Sedation in the Intensive Care Unit:  A Systematic Review FREE

To minimize patient discomfort in the intensive care unit (ICU), sedation has become an integral part of critical care practice. Sedation reduces the stress response, provides anxiolysis, improves tolerance of ventilatory support, and facilitates nursing care.^1- 3 Unfortunately, sedatives have adverse effects, have the potential to prolong mechanical ventilation, and may increase health care costs. An ideal sedative agent would have rapid onset of action, be effective at providing adequate sedation, allow rapid recovery after discontinuation, be easy to administer, lack drug accumulation, have few adverse effects, interact minimally with other drugs, and be inexpensive.

The lack of a recognized ideal sedative has resulted in varied approaches to sedation both among and within intensive care units (ICUs). In a survey of 164 ICUs in the United States,⁴ 18 different sedative agents were used. The most common agents listed were morphine sulfate, lorazepam, midazolam, diazepam, and haloperidol. Intensive care unit consultants in the United Kingdom reported use of 11 different agents in another survey.⁵ Similarly, Dasta et al⁶ reported the use of 23 different drugs for sedation, anxiety, and pain relief in their surgical ICU. These surveys indicate wide practice variation in sedative administration for critically ill patients.

A number of narrative review articles have summarized the principles of sedation and the range of available agents.^1,3,7- 16 In 1992, the Society of Critical Care Medicine (SCCM) Task Force developed practice parameters for sedation in the ICU based on available scientific data, clinical expertise, and experience.¹⁷ After careful review and discussion, this group recommended the use of midazolam or propofol for short-term (≤24 hours) sedation, lorazepam for longer-term (>24 hours) sedation, and haloperidol for delirium. The level of evidence available from the literature for these reviewers was somewhat limited at the time, and these useful clinical recommendations required heavy reliance on expertise. A survey by Rhoney and Murry¹⁸ presented at the SCCM's meeting in 1998 found that most ICUs did not use protocols, and practice often did not adhere to the society's recommendations.

There have been no recent systematic reviews on sedation for critically ill patients. While adequacy of sedation is an important outcome, additional outcomes of relevance are time to extubation and length of ICU stay. The objective of this systematic review was to answer the question: In ventilated ICU patients, which sedatives are associated with best level of sedation, shortest time to extubation, and shortest length of ICU stay? We divided the intervention into short-term and longer-term sedation, as suggested by the SCCM guidelines.¹⁷ We also have recorded the hemodynamic effects of sedative agents when available. We divided agents using the following taxonomy: (1) benzodiazepines (eg, diazepam, midazolam, lorazepam), (2) opiates (eg, morphine, fentanyl), (3) neuroleptics (haloperidol, methotrimeprazine), and (4) anesthetic agents (propofol and inhalational agents such as isoflurane).

We searched MEDLINE, EMBASE, and the Cochrane Collaboration for articles from 1980 to June 1998, using the key words hypnotics and sedatives, propofol, midazolam, benzodiazepines, haloperidol, and antipsychotic agent, each combined with intensive care unit and critical care. We wrote to experts and first authors of selected articles as well as 18 pharmaceutical companies. We hand searched Anesthesiology, Anesthesia and Analgesia, Canadian Journal of Anaesthesia, British Journal of Anaesthesia, Anaesthesia and Intensive Care, and Anaesthesia from 1980 to June 1998. Reference lists of retrieved articles were reviewed, and personal files were searched.

To identify published studies for inclusion in this analysis, study populations included mechanically ventilated adult patients requiring short-term or longer-term sedation; interventions included a comparison of at least 2 sedative drugs; outcomes showed quality of sedation, time to extubation, or length of stay in the ICU; and study designs were randomized trials. We excluded studies available in abstract format only. Our study population did not include those undergoing withdrawal of life support.

Duplicate data abstraction was conducted by 2 of 3 investigators (M.E.O., R.A.S., S.P.K.). In reporting information on instruments used to evaluate quality of sedation, we used a taxonomy recently proposed for this purpose by De Jonghe and colleagues.¹⁹ The selected publications were critically appraised using 8 validity criteria (Table 1). Differences of opinion were settled by consensus after consultation with a third investigator. We used criteria applicable to most systematic reviews of randomized controlled trials (criteria 1-4) in addition to criteria of specific importance in studies of sedative agents (criteria 5-8).

We analyzed the study results separately for short- and longer-term sedation, cardiac surgery patients and other ICU patients, and by agents being compared (Table 2; an extended version of the table with additional information is also available). The studies were too clinically heterogeneous to permit statistical pooling.

We identified 49 randomized controlled trials,^20- 68 41 from personal file, MEDLINE, and EMBASE searches, 7 more from hand searching anesthesia journals, and 1 from an expert in the field. Of the 49 trials, 17 were excluded^52- 68 (4 addressed a different population^52- 56 [2 normal volunteers, 1 nonventilated ICU patients, 1 intraoperative cardiac surgery anesthesia patients], 2 evaluated different outcomes^57- 58 [renal function of isoflurane, urine, and plasma catecholamines], and 6 investigated different dosing schedules or modes of administering a single agent^59- 68), leaving 32 trials for analysis.^20- 51

Short-term sedation was addressed in 20 studies,^{20- 34,41- 42,44- 45,50} whereas 14 studies included patients receiving sedation for more than 24 hours^{32- 33,35- 40,43,46- 49,51} (Table 2). Populations included cardiac surgery^20- 28,44 other surgical ICU,^{30- 31,34,45,47} trauma,^39,46 medical ICU,^35,40 and mixed ICU patients.^{29,32- 33,36- 38,41- 43,48- 49,51} The relative effectiveness of the anesthetic agent propofol, and the benzodiazepine midazolam, were compared in 20 of these 32 clinical trials.^20- 39 Other agents included the benzodiazepine lorazepam^40- 41; the opiates alfentanil,^44,51 pethidine,⁴⁴ papaveretum,⁴⁹ fentanyl,^46- 47,50 morphine,^48,51 and lytic solution⁴⁵ (pethidine, promethazine, and dihydroergotamine); and the anesthetic agents isoflurane^42- 43 and ketamine.^46,50

A variety of scoring systems were used to assess quality of sedation in 30 of the 32 trials (Table 2). The Ramsay score⁶⁹ was used most frequently, in 18 trials (60%). This instrument consists of 1 item and is evaluated using a numerical scale of 1 to 6. Other scoring systems included 2 (6%) modified Ramsay scores (1 item, numerical scale), 3 (10%) Cook and Palma modified Glasgow Coma Scales⁷⁰ (4 items, numerical scale), and 9 (30%) of 30 scoring systems created for each specific study (variable number of items, numerical scales). One trial used both the Ramsay and the Cook and Palma score,³² and another used a scoring system that was derived from merging the Ramsay and Cook scores (however, the scoring system was not provided in the article).³³

The validity of included trials is summarized in Table 1. Masking of allocation to treatment was not documented in any trial. Blinding health care workers was conducted in 5 (16%) of 32 trials. Standardized cointerventions that could affect the outcomes of time to extubation and length of ventilation or ICU stay were variably reported (weaning strategy, 3 [9%] of 32 trials; anesthesia for postoperative patients, 7 [78%] of 9; use of analgesia, 7 [22%] of 32; and neuromuscular blockers, 13 [41%] of 32 trials). Intention-to-treat analysis was used in 26 (81%) of 32 trials.

Cardiac Surgery Patients. Nine trials were identified that examined the relative effectiveness of propofol to midazolam in post–cardiac surgery patients.^20- 28 One trial compared the use of 2 different opiate preparations⁴⁴ and found no difference in quality of sedation or time to extubation when either pethidine or alfentanil was used.

Of the 7 trials^{20- 22,25- 28} providing data comparing quality of sedation of propofol vs midazolam, 5 reported no difference,^22,25- 28 and 2 found propofol to be more effective than midazolam.^20- 21 Trials reporting a difference between propofol and midazolam were older and fulfilled fewer validity criteria. Time to extubation was shorter for patients taking propofol than midazolam in 5 of 8 studies reporting this outcome^{20- 22,24,28} but was no different in the remaining trials.^23,25,27 Six of 7 trials reported no difference in duration of ventilation between propofol and midazolam,^21- 25,27 whereas 1 trial favored propofol rather than midazolam,²⁰ and the 2 trials reporting length of ICU stay found no difference.^27- 28 One trial reported a statistically significant increase in incidence of hypotension in the propofol vs the midazolam arm.²⁶

Surgical or Mixed ICU Patients. Nine trials included patients from surgical or mixed ICUs who received sedation for less than 24 hours.^{29- 34,41- 42,45} Of these, 6 compared the effectiveness of propofol vs midazolam,^29- 34 1 compared midazolam to lorazepam,⁴¹ 1 compared midazolam to isoflurane,⁴² and 1 compared propofol to lytic solution⁴⁵ (pethidine, promethazine, and dihydroergotamine).

Three of the 6 trials found propofol to be more effective than midazolam in the quality of sedation achieved,^31- 33 with the remainder reporting no difference.^29- 30,34 All 3 trials that reported time to extubation found propofol to be more effective than midazolam.^29,32- 33 No trials comparing propofol with midazolam reported on duration of ventilation or ICU stay. While only 1 trial found significantly lower blood pressure in the propofol arm compared with midazolam (more hypotension in the first 30 minutes),³⁴ 2 other trials noted a greater but nonsignificant incidence of hypotension with propofol compared with midazolam.^29,31

The trial comparing midazolam with lorazepam did not find any difference in quality of sedation but did not provide corresponding data.⁴¹ This trial also did not report a difference in hemodynamic effects, and no other outcomes were noted. The trial comparing isoflurane with midazolam reported a superior quality of sedation and shorter time to extubation for isoflurane with no difference in effect on hemodynamics.⁴² Propofol and lytic solution provided a similar quality of sedation and effect on hemodynamics; however, propofol resulted in a faster time to extubation.⁴⁵ Finally, the combination of ketamine and midazolam in patients requiring exogenous catecholamines for blood pressure support resulted in lower doses of these inotropes or vasopressors than those required by patients receiving fentanyl and midazolam.⁵⁰

Of the 14 trials evaluating sedation for more than 24 hours, 7 compared propofol with midazolam,^{32- 33,35- 39} 1 compared midazolam with lorazepam,⁴⁰ 1 compared midazolam with isoflurane,⁴³ 1 compared ketamine with fentanyl,⁴⁶ and 4 studied a combination of agents.^47- 49,51 All studies examined surgical or mixed ICU patients.

The quality of sedation for propofol and midazolam was similar in 3 of the 6 trials reporting this outcome.^35,37,39 One trial favored midazolam rather than propofol,³⁸ and 2 favored propofol rather than midazolam^32- 33; however, statistical analysis was provided for only 1 of these trials.³² Time to extubation was shorter in the propofol group compared with the midazolam group in all 4 trials reporting it,^{32- 33,37- 38} although 1 trial did not report the statistical analysis³³ and a second did not find the difference statistically significant.³⁸ No trials of midazolam vs propofol reported duration of ventilation, and the single trial reporting length of ICU stay found no difference between propofol and midazolam.³⁹ One of the 4 trials reporting hemodynamic effects found more hypotension in the propofol group compared with the midazolam group.³⁸

The trial comparing midazolam with lorazepam found no difference in quality of sedation or hemodynamic effects.⁴⁰ Isoflurane provided a shorter time to extubation than midazolam but no difference in quality of sedation, hemodynamics, or duration of ICU stay.⁴³ Ketamine resulted in higher blood pressure and heart rate than fentanyl in patients with head injuries.⁴⁶

Of the 4 trials examining different combinations of sedatives, the only trial reporting a significant difference in outcome was the combination of alfentanil and propofol, which, compared with morphine and midazolam, provided better quality of sedation, shorter time to extubation, and decreased length of ICU stay with no difference in hemodynamic effects.⁵¹

The most striking finding in this review was the relatively small portion of sedative agents reported as being used in surveys of both North American and European practice^4- 6 that have been evaluated rigorously by more than 1 or 2 randomized controlled trials. A large number of trials have been published comparing midazolam with propofol in several patient groups, while few studies have been conducted using other common agents. Propofol is at least as effective as midazolam in providing desired levels of sedation. In addition, once the decision has been made to wean patients from the drug, propofol appears to consistently result in a faster time to extubation in patients receiving either short-term or long-term sedation. It is not clear, however, that this shorter time to extubation actually leads to a decrease in total time of ventilation or a shorter ICU stay. While there has been some suggestion that propofol may lead to lower cost in short-term sedation, the opposite may occur in patients requiring longer sedation. Finally, use of propofol has been consistently found to result in more problems with hypotension than does midazolam. How often this results in a clinically important adverse event is not clear. In summary, propofol appears to offer some advantage in the setting where rapid waking of the patient is desired, but more study is required to determine whether the increased cost and potential to cause hypotension are outweighed by this benefit. It may be that the dosage of the drug is more important, or at least as important as the drug itself.

No difference was found in the effectiveness of midazolam compared with lorazepam in 2 trials, suggesting more trials be conducted that examine both the relative effectiveness of these 2 commonly used agents and their associated costs. Isoflurane was found to lead to a shorter time to extubation than did midazolam and better quality of sedation in patients who received it for less than 24 hours. Ketamine was as effective a sedative agent as fentanyl was in head injury patients and had a more favorable impact on blood pressure (less hypotension). Ketamine may also have had a role in patients requiring higher doses of inotropes or vasopressors. Both isoflurane and ketamine are used infrequently in the ICU setting but may warrant additional study. No other firm recommendations can be made from this review. The small number of trials comparing any 2 agents limits the interpretation of these studies.

Cost is becoming an increasingly important factor in deciding whether to adopt new therapies. We did not include reported cost evaluations available in a few of these trials^{28,32- 33,37- 38,51} because of the limited generalizability of cost data outside the study center. These concerns arose as a result of the relatively brief descriptions available on the costing methods. Recommendations on how to conduct economic evaluations that result in both internal validity (valid for the specific research center) and external validity (generalizability beyond the research center) have been developed and published.^71- 74 Future investigations of new and current sedative agents will need to incorporate valid assessments of both their relative effectiveness and associated costs.

Strengths of this review include the systematic approach to searching the literature, selecting the relevant studies, and independent duplicate assessment of trial validity. Our inability to fully assess the validity of the methods of unpublished studies resulted in their exclusion from this review. Heterogeneity of patients, drugs, and dosing regimens precluded meta-analysis of study results. As with all systematic reviews, readers are referred to the original publications for further detail.

The outcome of primary interest was the quality of sedation provided to the patient. While most trials assessed this end point using the Ramsay scale^{19- 26,28- 33,36,39,41- 42,44,48} (a 6-point scale, with measures ranging from anxious or agitated to asleep with no response⁶⁹), this commonly used sedation scale has never been rigorously validated.¹⁹ A recent systematic review ¹⁹ of sedation scoring systems confirms that some of the other scales used in these trials also lack formal validation. All scales in use are numerical, and most assess only a single item. For example, although the Ramsay scoring system is a 6-point numerical scale that includes measures of sedation and agitation, its format restricts these to being assessed as 1 item. While other scales, such as the one of Cook and Palma,⁷⁰ have a greater number of items, allowing for evaluation of an increased amount of clinical information, they have not been validated. These aspects of scoring systems lead to concern regarding the reproducibility and validity of results and difficulty in interpretation of results across studies. Additional difficulty arises when different definitions of ideal level of sedation are used; for example, some studies required Ramsay level 3, while others used levels 2 to 4, 3 to 5, 2 to 5, or 5. This review highlights the need for a reliable and valid sedation scoring system to improve the interpretability of future studies.

Blinding both patients and assessors to treatment is of greatest importance when subjective outcomes are used. In the trials reviewed, those assessing quality of sedation were frequently aware of the sedative agent the patients were receiving. Knowing which agent a patient is receiving can introduce the potential for bias in interpretation of the outcome of interest. While interpretation of quality of sedation was the most subjective outcome assessed, time to extubation and length of ICU sedation can similarly be influenced by knowing which sedative agent was received. It is possible that blinding of assessors had been performed in some studies and not reported; however, the importance of double-blinding has become so well recognized during the past decade that most investigators will report its use if it was done. All cointerventions, including anesthesia (if applicable), use of analgesics (other than those being directly studied) and neuromuscular blockers, and approach to weaning from mechanical ventilation, should be standardized or at least reported. These measures will help reduce bias in interpretation of end points, such as quality of sedation, time to extubation, and length of ventilation and ICU stay. These processes were not followed in many of the trials reviewed.

It is useful to consider how, with time, secular changes in practice can influence the interpretation of study results. For example, the approach to cardiac surgery patients has evolved during the past decade from a common practice of keeping patients intubated and sedated overnight to a more aggressive approach of allowing patients to awaken and be extubated as soon as possible, referred to as fast-tracking.⁷⁵ As a result of the potential for practice to change appreciably from the time a randomized controlled trial is conducted, cautious interpretation of the study is necessary and sound clinical judgment required in deciding how, or if, a study applies to current practice.

The lack of randomized trials for some of the sedative agents currently in use in ICUs raises the issue of considering observational studies to obtain the best evidence for such agents as haloperidol or diazepam. However, the potential for bias increases appreciably when studies using historical or nonrandomized controls are considered. Given the widespread use of sedation for ICU patients, there is a dearth of rigorous, adequately powered randomized controlled trials comparing commonly used agents in this setting. More economic evaluations are warranted in this field, given the diverse purchasing costs of different agents and their variable and incompletely evaluated effect of economic outcomes, such as duration of mechanical ventilation and duration of ICU stay.