1Introduction

Organization

We begin this report with background information regarding our perspective on the issues of weaning patients from mechanical ventilation. We then describe the methodology related to our definition of the study questions, our inclusion and exclusion criteria, our search, and our strategies for organizing and synthesizing our findings. Because the methodology of the individual studies and the types of questions they address are linked, we have organized the Results chapter of our report largely according to the types of studies. We address, in order, randomized controlled trials (RCTs) and nonrandomized controlled trials of weaning interventions, observational studies addressing predictors of weaning success, qualitative studies of the experience of health providers and patients involved in the weaning process, and studies using quantitative approaches to examine patient experience. Because the implications for future research are specific to each of these areas and within the controlled trials are specific to the interventions, we present our thoughts about future research as part of those sections of the conclusion. We deal with the strengths and limitations of our systematic review as the penultimate section of Chapter 4, Conclusions and Future Research. We end our report with a distillation of our most important inferences and thoughts about directions for future research in this field.

Mechanical Ventilation and Weaning from Mechanical Ventilation

Mechanical ventilation refers to the use of life support technology to perform the work of breathing for patients who are unable to breathe effectively on their own. Patients requiring mechanical ventilation include: (1) critically ill patients with advanced and potentially reversible respiratory failure due to pulmonary or non-pulmonary processes, (2) patients who are only temporarily unable to ventilate adequately on their own following general anesthesia, and (3) patients who have chronic respiratory or neuromuscular disorders that may prevent them from breathing effectively without mechanical support. Our report will address patients in all three groups.

Weaning from mechanical ventilation generally refers to the progressive reduction in mechanical support that is delivered to patients in the first two groups as they progressively increase their own contributions to breathing. Therefore, the settings relevant to this proposal are intensive care unit (ICUs), intermediate care units, stepdown units, postanesthetic recovery units, and other facilities in which the goal remains to discontinue mechanical support. Patients who are unable to be successfully liberated from mechanical ventilation in acute care settings and who are looked after in chronic ventilatory care settings are not the subject of this report.

According to conventional terminology, mechanical ventilation refers to the use of the ventilator to deliver inspired air through an artificial airway that is either the endotracheal tube or a tracheostomy tube. However, technology is now available that allows the provision of mechanical support without the use of an artificial airway. This technology is referred to as noninvasive positive pressure ventilation (NPPV). The advent of NPPV has, therefore, broadened the domain of mechanical ventilation and weaning from mechanical ventilation, as well as the scope of clinical research in these areas.

Questions On Weaning From Mechanical Ventilation

Our review addresses the five questions specified by AHRQ.

1.

When should weaning be initiated?

2.

What criteria should be used to initiate the weaning process?

3.

What are the most effective methods of weaning from mechanical ventilation?

4.

What are the optimal roles of nonphysician health care professionals in facilitating safe and expeditious weaning?

5.

What is the value of clinical practice algorithms and computers in expediting weaning?

This report summarizes the clinical literature that directly or indirectly addresses each of these questions. Some of these questions can be comprehensively addressed by the currently available literature (e.g., question 4 regarding multidisciplinary roles in facilitating weaning), whereas other questions are also well informed by physiologic studies outside the scope of this report (e.g., question 1 regarding when weaning should be initiated). Given that different ways of thinking about weaning could lead to different criteria for when weaning has started, the framing of the first two questions provides special challenges. We have dealt with this issue by reflecting on evidence that bears on the entire weaning process. We now provide some introductory comments about weaning from mechanical ventilation and our frame of reference for approaching these questions.

Importance of Optimizing the Weaning Process

The weaning process and definitions of weaning success and failure are variably defined by investigators in this field. According to the most liberal definition, attempts to reduce the level of support may begin as soon as mechanical ventilation begins, and virtually the whole period on the ventilator could be considered part of "weaning process." A much more conservative definition would reserve the term "weaning" for the final stages of mechanical ventilation in which the attending health care workers believe the possibility of the patient's breathing without assistance is near or imminent.

Weaning can be considered a gradual decrease in mechanical support, effected by: (1) increasing periods of unassisted breathing, (2) unassisted breaths alternating with progressively fewer ventilator breaths, or (3) reductions of ventilatory support in breaths triggered by patient effort. Unassisted breathing can be: (1) for short periods as part of a process designed to gradually decrease ventilatory support, (2) for periods typically up to 2 hours for assessing discontinuation readiness, or (3) as termination to the weaning process. Since many patients can be successfully liberated from mechanical ventilation after a brief spontaneous breathing trial, the term "weaning" might not be suitable in this instance. For some investigators, "weaning success" is defined as sustained spontaneous, unassisted breathing with or without an artificial airway, and for others it is defined as sustained extubation. Since our remit was neither to develop a new weaning definition nor select one particular definition, we considered weaning as broadly described by investigators in this field. Therefore, from the studies included in this systematic review, we recorded the weaning processes and outcomes that evaluated how best to predict and maintain liberation from mechanical ventilation.

By one estimate, patients spend, on average, approximately 41 percent of their time receiving mechanical ventilation in the weaning process (Esteban, Alia, Ibanez, et al., 1994). Efforts to reduce the duration of weaning are important because mechanical ventilation is associated with considerable morbidity, pulmonary barotrauma (Meade and Cook, 1995; Meade, Cook, Kernerman, et al., 1997; Slutsky and Trembly, 1998), ventilator-associated pneumonia (Cook, Walters, Brun-Buisson, et al., 1998; Papazian, Bregeon, Thirion, et al., 1996; Vincent, Bihari, Suter, et al., 1995), and mortality (Ely, Baker, Evans, et al., 1999; Fagon, Chastre, Vuagnat, et al., 1996). In addition to increasing morbidity and mortality, a long duration of mechanical ventilation is associated with increased health care costs, both because of a longer stay in the ICU and because of the costs associated with mechanical ventilation itself (e.g., more nursing and respiratory therapy time and more ventilator management interventions).

At the same time, premature discontinuation of mechanical ventilation can also contribute to failed extubation (a failed extubation is typically characterized by the need for reintubation within a short period of time) and thus contribute to ongoing risk of morbidity and ICU complications in critically ill patients. Failed trials of extubation can contribute to respiratory muscle fatigue or injury, which may in turn delay future weaning attempts and therefore prolong mechanical ventilation even further, possibly incurring the need for more sedation and potentiating cardiac ischemia. Aside from patients' distress if they must breathe unassisted before they are ready, reintubation is associated with airway trauma, oropharyngeal or gastric aspiration, acute lung injury, risk of cardiovascular compromise, and the adverse sequelae of episodic hypoxemia. Reintubation has also been associated with increased risk of mortality after controlling for severity of illness and comorbid conditions (Epstein and Ciubotaru, 1997; Esteban, Alia, Ibanez, et al., 1994).

In considering these issues, it is worthwhile to note that the goals of shortening the duration of weaning and avoiding premature termination of mechanical ventilation will always, to some extent, compete. We can think of alternative thresholds for decreasing and discontinuing ventilatory support. Clinicians choosing a low threshold will reduce support whenever patients might tolerate it and will therefore minimize the duration of mechanical ventilation. Clinicians who choose a high threshold will not reduce support unless they are confident patients will tolerate support reduction and may seldom need to reintubate patients.

In choosing thresholds for reducing mechanical support, clinicians thus face a tradeoff between decreasing the duration of ventilation at the risk of a larger number of weaning failures and reintubations or minimizing weaning failures at the risk of unnecessary prolonged ventilation. Defining the optimal tradeoff between shortening average times on the ventilator and minimizing the rate of reintubation presents a challenge for health care workers managing critically ill patients, particularly since clinicians currently disagree on what constitutes an "acceptable" rate of failed extubation.

This disagreement flows in part from a lack of clinical data. Both extending the duration of ventilation and the necessity for reintubation have direct consequences important to patients and the health care system. At the same time, both (and particularly the latter) constitute to some extent surrogate endpoints. Clinicians place a priority in avoiding reintubation because of concern that it may increase the frequency of pneumonia, myocardial infarction, and death. Prolonged ventilation prior to extubation may have an impact on the rate of these same outcomes. Thus, the immediate effect of alternative weaning strategies on duration of ventilation before extubation and on reintubation rates is poorly understood. Only knowledge of their effects on the putative complications of prolonged ventilation or intubation can fully inform tradeoffs between alternative weaning strategies.

How Can Clinicians Conduct an Optimal Wean?

One strategy for optimizing the duration of mechanical ventilation relies on optimal recognition of when patients would tolerate reduction in mechanical support and when, ultimately, they would tolerate discontinuation. Research in this area examines patient characteristics that might predict their ability to tolerate reductions or discontinuation of mechanical support.

As noted above, clinicians need to trade off goals of reducing mean time on the ventilator with minimizing rates of reintubation and, thus, defining the optimal threshold is an important goal for weaning research. At the same time, it would be highly desirable to find an approach that simultaneously minimizes both the duration of ventilation and the need for reintubation. Currently, clinicians use a variety of indicators including some assessed qualitatively (e.g., diaphoresis, anxiety) and some assessed quantitatively (e.g., vital signs, such as heart rate, respiratory rate, blood pressure, and pulmonary mechanics, such as tidal volume, airway occlusion pressure, work of breathing, rapid shallow breathing index), or scores that combine two or more of these variables to guide weaning. Clinicians may not only have different thresholds, they may also choose different patient characteristics to determine whether a patient has a decreased need, or no further need, of the support of a ventilator. Whatever patient characteristics they use, in making the final decision about reducing or discontinuing mechanical support, they intuitively weigh these factors.

It is therefore no surprise that a physician's judgment of a patient's readiness for weaning is often inaccurate. In one study, one-half of the patients whom physicians considered to be incapable of sustaining spontaneous ventilation subsequently tolerated a weaning trial (Stroetz and Nubmayr, 1995). All this raises the possibility that different health care workers may do a better job of rapid detection of problems with reductions in ventilatory support.

Perhaps a deeper quantitative understanding of the results of tests that may predict weaning success would improve clinicians' accuracy. Traditional indices of diagnostic test power include sensitivity (the proportion of those who successfully wean who have "positive" test results) and specificity (the proportion of those who do not successfully wean who have "negative" test results). Sensitivity and specificity are limited in that they rely on a single cutpoint or threshold and they do not provide an easy way to go from the pretest likelihood or probability of successful weaning, through the tests results, to the posttest probability. Likelihood ratios (the ratio of the proportion of patients with a test result who successfully wean to the proportion of patients with the same test result who do not successfully wean) allow consideration of multiple cutpoints or thresholds. Furthermore, a simple nomogram allows the transition from the pretest probability of success, through the test result, to the posttest probability.

Algorithms present an alternative approach that does not rely on clinicians' sophisticated use of information from diagnostic tests. Ventilation researchers have suggested standardized algorithms for advancing the wean which include criteria for cutting the rate of reduction of support or returning to a previous greater level of support (Ely, Baker, Dunagan, et al., 1996; Kollef, Shapiro, Silver, et al., 1997). Recently, computer technology for implementing these algorithms has become available (Strickland and Hasson, 1993). Clinician-guided or computer-guided protocols could, in theory, increase the rate at which patients wean without a consequent increase in premature reduction of support. On the other hand, protocols could, in theory, also make the weaning process formulaic in a manner that decreases caregiver sensitivity to subtle clinical findings that might predict adverse effects and attendant weaning failure. In one protocol-guided study, one-third of patients who failed to pass the daily screen were still successfully extubated, highlighting that rigid adherence to the protocol may not be ideal, since, in this case, it would have unnecessarily committed these patients to prolonged ventilation (Ely, Baker, Dunagan, et al., 1996).

A second fundamental strategy for optimizing the weaning process follows the decision that a patient can tolerate reduction of mechanical support and focuses on the choice of the method of reduction. Implicit in these options is the hypothesis that some methods of gradual support are better than others because of differential mechanics or respiratory muscle conditioning that might favor a patient's capacity to breath unassisted. Clinicians have a menu of ventilation modes from which they can choose. These include, but are not restricted to, T-piece trials, synchronized intermittent mandatory ventilation (SIMV), pressure support (PS), continuous positive airway pressure (CPAP), combinations of the foregoing, and newer approaches to weaning such as volume support, proportional assist ventilation (PAV), and NPPV. One of these methods, T-piece trials, can be used in two different ways, one of which is aligned with a traditional description of weaning (conceptualized as a gradual reduction in support) and one of which is not. For example, multiple periods of breathing through a T-piece that are gradually increased in length is more consistent with the former approach (Esteban, Alia, Ibanez, et al, 1994), whereas once or twice a day discontinuation assessments with the specific goal of achieving discontinuation from mechanical ventilation is more consistent with the latter approach (Brochard, Rauss, Benito, et al., 1994). Most clinicians are familiar with each of these strategies and use a subset of them to various degrees in practice.

It may be that some approaches to weaning work better than others in reducing the duration of weaning. Ascertaining the best approaches, and disseminating such information, could improve the quality of care of mechanically ventilated patients, and reduce morbidity and costs.