| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
Divisions of *Surgical Intensive Care, Anaesthesiological Investigations, and Anaesthesiology, Department APSIC, University Hospitals of Geneva, Geneva, Switzerland
Address correspondence and reprint requests to B. Walder, MD, Division of Surgical Intensive Care, University of Geneva Hospitals, Rue Micheli-du-Crest 24, CH-1211 Genève 14, Switzerland. Address e-mail to Bernard.Walder{at}hcuge.ch
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Implications: The duration of adequate sedation time is longer with propofol compared with midazolam. In postoperative patients with sedation <36 h, weaning is faster with propofol. The clinical relevance of these differences remains unclear. Specific adverse effects are more common with propofol. For rational decision making, reliable data are needed on cost, mortality, intensive care unit populations other than postoperative, and long-term sedation.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Ideally, sedation in the ICU allows for a comfortable and cooperative patient, decreases the levels of anxiety and stress, reduces insomnia and the risk of awareness during stressful interventions, and normalizes metabolism and hemodynamics. The ideal hypnotic has favorable kinetics that enable rapid onset and easy targeting of sedation and quick offset from sedation. This hypnotic would also have an acceptable adverse effect profile. Under these conditions, the hypnotic is expected to shorten an ICU stay and thus decrease cost, reduce morbidity, and even mortality.
Propofol and midazolam are IV hypnotics that modulate the 
-aminobutyric acid type A receptor (6,7). Favorable characteristics of propofol are the lack of accumulation and a short recovery time, both of which are essential for early neurological examination after discontinuation (2). Adverse effects of propofol are pain on injection (8), arterial hypotension (9), bradycardia (10), bloodstream infection (11), change in serum lipids (12), and excitation of the central nervous system (13), including seizures in susceptible patients (14). Midazolam is not painful on injection, and there is no increased risk of bradycardia, bloodstream infection, hypertriglyceridemia, or excitation of the central nervous system. However, drug accumulation may occur particularly in patients with renal failure (15), and in patients receiving antimycotics (16). This may lead to a prolonged weaning from mechanical ventilation.
More than 10 years ago, Aitkenhead et al. (17) published the first randomized controlled trial on the role of propofol and midazolam in mechanically ventilated ICU patients. Since, numerous similar reports have been published. The relative benefit and harm of propofol and midazolam in ICU patients, however, remain controversial (18,19). This may be related to differences in study populations, varying trial sizes, inconsistent evaluation of the level of sedation, and variability in end points among trials. The aim of this systematic review was to synthesize the best currently available evidence on any difference between propofol and midazolam in mechanically ventilated, critically ill patients and to quantify these differences.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Search Strategy, Inclusion, and Exclusion Criteria
A systematic search was performed for full reports, published in peer-reviewed journals, of randomized comparisons of propofol with midazolam in critically ill, mechanically ventilated adult patients in the ICU. We searched MEDLINE (PubMed and KnowledgeFinder® 4.19, from 1966, date of last search June 15, 1999), EXCERPTA MEDICA (from 1984, date of last search January 6, 1999), and the COCHRANE LIBRARY (1999, issue I). Searches were without language restriction, using the free text and MeSH terms propofol, midazolam, critically ill, critical care, intensive care, random, and combinations of these words. References lists of retrieved reports and review articles (21,22) were checked. Data from abstracts from scientific meetings, animal studies, and review articles were not considered. We did not contact the manufacturers of propofol and midazolam. Authors were contacted when there was ambiguity about the data.
Scoring and Data Abstraction
Abstracts of retrieved reports were screened by one author (BW). All articles that did not clearly meet our inclusion criteria (e.g., randomized comparisons of propofol with midazolam in critically ill, mechanically ventilated adult patients in the ICU) were excluded at this stage. All authors independently read all remaining trials and assessed their methodological validity using the 5-point, 3-item Oxford scale (23). The scale takes into account proper randomization, double-blinding, and reporting of withdrawals. The minimum score of an included randomized controlled trial is 1, the maximum score is 5. Consensus was reached by discussion.
Data abstraction of all included reports was done by two investigators independently (BW and NE) and checked by the others. Information was extracted on the methodological validity of the trials (see "Study End Points"), inclusion and exclusion criteria of patients, number of included and analyzed patients, patient characteristics (age, clinical setting, pathologies), indication for sedation, baseline level of sedation (i.e., before the start of the treatment), definition of target sedation, duration of sedation, regimens of propofol and midazolam, concomitant analgesia, and sponsorship.
Study End Points
There was a pre hoc decision that the following end points were of primary interest: efficacy of sedation, weaning time from mechanical ventilation, adverse drug reactions, length of ICU stay, mortality, and cost estimations (drug costs, ICU costs).
Efficacy of sedation (in %) for a trial was defined as the average duration of adequate sedation (in hours) divided by the average total duration of sedation (in hours) in that trial. The definition of adequate sedation was taken from the original trials, independent of the sedation scale used. Weaning time was defined as the delay from the end of the drug administration to extubation. Data on adverse drug reactions were extracted when they were reported in dichotomous form. A clinically relevant hypertriglyceridemia was defined as at least a doubling of the normal values as reported in the original trials. A clinically relevant arterial hypotension was defined as a systolic blood pressure < 90 mm Hg or the need for an intervention.
Analysis
Meta-analyses were performed by using weighted mean difference (WMD) for continuous data, relative risk (RR) (24) and number-needed-to-treat (NNT) for dichotomous data with fixed and random effects models (Revman version 4.0; Cochrane Library, Oxford, England). For combined data, a fixed effect model (25) was used when data were homogeneous (P > 0.1); otherwise, we used a random effects model (26). It was assumed that a 95% confidence interval (CI) of the WMD which did not include 0, and a 95% CI of the RR which did not include 1 indicated a statistically significant difference between propofol and midazolam. To estimate the clinical relevance of any difference between propofol and midazolam, we calculated the NNT with 95% CI for dichotomous data (27), using the weighted means (weighted by trial size) of the pooled propofol and midazolam event rates (28). A positive NNT indicates how many patients have to be treated with propofol for one to show a defined end point, who would not have done so had they been treated with midazolam.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
|
The median Oxford validity score of all reports was 2 (range, 1 to 4) (Table 1). One report scored 4, 5 scored 3, 9 scored 2, and 10 scored 1. In one trial (54), there was an effort to blind the study drugs.
|
36 h) and "long-term" studies (
54 h). In 16 trials (12,17,19,31,33,40,41,46,48–52,54–56), the depth of sedation was evaluated by using the Ramsay scale (57) (a 6-point scale ranging from 1 = anxious, agitated, restless, to 6 = no response to glabellar tap or loud auditory stimulus) (Table 1). In 7 trials (1,18,40,41,48,50,52), the level of sedation at baseline (i.e., before the start of intervention) was reported. All trials except one (45) defined a target sedation; when the Ramsay scale was used, a score of 3–4 was most frequently targeted.
The average number of patients per trial was 60 (range, 9 to 312). The average range of ages of patients per trial was 37 to 69 yr. Most patients were surgical, medical, or trauma. For 172 patients, no clinical setting could be identified. In the 13 trials with data on short-term sedation, two thirds of patients were surgical, and a minority only medical or trauma ( Table 2). Six of these trials included postoperative cardiac patients only (351 patients, 71% of all surgical patients with short-term sedation) (48–51,54,55). In the 8 trials with data on long-term sedation, patients were equally distributed among surgical, medical, and trauma (Table 2).
|
In 23 trials, the average maintenance doses of propofol and midazolam were mentioned; it was between 0.6 and 3 mg · kg-1 · h-1 for propofol, and was between 0.012 and 0.3 mg · kg-1 · h-1 for midazolam (Table 1). In all trials, concomitant analgesia was well controlled (i.e., trials with unequal analgesia consumption among groups (35,36) had been excluded from the analysis). In 16 trials, morphine was used as an analgesic; 10 trials reported the average dose (range, 0.005 to 0.046 mg · kg-1 · h-1). In 3 trials, no opioids were administrated, and in 7 trials opioids other than morphine were used (for instance, IV piritramid, papaveretum, fentanyl, alfentanil, sufentanil, or epidural fentanyl).
Analysis of Efficacy Data
An intention-to-treat analysis of efficacy data was not possible, because most trials reported the number of analyzed patients only, but not the number of enrolled patients.
Efficacy of Sedation
Efficacy of sedation was reported in 19 trials (17 reports). In those, efficacy of sedation was a minimum of 54% to a maximum of 97% with propofol and was a minimum of 26% to a maximum of 95% with midazolam ( Fig. 2A). The event scatter suggested that propofol performed better than midazolam. There was no intention to analyze these data meta-analytically. There was one outlier, a small study in which efficacy of sedation was only 26% with midazolam, compared with 54% with propofol (40). This was the only trial in which efficacy of sedation was analyzed during physiotherapy. When this study was excluded, efficacy of sedation was in the range of 60%–97% with propofol and was 47%–95% with midazolam.
|
Weaning Time
Weaning time was reported in 12 trials (663 patients) in 10 reports. In 9, sedation times were short (36 h); most patients were postoperative (17,19,45,48–51,54,55). The event rate scatter ( Fig. 3) suggested a relatively shorter weaning time with propofol in 6 trials (17,19,45,50,51,55), a shorter weaning time with midazolam in 1 (49), and equivalence in 1 (54). The average weaning times per trial after propofol sedation were 5 to 258 min (0.8 to 4.3 h) and after midazolam were 92 to 432 min (1.5 to 7.2 h); for combined data, the WMD was 134 min (95% CI, 46–221 min), or 2.2 h (95% CI, 0.8–3.7 h) in favor of propofol.
|
54 h). In one study with 108 patients (12), after an average 141 h of sedation, average weaning time with propofol was 35 min (0.58 h), and after midazolam was 98 min (1.6 h); the WMD was 63 min (95% CI; 47–80 min) or 1 h (95% CI; 0.8–1.3 h). Two small trials came from the same report (19). In the first, the average duration of sedation in 28 patients was 115 h; with propofol, weaning time was 24 min (0.4 h) and with midazolam was 810 min (13.5 h); the WMD was 786 min (95% CI; 650–922 min) or 13.1 h (95% CI; 10.8–15.4 h) in favor of propofol. In the second, the average duration of sedation in 20 patients was 339 h; with propofol, weaning time was 48 min (0.8 h) and with midazolam was 2196 min (36.6 h); the WMD was 2148 min (95% CI, 1839–2457 min), or 35.8 h (95% CI, 30.6–41 h) in favor of propofol. In both trials, target sedation was defined as a Ramsay score 2–5.
Other End Points
The length of ICU stay was reported in four trials (18,46,53,54). No meaningful conclusions could be drawn. Costs were estimated in five reports; authors of three concluded that propofol was more cost effective (12,19,45), and authors of two concluded that midazolam was more cost effective (18,54).
Authors of 15 reports favored propofol for sedation in ICU patients (1,12,17,19,31,39,40,44–47,50,51,55,56). One trial was in favor of midazolam (18). The authors of eight reports did not express any opinion.
Ten trials were sponsored by the manufacturers of propofol or midazolam. Of nine trials supported by the manufacturer of propofol, five reported a benefit for propofol (1,17,40,50,51). Conclusions of the only trial which was supported by the manufacturer of midazolam were in favor of midazolam (18). Six trials (138 patients) were published in a supplement of a peer-reviewed journal which was sponsored by the manufacturer of propofol (39,41,44,46,47,56).
Adverse Drug Reactions
Specific adverse drug reactions (arterial hypotension, hypertriglyceridemia) were significantly more often reported with propofol compared with midazolam (Table 3). No trial reported on bloodstream infection in relation to the administration of the study drugs. Data on mortality were reported in 8 trials (1,12,17,18,33,42,46,53). Two further trials reported on death but without group assignment, and original authors were unable to provide more precise data (19,40). Of patients receiving propofol, 32 of 331 (9.7%) died; of patients receiving midazolam, 40 of 334 (12.0%) died; the RR was 0.8 (95% CI, 0.5 to 1.3).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Further, there was strong evidence that weaning times were shorter with propofol after short-term (36 hours) sedation. Again, the clinical relevance of this result was not obvious; the difference in the average weaning time, although statistically significant, was approximately 2 hours. With both hypnotics, there was a large variability in weaning times across trials (Fig. 3). In view of different weaning protocols and extubation criteria used in different institutions, this variability was not unexpected. However, we have to assume that, within one trial, the same methods of weaning and identical criteria of extubation were applied to all randomized patients. Thus, the relative difference in weaning times between propofol and midazolam should not be influenced by local techniques and methods. Interestingly, the range of times needed to wean all patients from mechanical ventilation (i.e., the longest average value minus the shortest average value) was 253 minutes (4.2 hours) with propofol and was 340 minutes (5.7 hours) with midazolam Figure 3). Thus, after short-term sedation with propofol, weaning takes place early (earliest after 5 minutes) in one trial, but more than 4 hours later in another. After short-term sedation with midazolam, extubation is unlikely to happen earlier than 1.5 hours or later than 6 hours after stopping the drug. Thus, although the earliest patients may be extubated faster after short-term sedation with propofol, the range of time spent to get all patients extubated is not much different from midazolam.
Data on long-term sedation ( 54 hours) were inconclusive, mainly because of two small trials that reported very short weaning times after propofol sedation and extremely long weaning times after midazolam sedation (19). Interpretation of the results of these trials is difficult because a wide variation of definitions of target sedation was used [i.e., a Ramsay score (57) of 2 = "fully awake" to 5 = "deeply sedated"].
Against these data on efficacy of sedation and weaning time, data on adverse drug reactions have to be weighted. One of 12 patients receiving propofol is likely to have an episode of arterial hypotension, who would not have been hypotensive had all patients received midazolam (Table 3). Supplementary fluid, administration of a vasopressor, decrease of the infusion rate of the sedative drug, or elevation of the lower extremities has been used as corrective measurement in the original trials. In clinical practice, the workload for ICU staff could be larger with propofol compared with midazolam, because additional interventions are needed to correct arterial hypotension. The clinical impact of propofol-induced hypertriglyceridemia is unknown. It is likely that the risk will be decreased with the recently introduced propofol 2% solution compared with propofol 1% used in the analyzed studies.
Data on the length of ICU stay, costs, and mortality were rarely reported in these trials. However, these end points are important for rational decision making. Fourteen years after the publication of the first randomized controlled trial on the subject (48), it is still unclear if propofol shortens the length of ICU stay in mechanically ventilated patients, or if it decreases cost compared with midazolam. Every day, thousands of mechanically ventilated ICU patients need sedation; thus, cost-effectiveness analysis is an important issue. Such analyzes, however, have to consider numerous factors, including increased workload caused by frequent but minor averse drug reactions (arterial hypotension, bradycardia) or rare but serious adverse drug reactions (bacteremia), direct drug costs, length of stay in the ICU, or different underlying diseases. As these factors were only inconsistently reported in the original studies; no sensible conclusions could be drawn.
An important limitation of the clinical applicability of the results of this meta-analysis was the selection of patients in the original reports. Most patients were sedated for shorter than 36 hours. In these trials, two thirds of patients were surgical, mostly cardiac. In trials on long-term sedation, only one third of patients were surgical. Thus, because most data are from short-term sedation in postoperative patients, they may not be applicable to long-term sedation in other patient groups. Another limitation was a result of the exclusion of high-risk patients in the original reports. In many trials, patients with liver or renal failure, or obesity were not included. Thus, strictly speaking, these data only apply to low-risk, critically ill patients. Further research should be directed to high-risk, long-term sedated ICU patients including ICU stay, costs, and mortality.
The methodological design of the trials and the quality of data reporting was often poor. For instance, the median Oxford validity scale was low, 6 of 25 reports scored only 3 or 4. In only one trial (54), there was an effort to blind the study drugs, leaving all other trials open to the potential risk of observer bias. However, blinding of these drugs is not simple because of their obvious difference in appearance; a double-dummy design would be needed.
In only 11 trials, the number of and reasons for withdrawals were mentioned. There is also the problem of trial size. Small trials can suffer significantly from the random play of chance (60). Of the 27 trials available for analysis, only 11 had more than 50 analyzable patients, and only 5 more than 100. Small size overestimates treatment effects in other settings (61); similar effects cannot be eliminated here. Finally, only 7 trials reported a baseline measurement of sedation. It is obvious that a baseline measurement is needed to ensure comparability of study groups. Of further concern was that 10 trials were supported by the manufacturer of the two drugs, and 6 were published in a sponsored supplement of a peer-reviewed journal. It cannot be eliminated that sponsorship had an impact on the quality of data reporting.
Effective and adequate sedation in critically ill patients undergoing mechanical ventilation is possible with both propofol and midazolam. In postoperative patients with a small incidence of concomitant organ diseases, who are undergoing sedation for no longer than 36 h, adequate sedation time is longer, and weaning time is shorter with propofol compared with midazolam. The clinical relevance of these differences remains unclear. Specific adverse effects happen more often with propofol. For rational decision making, more valid data are needed on sedation in high-risk populations (for instance, with multiple organ dysfunction syndrome), on long-term sedation (>48 h), on cost, ICU stay, and mortality.
|
Acknowledgments |
|---|
We thank Daniel Haake from the Documentation Service of the Swiss Academy of Medical Sciences for his help in searching electronic databases.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. N. Sessler and K. Varney Patient-Focused Sedation and Analgesia in the ICU Chest, February 1, 2008; 133(2): 552 - 565. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Arroliga, F. Frutos-Vivar, J. Hall, A. Esteban, C. Apezteguia, L. Soto, A. Anzueto, and for the International Mechanical Ventilation Study Use of Sedatives and Neuromuscular Blockers in a Cohort of Patients Receiving Mechanical Ventilation Chest, August 1, 2005; 128(2): 496 - 506. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Saito, Y. Terao, M. Fukusaki, T. Makita, O. Shibata, and K. Sumikawa Sequential Use of Midazolam and Propofol for Long-Term Sedation in Postoperative Mechanically Ventilated Patients Anesth. Analg., March 1, 2003; 96(3): 834 - 838. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
