in times to extubation and following commands

Authors

  • Ruth E. Wachtel
    • Department of AnesthesiaUniversity of Iowa
    • Department of AnesthesiaUniversity of Iowa
    • Department of Health Management & PolicyUniversity of Iowa
  • Richard H. Epstein
    • Department of AnesthesiologyJefferson Medical College
  • Johannes Ledolter
    • Department of Management SciencesUniversity of Iowa
Reports of Original Investigations

DOI: 10.1007/s12630-011-9519-1

Cite this article as:
Wachtel, R.E., Dexter, F., Epstein, R.H. et al. Can J Anesth/J Can Anesth (2011) 58: 714. doi:10.1007/s12630-011-9519-1

Abstract

Purpose

We performed a meta-analysis to compare the operating room recovery time of desflurane with that of propofol.

Methods

Studies were included in which a) humans were assigned randomly to propofol or desflurane groups without other differences between groups (e.g., induction drugs) and b) mean and standard deviation were reported for extubation time and/or time to follow commands. Since there was heterogeneity of variance between treatment groups in the log-scale (i.e., unequal coefficients of variation of observations in the time scale), generalized pivotal methods for the lognormal distribution were used as inputs of the random effects meta-analyses.

Results

Desflurane reduced the variability (i.e., standard deviation) in time to extubation by 26% relative to propofol (95% confidence interval [CI], 6% to 42%; P = 0.006) and reduced the variability in time to follow commands by 39% (95% CI, 25% to 51%; P < 0.001). Desflurane reduced the mean time to extubation by 21% (95% CI, 9% to 32%; P = 0.001) and reduced the mean time to follow commands by 23% (95% CI, 16% to 30%; P < 0.001).

Conclusions

The mean reduction in operating room recovery time for desflurane relative to propofol was comparable with that shown previously for desflurane relative to sevoflurane. The reduction in variability exceeded that of sevoflurane. Facilities can use the percentage differences when making evidence-based pharmacoeconomic decisions.

Méta-analyse des temps moyens du desflurane et du propofol et de leur variabilité au niveau des temps jusqu’à l’extubation et la réponse à un ordre

Résumé

Objectif

Nous avons réalisé une méta-analyse afin de comparer le temps de récupération en salle d’opération après une administration de desflurane par rapport au propofol.

Méthode

Les études dans lesquelles a) des patients ont été randomisés en groupes recevant du propofol ou du desflurane sans autre différence entre les groupes (par ex. médicaments d’induction) et b) la moyenne et l’écart type étaient rapportés pour le temps jusqu’à extubation et/ou le temps nécessaire à la réponse à un ordre, ont été retenues. En raison de l’hétérogénéité du point de vue de la variance entre les groupes de traitement sur une échelle logarithmique (c.-à-d. des coefficients inégaux de variation des observations dans l’échelle de temps), des méthodes pivot généralisées pour la distribution log-normale ont été utilisées pour saisir les méta-analyses d’effets aléatoires.

Résultats

Le desflurane a réduit la variabilité (c.-à-d. l’écart type) en matière de temps jusqu’à extubation de 26 % par rapport au propofol (intervalle de confiance [IC] 95 %, 6 % à 42 %; P = 0,006) et a réduit la variabilité en matière de temps jusqu’à la réponse à un ordre de 39 % (IC 95 %, 25 % à 51 %; P < 0,001). Le desflurane a réduit le temps moyen jusqu’à extubation de 21 % (IC 95 %, 9 % à 32 %; P = 0,001) et réduit le temps moyen jusqu’à la réponse à un ordre de 23 % (IC 95 %, 16 % à 30 %; P < 0,001).

Conclusion

La réduction moyenne du temps de récupération en salle d’opération lors de l’administration de desflurane par rapport à du propofol était comparable à celle précédemment démontrée lors de la comparaison du desflurane au sévoflurane. La réduction de la variabilité a dépassé celle du sévoflurane. Les établissements peuvent utiliser les différences de pourcentage pour prendre des décisions pharmaco-économiques fondées sur des données probantes.

Many hospitals strive to reduce their non-operative operating room (OR) time, i.e., time in the OR when surgery is not being performed. Reducing non-operative time can reduce labour costs for ORs with more than eight hours of cases.1-5

Many surgeons focus on non-operative time. Vitez and Macario asked surgeons to score the importance of particular attributes of anesthesia groups using a scale from 0 to 4; 0 = “no importance”, and 4 = “a factor that would make me switch groups/hospitals”.6 The mean score was 4.0 for “ability to calmly manage a crisis”. The mean score was only slightly less (3.9) for “patient quick to awaken”, demonstrating the importance surgeons place on promptly beginning the next case.

We previously used data from an anesthesia information management system to model the time from end of surgery to tracheal extubation.7 We applied that knowledge to perform meta-analyses of trials comparing extubation times following maintenance with desflurane and sevoflurane.7 Desflurane reduced the mean extubation time relative to sevoflurane by 25% and reduced the standard deviation by 21%.7 Desflurane reduced the mean extubation time relative to isoflurane by 34% and reduced the standard deviation by 36%.8

In our earlier analyses, we assumed that the coefficient of variation does not vary according to treatment, i.e., type of anesthetic. The assumption held for desflurane vs sevoflurane (see the Results section 2 and Fig. 5 of reference 7). However, we illustrate in the Appendix that the assumption does not hold true for desflurane vs propofol. We modified the statistical analysis by using generalized pivotal methods to account for differences in the coefficient of variation between groups. To explain the method, we use data from a small observational study of the times required to prepare propofol for the next case. In this article, we applied generalized pivotal methods to compare OR recovery times between desflurane and propofol.9-34

Methods

To identify published manuscripts comparing OR recovery times after desflurane and propofol in humans, we searched PubMed on January 10, 2011 using the following criteria: desflurane AND (propofol OR Diprivan) AND (extubation OR extubate OR command OR recover OR recovery OR cost), limited to humans. The search yielded 168 articles. A search of Web of Knowledge without limits yielded 424 articles, and a search of the Cochrane Library yielded no additional articles. One author (R.E.W.) read the titles and abstracts of the articles and identified 82 articles that potentially satisfied our inclusion criteria: a) humans assigned randomly to desflurane or propofol groups without other differences between groups, e.g., induction drugs; b) mean and standard deviation reported for extubation time and/or time to follow commands; and c) peer-reviewed publication, i.e., exclusion of letters, editorials, and meeting abstracts. No restrictions were placed on date or language. The references of the articles were also searched in an attempt to identify additional articles, and none were found. Two OR endpoints were included because recovery times can be sensitive to the selection of the endpoint.7,8,35 Two authors (R.E.W., F.D.) independently reviewed the 82 articles and independently abstracted data from the 26 articles meeting the inclusion criteria, including covariates and measures of study quality (Table 1).36 Overall, 56 articles were excluded: 20 because neither endpoint was reported; 13 because the articles did not contain original data; 13 because the two groups were not matched (e.g., the desflurane group received nitrous oxide but the propofol group did not);1 seven because the articles did not report standard deviations or standard errors; and three because patients had not been randomized. There were two discrepancies in data extraction involving two of the remaining 26 articles. One discrepancy was an error by R.E.W. caught by F.D., and the other was an error by F.D. caught by R.E.W. For the first error, a weighted average was copied incorrectly from the preceding row, and for the second error, the author judged incorrectly that a target-controlled infusion had been used.
Table 1

Characteristics of studies listed in sequence of increasing observed effect of desflurane vs propofol

Reference

n Subjects

Propofol

n Subjects

Desflurane

Sequence

Generation

Remifentanil

Target Infusion

Titrated

BIS or AEP

LMAD

Year

Mean

Age (yr)

Mean

Weight (kg)

Mean

Duration (min)

7

30

30

   

Yes

 

2007

43

60

99

10

30

30

Yes

 

Yes

  

2009

56

72

91

11

20

20

Yes

Yes

   

2007

26

67

82

12

23

22

     

1992

34

75

49

13

25

25

 

Yes

   

2001

48

75

141

14

32

31

 

Yes

Yes

  

2001

18

59

75

15

40

40

 

Yes

 

Yes

 

2003

46

66

96

16

15

15

     

1991

37

74

31

17

13

15

     

1993

34

64

20

18

25

25

Yes

Yes

 

Yes

Yes

2006

42

79

51

19

35

35

 

Yes

Yes

  

2001

40

76

67

15

40

40

 

Yes

   

2003

48

69

91

20

30

30

    

Yes

1998

44

65

25

21

23

23

     

1991

30

64

62

22

50

54

Yes

    

2002

35

66

32

23

14

16

     

1993

28

78

91

24

30

30

Yes

  

Yes

Yes

2002

75

74

48

25

14

14

     

1997

77

67

201

26

20

20

 

Yes

   

1998

36

67

63

27

35

40

Yes

  

Yes

Yes

2001

55

68

38

28

17

17

Yes

  

Yes

 

2001

30

70

79

29

29

31

Yes

    

1998

27

66

68

30

40

40

Yes

    

1998

29

71

71

31

18

18

 

Yes

Yes

  

2004

50

76

342

32

20

20

     

1996

24

79

 

33

11

12

   

Yes

 

2000

40

125

170

34

100

100

     

2007

52

72

67

n = sample size; LMAD = laryngeal mask airway device with all tracheal intubations with an endotracheal tube. Sequence Generation means that the patients were randomized to groups using either a random number table or a computer random number generator. Sequence allocation concealment (e.g., opaque envelopes) is not listed because it was reported only in one study.9 Remifentanil is use of remifentanil infusion for intraoperative analgesia. Target Infusion refers to computer-controlled infusion to deliver predicted constant plasma concentration or pre-specified declining dose per minute, uninfluenced by observed hemodynamics. Titrated BIS or AEP refers to dose adjustment to maintain bispectral index or auditory evoked potentials9 at pre-specified levels. Mean Duration refers to the duration of anesthesia or duration of surgery when duration of anesthesia was unavailable. Reference 15 appears twice because the article included studies with and without BIS

Percentage reductions in mean time and 95% confidence interval (CI) were calculated as described in the Appendix using Microsoft® Excel, Visual Basic for Applications.37 Percentage reductions in standard deviation and confidence interval were also calculated. The correlation between these two summary measures was studied, and the covariates were explored using Kendall’s rank correlation coefficient. Meta-regression was not used because the covariates that we expected to influence results (e.g., obese patients undergoing longer anesthetics would have larger differences) were not binary study characteristics but were measured variables with sampling error. The P values are two-sided and exact (StatXact® 9, Cytel Software Corporation, Cambridge, MA, USA). Fail-safe calculations assessed whether publication bias could have influenced conclusions.38

Economic interpretation of the meta-analysis results depends on the influence of time of emergence from general anesthesia on OR time. The Institutional Review Board at the University of Iowa approved observation of anesthesia providers at the ambulatory surgery centre. The times to prepare propofol for use in infusion syringe pumps were recorded by timing anesthesia providers as they drew up 50 mL of propofol and purged air from the attached extension tubing. Observational details and analyses of preparation times are described in the Appendix. In addition, activities of OR staff, including nurses, were observed from the time of end of surgery to tracheal extubation.

Results

There were few substantive differences in quality among the studies. None of the studies were blinded for desflurane vs propofol, and all studies were randomized. All patients received the drugs to which they were randomized (Table 1). Nine of the 26 studies reported randomization using either a random number table or a computer random number generator.

Desflurane reduced the variability (standard deviation) in time to extubation by approximately 26% relative to propofol, the variability in time to follow commands by 39%, the mean time to extubation by approximately 21%, and the mean time to follow commands by 23% (Tables 2-3, Figure). Heterogeneity among studies for each endpoint (P < 0.001) was unexplained by other measured variables (Table 4).
Table 2

Times from end of surgery to extubation and from end of surgery to follow commands for each study in Table 1

Reference

Propofol (min) Mean (SD)

Desflurane (min) Mean (SD)

Desflurane Reduction in Mean

Desflurane Reduction in SD

Extubation

Commands

Extubation

Commands

Extubation

Commands

Extubation

Commands

7

8.2 (3.0)

 

13.7 (5.0)

 

−68%

 

−67%

 

10

6.4 (4.2)

8.0 (0.8)

7.6 (0.7)

9.2 (0.7)

−14%

−15%

84%

14%

11

6.8 (3.7)

7.8 (3.7)

7.3 (3.4)

8.7 (3.3)

−8%

−12%

9%

11%

12

 

10.6 (6.3)

 

11.0 (5.5)

 

−4%

 

13%

13

5.5 (3.3)

 

5.7 (2.5)

 

−3%

 

25%

 

14

10.4 (3.0)

 

10.2 (5.1)

 

0%

 

−71%

 

15

6.8 (4.6)

 

6.5 (4.1)

 

5%

 

11%

 

16

 

10.0 (4.8)

 

9.4 (4.4)

 

4%

 

9%

17

 

5.3 (2.3)

 

5.0 (0.9)

 

6%

 

62%

18

6.9 (2.6)

6.6 (2.8)

6.4 (2.6)

6.0 (2.2)

6%

9%

0%

22%

19

6.3 (2.1)

8.7 (2.9)

5.3 (2.5)

7.3 (1.9)

15%

16%

−20%

35%

15

10.5 (5.9)

 

8.3 (6.1)

 

21%

 

−4%

 

20

5.6 (2.9)

6.6 (3.0)

4.4 (1.5)

5.1 (1.5)

22%

23%

49%

50%

21

 

8.3 (3.9)

 

6.4 (2.4)

 

23%

 

39%

22

 

4.6 (2.2)

 

3.5 (1.8)

 

23%

 

18%

23

 

12.2 (15.5)

 

9.1 (3.1)

 

26%

 

83%

24

8.7 (3.8)

10.5 (3.9)

6.1 (3.1)

7.7 (3.0)

29%

26%

18%

23%

25

9.9 (6.5)

14.3 (8.0)

6.9 (3.0)

7.4 (3.2)

32%

48%

55%

61%

26

9.8 (4.0)

10.6 (4.5)

6.7 (2.8)

7.2 (2.8)

31%

32%

30%

38%

27

 

6.0 (2.0)

 

4.0 (2.0)

 

32%

 

0%

28

 

8.0 (4.0)

 

5.0 (4.0)

 

35%

 

−3%

29

 

7.0 (6.0)

 

4.0 (2.0)

 

42%

 

67%

30

8.9 (5.3)

8.3 (6.9)

5.1 (3.3)

4.7 (2.6)

41%

42%

38%

63%

31

13.2 (2.3)

 

7.5 (1.3)

 

43%

 

43%

 

32

 

15.1 (1.8)

 

6.4 (0.4)

 

58%

 

78%

33

13.2 (7.6)

 

5.6 (1.4)

 

58%

 

82%

 

34

6.2 (3.2)

 

2.3 (1.6)

 

64%

 

49%

 

SD = standard deviation. Reduction in Mean refers to the reduction in the lognormal mean of time to extubation or time to follow commands, calculated using pivotal methods. Negative reductions indicate that values were larger in the desflurane group compared with the propofol group (see Appendix equations (1) and (10)). The corresponding standard errors of the estimates are not reported because they are in the log-scale. Reduction in SD refers to the reduction in the standard deviation of the time to extubation or time to follow commands, also calculated using pivotal methods (see Appendix equations (16) and (17)). The inverses of these standard errors are proportional to the areas of the circles in the Figure

Table 3

Desflurane reductions in operating room recovery times relative to propofol

 

All Studies

Excluding the studies with the largest and smallest reductions

Mean time to extubation

21% (95% CI, 4% to 36%; P = 0.010)

21% (95% CI, 9% to 32%; P = 0.001)

Mean time to follow commands

25% (95% CI, 5% to 41%; P = 0.008)

23% (95% CI, 16% to 30%; P < 0.001)

Standard deviation of time to extubation

30% (95% CI, 6% to 48%; P = 0.008)

26% (95% CI, 6% to 42%; P = 0.005)

Standard deviation of time to follow commands

40% (95% CI, 26% to 52%; P < 0.001)

39% (95% CI, 25% to 51%; P < 0.001)

Results are reported as mean with 95% confidence interval (CI). Fail-safe analyses showed that single additional studies with 0% difference would require > 3,000 patients per group for a resulting P > 0.0538

https://static-content.springer.com/image/art%3A10.1007%2Fs12630-011-9519-1/MediaObjects/12630_2011_9519_Fig1_HTML.gif
Figure

Reduction in variability in time to follow commands with desflurane instead of propofol. The value along the vertical axis is the reduction in the standard deviation of the time to follow commands by using desflurane instead of propofol, calculated using equations (11) to (17). The dotted horizontal red line at 39% is the weighted mean estimate reported in the Results and the right-hand column of Table 3. The solid horizontal red line shows 0% increase. Each circle shows the point estimates of the reductions in variability from a study as reported in Table 2. However, the relationship in Table 2 is less apparent because the table is sorted in ascending sequence of the percentage reduction in the mean time to extubation. The fact that 17 of the 19 studies are displayed above the solid horizontal 0% line highlights that the studies showed significant reductions in the variability of time to follow commands. The area of each circle is proportional to the precision of that estimate (i.e., 1 divided by the square of the standard error of the proportional reduction in standard deviation). Studies with greater precision appear as larger circles. As described below equation (10), the standard error is calculated by dividing the width of the 75% confidence interval by the corresponding inverse of the standard normal distribution. This graph is novel because previous studies did not estimate the standard error of the reduction in variability for each study in which desflurane was compared with sevoflurane and isoflurane. We previously estimated the standard error based on a pooled quantity from secondary observations of extubation times (see Appendix of reference 7). The graph is also novel because none of the prior studies reported a significant reduction in the standard deviation because the statistical methodology described in this article had not previously been developed. The Figure also shows the estimated reduction in the mean time to follow commands by using desflurane instead of propofol, plotted along the horizontal axis. The standard error of that estimate is not shown, as the focus of the plot is the reduction in variability along the vertical axis. The methodologically important finding of the Figure is highlighted by the line of equality. For several studies, the percentage reductions in the variability in the time to follow commands exceeded the reductions in the mean time to follow commands. Thus, there are unequal coefficients of variation between treatment groups, which differs from Fig. 5 of reference 7 for time to extubation with desflurane vs sevoflurane. For statistical details, see the Appendix after equation (9)

Table 4

Association between independent variables in Tables 1 and 2 and percentage reductions with desflurane (Table 2)

 

Mean Extubation

Mean Commands

SD Extubation

SD Commands

Kendall’s Correlation

Uncorrected P value

Kendall’s Correlation

Uncorrected P value

Kendall’s Correlation

Uncorrected P value

Kendall’s Correlation

Uncorrected P value

Variables of Table1from left to right

Total sample size

−0.07

0.74

0.02

0.92

−0.25

0.18

−0.06

0.75

Sequence generation

−0.18

0.44

0.03

0.90

0.04

0.88

−0.32

0.11

Remifentanil

−0.24

0.28

−0.24

0.26

−0.48

0.02

−0.14

0.53

Target infusion

−0.14

0.55

−0.31

0.14

−0.07

0.78

−0.10

0.66

Titrated BIS or AEP

−0.09

0.72

0.12

0.60

−0.16

0.51

−0.39

0.05

LMAD

0.03

0.95

−0.02

0.96

−0.03

0.95

−0.16

0.47

Year

−0.34

0.07

−0.18

0.33

−0.15

0.44

−0.33

0.06

Mean age (yr)

0.12

0.54

−0.13

0.45

0.34

0.06

−0.24

0.16

Mean weight (kg)

0.25

0.18

0.03

0.89

0.21

0.27

−0.03

0.89

Mean duration (min)

−0.06

0.78

0.22

0.23

0.13

0.49

0.14

0.45

Variable of Table2columns 2 and 3

Propofol value (min)

0.30

0.11

0.21

0.24

0.30

0.10

0.20

0.25

Variables from Table 1 - see Table 1 for definitions. Propofol value (min) - on the far left, it refers to the mean time to extubation for propofol, and on the far right, it refers to the standard deviation of the time to follow commands. To interpret the sign of Kendall’s rank correlation coefficient for the binary variables (e.g., use of remifentanil), absence was coded as 0 and presence as 1. Given 40 comparisons, Bonferroni correction of P = 0.05 would treat P ≤ 0.001 as statistically significant. SD = standard deviation. BIS = bispectral index; AEP = auditory evoked potentials; LMAD = laryngeal mask airway device

We observed seven cases in which a propofol anesthetic was used. In all cases, at least one OR nurse or surgical technologist was performing no discernable activity for at least 100 sec prior to tracheal extubation (95% CI > 66% of cases). The time to draw up propofol and set up an infusion pump averaged one minute (see Appendix).

Discussion

Desflurane proportionally reduces the mean time to extubation and time to follow commands relative to propofol (21% and 23%), approximately the same as sevoflurane (25% and 19%)7 but less than isoflurane (34% and 34%).8 Clinicians provide anesthesia care in heterogeneous ways (Table 1) and meta-analysis of economic endpoints provides managerial insight into overall (pooled) effect (Table 3).39 The principal limitation is that since drug (treatment) effect is proportional,7 for results to be useful economically to a facility, results need to be converted to absolute reductions in time using the facility’s patients’ typical OR recovery times. For example, a 20% reduction in the mean time represents 1.5 min for patients with the brief mean interval of 7.5 min vs 2.5 min for patients with the long mean interval of 12.5 min.7 Differences between anesthetic agents in OR recovery times are studied since they can limit OR throughput, based on non-anesthesia OR personnel waiting for the patient to be extubated during emergence for most (> 66%) cases. Outside of ORs there typically are additional personnel (e.g., housekeepers and post-anesthesia care unit nurses) waiting for the end of cases, since surgical suites appropriately staff for multiple ORs in which cases end simultaneously.40,41 Additional personnel (e.g., housekeepers and postanesthesia care unit nurses) are typically outside of ORs waiting for cases to end, since surgical suites are appropriately staffed for multiple ORs on the basis of cases that end simultaneously.40,41

Achievable reductions in direct OR costs resulting from time savings in the OR can be calculated as described in the Discussion of reference 7. Specific values are unique to each facility (e.g., application of our results depends on the number of ORs with more than eight hours of cases daily). Other endpoints, such as time to home discharge and nausea, have previously undergone meta-analysis42-44 and are also of value when comparing the overall impact of the selection of anesthetic drugs. Selection of propofol adds approximately one minute to fill a syringe for infusion and to set up the infusion pump (see Appendix).

The novel findings of our study are twofold. First, as shown in the Figure, the reductions in the variabilities in OR recovery time are larger when desflurane is compared with propofol (26%, time to extubation and 39%, time to follow commands) than when desflurane is compared with sevoflurane (21% and 22%, respectively).7 Second, as is the focus of the Appendix, the reductions in the variabilities relative to propofol (26% and 39%) are larger than the corresponding mean reductions (21% and 23%). Such results are striking when considered in light of the traditional weighted mean difference meta-analysis that assumes a 0% reduction in variability. The pharmacokinetic/dynamic basis for the difference between reductions in standard deviation and mean is unknown. Variability matters clinically, as it contributes to the incidence of prolonged extubation times (e.g., > 15 min). Anesthesiologists rate recovery from propofol as poor when such prolonged extubations occur.45 Resulting intangible OR costs include significantly longer times to incision of to-follow cases7 (e.g., from surgeons leaving surgical suite46). The methods described in the Appendix and summarized in the Figure can be used in future clinical trials and meta-analyses of such trials with the reduction in variability of task duration as a primary study endpoint.

In conclusion, the mean reduction in OR recovery times for desflurane relative to propofol was comparable with that shown previously for desflurane relative to sevoflurane. The reduction in variability with propofol exceeded that compared with sevoflurane. Facilities can use the percentage differences when making evidence-based pharmacoeconomic decisions.

Footnotes
1

One article was unclear about whether the desflurane and propofol groups had both received nitrous oxide. An e-mail to the author clarified the protocol.

 

Acknowledgements

We appreciate the assistance of Emine Bayman PhD who used WinBUGS code to perform another analysis of the standard deviations. We thank Martin Mueller MD for reviewing the articles written in German.

Competing interests

This research was supported by Baxter Healthcare Corporation. Baxter Healthcare Corporations’ physicians and scientists made recommendations about the study design prior to providing funding, and they reviewed the manuscript once written. They were not involved in the conduct of the study; collection, analysis, or interpretation of the data; or preparation of the manuscript. Drs. Dexter and Epstein have previously performed research funded by Abbott Laboratories, and they previously performed research funded by Baxter Healthcare Corporation. Drs. Wachtel and Dexter receive no funds personally, other than their salaries from the University of Iowa. They do not receive travel expenses or honoraria from any source other than the University, and they have tenure with no incentive program.

Copyright information

© Canadian Anesthesiologists' Society 2011