Chest Email Content Delivery
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Guest Access | Sign In via User Name/Password
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Meade, M.
Right arrow Articles by Epstein, S.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Meade, M.
Right arrow Articles by Epstein, S.
(Chest. 2001;120:400S-424S.)
© 2001 American College of Chest Physicians

Predicting Success in Weaning From Mechanical Ventilation*

Maureen Meade, MD; Gordon Guyatt, MD; Deborah Cook, MD; Lauren Griffith, MSc; Tasnim Sinuff, MD; Carmen Kergl, RRT; Jordi Mancebo, MD; Andres Esteban, MD and Scott Epstein, MD

* From the Departments of Medicine (Drs. Meade, Guyatt, Cook, and Sinuff) and Clinical Epidemiology & Biostatistics (Ms. Griffith), McMaster University, Hamilton, Canada; the Department of Respiratory Therapy (Ms. Kergl), Hamilton Health Sciences Corporation, Hamilton, Canada; the Department of Intensive Care (Dr. Mancebo), University of Barcelona, Hospital de Sant Pau, Barcelona, Spain; the Department of Intensive Care (Dr. Esteban), Hospital Universitario de Getafe, Madrid, Spain; and the Department of Medicine (Dr. Epstein), New England Medical Center, Tufts University, Boston, MA.

Correspondence to: Deborah Cook, MD, McMaster University, Faculty of Health Sciences Center, Department of Clinical Epidemiology, 1200 Main St West, Hamilton, Ontario, Canada; e-mail: debcook{at}mcmaster.ca


   Abstract
TOP
 Abstract
Introduction
 Materials and Methods
 Results
 Conclusion
 References
 
We identified 65 observational studies of weaning predictors that had been reported in 70 publications. After grouping predictors with similar names but different thresholds, the following predictors met our relevance criteria: heterogeneous populations, 51; COPD patients, 21; and cardiovascular ICU patients, 45. Many variables were of no use in predicting the results of weaning. Moreover, few variables had been studied in > 50 patients or had results presented to generate estimates of predictive power. For stepwise reductions in mechanical support, the most promising predictors were a rapid shallow breathing index (RSBI) < 65 breaths/min/L (measured using the ventilator settings that were in effect at the time that the prediction was made) and a pressure time product < 275 cm H2O/L/s. The pooled likelihood ratios (LRs) were 1.1 (95% confidence interval [CI], 0.95 to 1.28) for a respiratory rate [RR] of < 38 breaths/min and 0.32 (95% CI, 0.06 to 1.71) for an RR of > 38 breaths/min, which indicate that an RR of < 38 breaths/min leaves the probability of successful weaning virtually unchanged but that a value of > 38 breaths/min leads to a small reduction in the probability of success in weaning the level of mechanical support. For trials of unassisted breathing, the most promising weaning predictors include the following: RR; RSBI; a product of RSBI and occlusion pressure < 450 cm H2O breaths/min/L; maximal inspiratory pressure (PImax) < 20 cm H2O; and a knowledge-based system for adjusting pressure support. Pooled results for the power of a positive test result for both RR and RSBI were limited (highest LR, 2.23), while the power of a negative test result was substantial (ie, LR, 0.09 to 0.23). Summary data suggest a similar predictive power for RR and RSBI. In the prediction of successful extubation, an RR of < 38 breaths/min (sensitivity, 88%; specificity, 47%), an RSBI < 100 or 105 breaths/min/L (sensitivity, 65 to 96%; specificity, 0 to 73%), PImax, and APACHE (acute physiology and chronic health evaluation) II scores that are obtained at hospital admission appear to be the most promising. After pooling, two variables appeared to have some value. An RR of > 38 breaths/min and an RSBI of > 100 breaths/min/L appear to reduce the probability of successful extubation, and PImax < 0.3, for which the pooled LR is 2.23 (95% CI, 1.15 to 4.34), appears to marginally increase the likelihood of successful extubation. Judging by areas under the receiver operator curve for all variables, none of these variables demonstrate more than modest accuracy in predicting weaning outcome. Why do most of these tests perform so poorly? The likely explanation is that clinicians have already considered the results when they choose patients for trials of weaning.

Key Words: extubation • mechanical ventilation • meta-analysis • methods • modes • reintubation • systematic reviews • weaning


   Introduction
TOP
 Abstract
 Introduction
Materials and Methods
 Results
 Conclusion
 References
 
Critical -care clinicians must carefully weigh the benefits of rapid liberation for mechanical ventilation against the risks of premature trials of spontaneous breathing and extubation. The need for accurate prediction applies to all phases of weaning, beginning with reductions in mechanical support, as patients are increasingly able to support their own breathing, followed by trials of unassisted breathing, which often precede extubation, and ending with extubation.

Patients may fail to wean as a result of impaired respiratory center drive or, more frequently as a result of neuromuscular abnormalities including respiratory muscle fatigue, impaired lung mechanics, or impaired gas exchange capability. Patients may successfully be weaned to minimal levels of respiratory support but may still fail extubation as a result of airway abnormalities. Based on experimental data in healthy individuals1 and animals,2 and based on observational data from patients that suggest the development of respiratory muscle fatigue during unsuccessful weaning,3 4 5 6 some investigators postulate that failed trials of discontinuation of mechanical ventilation may precipitate respiratory muscle injury and, ultimately, prolong the duration of mechanical ventilation. Therefore, criteria have been sought to identify patients who are likely to fail, so thatpremature trials of spontaneous breathing can be avoided. Moreover, failed trials of extubation have been associated with excess hospital mortality, prolonged ICU and hospital stays, and increased need for tracheostomy.7 8

The predictors of weaning that clinicians currently use, and that investigators have studied, include an assortment of demographic characteristics (ie, age and diagnostic categories), subjective signs (ie, diaphoresis and agitation), vital signs and hemodynamic variables (ie, heart rate and BP), lung mechanics (ie, tidal volume and respiratory rate [RR]), gas exchange (ie, PaO2 and PaCO2 levels), and severity-of-illness measures (ie, biochemical variables, comorbidities, levels of respiratory support, and levels of nonrespiratory support). Investigators have tested these variables individually, as composite scores or derivations, and as complex systems. Since the reasons that patients fail weaning may vary among different patient populations, the predictors of weaning also may vary. For instance, predictor variables that are useful in patients undergoing cardiac surgery may differ from those that are of value in patients with COPD, and both may differ from predictors that are useful in a general case mix of ICU patients.

In this article, we separate studies into three groups according to the following target populations: a relatively heterogeneous mix of ICU patients; patients with COPD; and patients who had undergone cardiac surgery (Table 1 ). One can think of three stages in the weaning process. In the first, the clinician progressively reduces, in a stepwise fashion, the level of support. In the second stage, the patient undergoes a trial of unassisted breathing. In the final stage, the clinician extubates the patient. Investigators have addressed prediction at each stage of the process. Thus, we also classified studies according to what the investigators were trying to predict in the following way: the success of stepwise reductions in mechanical support; unassisted breathing trials; extubation; and the result of trials of unassisted breathing plus extubation (Table 1) . We include as trials of "unassisted" breathing those trials completed on a low level of pressure support to overcome the additional work of breathing through a ventilator circuit or those completed on a low level of continuous positive airway pressure to offset the loss of physiologic continuous positive airway pressure caused by the presence of an endotracheal tube.


View this table:
[in this window]
[in a new window]

  		Table 1. Populations and Outcomes

 

   Materials and Methods
TOP
 Abstract
 Introduction
 Materials and Methods
Results
 Conclusion
 References
 
Eligibility Criteria
We sought studies that included any patients receiving mechanical ventilation in an ICU setting that examined potential predictors of success in stepwise reductions in mechanical ventilation, trials of spontaneous breathing, extubation, or any combination of these outcomes.

We also included studies of the predictors of the duration of mechanical ventilation in cardiac surgery patients and COPD patients. Randomized trials and controlled clinical trials were included. We excluded predictors of self-extubation. Although they are representative of an important body of literature in this field, we excluded studies that were designed primarily to evaluate the reproducibility in the measurement of various predictors of weaning success or duration of ventilation.

Search for Relevant Studies
To identify relevant studies, we searched MEDLINE, Excerpta Medica Database, HEALTHStar, CINAHL (Cumulative Index to Nursing and Allied Health Literature), the Cochrane Controlled Trials Registry, and the Cochrane Data Base of Systematic Reviews from 1971 to September 1999, and personal files. We examined the reference lists of all included articles for other potentially relevant citations. In addition, we hand-searched the respiratory therapy journal Respiratory Care from 1997 to 1999. We did not explicitly search for unpublished literature. Our search strategies are available on request.

Two reviewers examined each title and abstract. Reviewers included either two of the investigators or one investigator and a senior respiratory therapist. We took a comprehensive approach and retrieved all articles that either reviewer considered to be possibly eligible. Two reviewers also examined the full text and made final decisions regarding eligibility based on the inclusion and exclusion criteria described above. These decisions were made unblinded to the source, authors, and conclusions of each study. Disagreements were resolved by consensus.

Data Abstraction
Data were abstracted and methodological quality was assessed in duplicate by two of five respiratory therapists and five intensivists. One of the senior investigators rechecked the final data abstraction.

Depending on the data available to us, we reported the results of studies of weaning predictors in a variety of ways. These include the following: the means or medians in patients who were successfully and unsuccessfully weaned; the proportions of patients with results more extreme than the specified thresholds; sensitivity, specificity, likelihood ratios (LRs), or predictive values; Pearson correlation coefficients; and {chi}2 tests, Student’s t tests, analysis of variance, and univariate and multivariate regressions. We implemented a process for data abstraction that allowed for the recording of all data types.

Study results may be influenced by the extent to which investigators control for important potential sources of bias in predicting weaning success and failure. Therefore, we also recorded aspects of study design, including the following: (1) whether investigators enrolled a representative sample of patients (or, alternatively, whether selection bias was evident); and (2) whether those making weaning decisions or assessing outcomes were aware of predictor variables (ie, blinding).

Finally, the applicability of study results depends on the adequate reporting of information related to patient populations and experimental methods. We recorded this information as well.

Relevant Predictors
Because of the very large number of predictor variables, our goal was a manageable presentation of the data. We present the results only for those studies in which predictors showed even a modest potential for differentiating success from failure in weaning. We developed a number of guidelines for what we considered to be a modest potential for differentiating success from failure.

  1. We present all clearly specified predictors for which results could be recorded in 2 x 2 tables if there was an associated biologically sensible LR of > 2 or < 0.5.
  2. When investigators presented results as means and SDs of the success and failure groups, we present predictors if the difference in means between the two groups was greater than one half of the smaller of the SDs of the two groups.
  3. When there was no information about the power of the predictor in terms of either LRs or the distributions of predictor results in the success and failure groups, we included predictors with a statistically significant association with the outcome of interest (for instance, on multiple regression analysis).

In many instances, a predictor met one of these three criteria in some but not all studies. When the results differed across studies within one of the three populations, we included the predictor, unless it was not predictive in the majority of studies and in the majority of patients. For example, if only one of many studies found a predictor to be of value, we present the results with this predictor if the study sample size was > 50% of the total sample size of all studies that examined that predictor.

Terminology
We use the following terminology in our interpretation and presentation of test results. We classify a test result as positive if it increases the likelihood of successful weaning (sensitivity is therefore the proportion of patients who have experienced successful weaning who have a positive test result) and as negative if it decreased the likelihood of successful weaning (specificity is then the proportion of patients whose weaning failed who had a negative test result). When using LRs, an LR of 1 means that the posttest probability is the same as the pretest probability and, thus, that the test result is unhelpful. Values of 1 to 2 (which raise probability as much as values of 1 to 0.5 lower the probability) change probability very little, values of 2 to 5 or 0.5 to 0.2 lead to small changes in probability, values of 5 to 10 or 0.2 to 0.1 lead to moderate changes in probability, and values of > 10 or < 0.1 lead to large changes in probability.

Statistical Analysis
If we identified more than one study examining a relevant predictor and presenting these data in a manner allowing the creation of a 2 x 2 table, we summarized data in the form of LRs.9 The majority of studies, however, presented data only as group means and SDs. To transform these data into LRs, we tested the assumption of normality by inspecting the mean and SD for skewness. To do so, we noted the occasions on which the value obtained by adding 2 SDs to the mean and subtracting 2 SDs from the mean yielded clinically implausible values. If we could assume normality, knowing the total sample size and the number of patients in the successfully and unsuccessfully weaned groups of patients, we estimated the number of patients in each cell of a 2 x 2 table. We used the predictor threshold that was most often provided by the investigators to create these LRs. We calculated confidence intervals (CIs) for all summary measures.10

Where appropriate, we pooled data across studies to narrow the 95% CIs around estimates of accuracy in prediction. Using these data transformations, we calculated the pooled LR of a positive test result, the pooled LR of a negative test result, the pooled sensitivity and specificity of a given predictor threshold, and an associated pooled odds ratio (OR).11 We did not pool LRs across studies in which some investigators presented their results as binary variables while others presented their results as continuous variables. Whenever we could pool three or more studies, we also constructed a summary receiver operating characteristic (ROC) curve. We tested ROC curves for the presence of a threshold effect (ie, the presence of a natural cutoff, or threshold, value), and for accuracy (using Q tests and area under the curves).

Definitions of Predictor Variables
Investigators defined two variables, maximal inspiratory pressure (PImax) and negative inspiratory force (NIF), in many different ways (ie, "PImax," "NIF," negative inspiratory pressure ["NIP"], and maximal inspiratory pressure ["MIP"]). For the purposes of this report, we refer to PImax when investigators described PImax that was measured in an occluded airway after 20 s starting from residual volume, and we refer to NIF when negative pressure was measured after at least 1 s of inspiratory effort against an occluded airway and the most negative value of three attempts was recorded.


   Results
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
Conclusion
 References
 
We identified 65 observational studies12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 of weaning predictors that were reported in 70 publications, of which 2 studies75 76 are not included in our tables. Of these, 41 studies included heterogeneous ICU populations, 6 included only patients with COPD, and 16 studies evaluated weaning predictors in the cardiovascular ICU (CVICU). We found 462 putative weaning predictors. After grouping together predictors with similar names but different thresholds (ie, grouping together RR, RR < 35 breaths/min, and RR < 38 breaths/min), the following numbers of predictors met our relevance criteria in each group: heterogeneous populations, 51; COPD patients, 21; and CVICU patients, 45.

In general, this literature is limited by a lack of blinding; that is, caregivers making decisions about weaning were aware, to some extent (either explicitly or vaguely), of the values of the predictor variables and may have included this information in their bedside assessments. For most predictor variables, this methodological limitation was unavoidable. Only a few notable studies appeared to achieve blinding of both caregivers (ie, those deciding on whether or not to reduce mechanical support, to end a trial of unassisted breathing, or to extubate) and outcome assessors. Lack of blinding, which was characteristic of the remaining studies, usually results in inflated estimates of predictive accuracy.

The importance of selection bias in this literature was difficult to assess because the reporting of patient selection in individual studies was not detailed. For the vast majority of studies, selection bias was not evident.

Many studies omitted to report information bearing on the applicability of their results. For instance, most studies did not mention whether patients with a tracheostomy were included, how decisions to perform tracheostomy were handled in the study protocol, or whether this procedure was taken into account during the analysis. Patients with a tracheostomy might fare differently on numerous tests for weaning, and systematically excluding these patients would alter the patient population and the corresponding test properties.

Weaning Predictors in Heterogeneous Patient Populations
Tables 2 3 4 4A 5 6 6A 6B 7 8 8A 8B 9 summarize the studies evaluating weaning predictors in heterogeneous populations of mechanically ventilated patients in ICUs.


View this table:
[in this window]
[in a new window]

  		Table 2. Predictors of Success in Reducing Mechanical Ventilatory Support*

 

View this table:
[in this window]
[in a new window]

  		Table 3. Pooled Results for Predictors of Success in Reducing Mechanical Ventilatory Support*

 

View this table:
[in this window]
[in a new window]

  		Table 4. Predictors of Success in Trials of Unassisted Breathing*

 

View this table:
[in this window]
[in a new window]

  		Table 4A. Continued

 

View this table:
[in this window]
[in a new window]

  		Table 5. Pooled Results for Predictors of Success in Trials of Unassisted Breathing*

 

View this table:
[in this window]
[in a new window]

  		Table 6. Predictors of Successful Extubation*

 

View this table:
[in this window]
[in a new window]

  		Table 6A. Continued

 

View this table:
[in this window]
[in a new window]

  		Table 6B. Continued

 

View this table:
[in this window]
[in a new window]

  		Table 7. Pooled Results for Predictors of Successful Extubation*

 

View this table:
[in this window]
[in a new window]

  		Table 8. Predictors of Successful Trials of Unassisted Breathing Followed by Extubation*

 

View this table:
[in this window]
[in a new window]

  		Table 8A. Continued

 

View this table:
[in this window]
[in a new window]

  		Table 8B. Continued

 

View this table:
[in this window]
[in a new window]

  		Table 9. Pooled Results for Predictors of Trials of Unassisted Breathing Followed by Extubation*

 
For stepwise reductions in mechanical support, the most promising weaning predictors are a rapid shallow breathing index (RSBI) of < 65 breaths/min/L made using the ventilator settings that were in effect at the time the prediction is made and a pressure time product of < 275 cm H2O/L/s (Table 2 ).12 13 14 15 The small sample size of the study that reported these results (40 patients) limits the associated strength of inference.

Table 3 presents the pooled results for the only predictor (RR) for which data were amenable to pooling. The pooled LRs are 1.1 (95% CI, 0.95 to 1.28) for an RR of < 38 breaths/min and 0.32 (95% CI, 0.06 to 1.71) for an RR of > 38 breaths/min, indicating that an RR of < 38 breaths/min leaves the probability of successful weaning virtually unchanged but a value of > 38 breaths/min leads to a small reduction in the probability of success in weaning the level of mechanical support. The wide CI around the LR leaves even this estimate open to considerable uncertainty.

For trials of unassisted breathing, the most promising weaning predictors from the review of individual studies (Table 4 )16 17 18 19 20 21 22 23 include the following: RR; RSBI; the product of RSBI and airway pressure 0.1 s after the occlusion of the inspiratory port of a unidirectional balloon occlusion valve (P0.1) (ie,RSB-P0.1 index) < 450 cm H2O breaths/min/L; PImax < 20 cm H2O; and a knowledge-based system for adjusting pressure support. Data allowed pooled estimates for two of these variables (RR and RSBI) (Table 5 ). Pooled results are consistent across studies that provided binary data and those that provided only continuous data. For both RR and RSBI, the power of a positive test result was very limited (highest LR, 2.23), while the power of a negative test result was substantial (LR, 0.09 to 0.23). Summary data suggest a similar predictive power of RR and RSBI.

In the prediction of successful extubation (Table 6 ),24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 an RR of < 38 breaths/min (sensitivity, 88%; 5specificity, 47%), an RSBI of < 100 breaths/min/L or 105 breaths/min/L (sensitivity, 65 to 96%; specificity, 0 to 73%), PImax, and APACHE (acute physiology and chronic health evaluation) II scores measured at hospital admission appear to be the most promising. After pooling (Table 7 ), two variables appeared to have some value. An RR of > 38 breaths/min and an RSBI of > 100 breaths/min/L appear to reduce the probability of successful extubation, and an inspiratory pressure/PImax ratio of < 0.3 (pooled LR, 2.23; 95% CI, 1.15 to 4.34) appears to marginally increase the likelihood of successful extubation.

Several studies evaluated the ability to predict the combined outcome of a successful trial of unassisted breathing followed by successful extubation. Predictor variables that showed some promise on review of the individual study results (Table 8 )43 44 45 46 47 48 49 50 51 52 include the following: duration of ventilation prior to weaning; an RR < 38 breaths/min (sensitivity, 92% [100 patients]); tidal volume, > 4 mL/kg (sensitivity: in 100 adults, 94%; in 84 children, 94%); an RSBI of < 100 breaths/min/L; an NIF of < -20 cm H2O; PImax; P0.1 of < 5.0 cm H2O (sensitivity, 87%; and specificity, 91% [in 67 patients]); and P0.1/PImax ratio. Several other studies suggested potentially powerful predictors but enrolled <= 30 patients. In all studies, the predictors were measured immediately prior to the trial of unassisted breathing or early during the initiation of the trial. In Table 9 , we present the results of pooled analyses. The RSBI yielded a statistically significant pooled LR of 1.58 (95% CI, 1.30 to 1.90), indicating that it remains a very weak predictor. P0.1/PImax ratio yielded a much more clinically useful pooled LR of 16.3 (95% CI, 2.35 to 113).

Summary ROC curves deal with the problem of different thresholds among studies. We show the summary ROC curves for several predictors of successful extubation (Figs 1 ,2 ,3 ) and of successful trials of unassisted breathing and extubation (Figs 4 ,5 ,6 ,7 ). Testing for the presence of a threshold effect indicated that none of these variables were associated with an ideal cut point or threshold level for weaning. Moreover, judging by the modest areas under the curve for all variables, none of these variables demonstrate more than modest accuracy in predicting weaning outcome, and none appear to perform any better than the others.



View larger version (18K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Summary ROC curve for minute ventilation predicting successful extubation.

 


View larger version (19K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2. Summary ROC curve for RR predicting successful extubation.

 


View larger version (22K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3. Summary ROC curve for RSBI predicting successful extubation.

 


View larger version (21K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4. Summary ROC curve for minute ventilation predicting successful trials of unassisted breathing.

 


View larger version (22K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5. Summary ROC curve for RR predicting successful trials of unassisted breathing and extubation.

 


View larger version (21K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 6. Summary ROC curve for tidal volume predicting successful trials of unassisted breathing and extubation.

 


View larger version (24K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 7. Summary ROC curve for RSBI predicting successful trials of unassisted breathing and extubation.

 
Weaning Predictors for Patients With COPD
Hilbert et al53 evaluated a number of variables, including RR, RSBI, P0.1, effective inspiratory impedance, and PaO2/fraction of inspired oxygen (FIO2) ratio, for the ability to predict success on a trial of extubation in 40 patients with COPD but found none to be of any value (Table 10 ).


View this table:
[in this window]
[in a new window]

  		Table 10. Predictors of Successful Extubation in Patients With COPD

 
Two groups of investigators evaluated predictors of success in trials of unassisted breathing followed by extubation in two relatively small studies (N = 26 and N = 31) (Table 11 ).54 55 A gastric intramucosal pH > 7.3 and a gastric intramucosal PaCO2 < 60 mm Hg showed some promise as weaning predictors in a study of 26 patients.


View this table:
[in this window]
[in a new window]

  		Table 11. Predictors of Successful Trials of Unassisted Breathing and Extubation in Patients With COPD*

 
Finally, Table 12 56 57 58 summarizes studies evaluating the prediction of successful extubation at 4 months in patients with COPD. Menzies et al56 examined the predictive power of a number of variables that were recorded in COPD patients in the first 3 days after their admission to an ICU. The investigators recorded variables immediately before a trial of unassisted breathing through a ventilator circuit. Nava et al57 enrolled only COPD patients who resided in a rehabilitation unit and who had received mechanical ventilation for at least 21 days. They examined the predictive power of variables recorded between 5 and 10 days after hospital admission during a period of clinical stability. The investigators almost invariably presented their results as differences in means and SDs between groups that did or did not wean from mechanical ventilation, a format that is not easily applied to patient-care decisions.


View this table:
[in this window]
[in a new window]

  		Table 12. Predictors of Extubation Within 4 Months for Patients With COPD

 
Weaning Predictors in the CVICU
Two groups of investigators59 60 have studied predictors for trials of unassisted breathing in the CVICU (Table 13 ). Neither report included threshold values that could be applied in the clinical arena, rather, all results were presented as means and SDs among patients who passed and failed trials of unassisted breathing. The investigators found a large number of predictor variables that were associated with successful trials.


View this table:
[in this window]
[in a new window]

  		Table 13. Predictors of Successful Trials of Unassisted Breathing in the CVICU*

 
Table 14 61 62 presents predictors of successful extubation, which also have been studied by two separate groups of investigators. Once again, there appears to be a large number of variables that are associated with successful extubation, although the investigators do not provide any threshold values.


View this table:
[in this window]
[in a new window]

  		Table 14. Predictors of Successful Extubation in the CVICU*

 
In a single study of 23 CVICU patients, Saito et al63 evaluated P0.1 < 4.0 cm H2O as a predictor of success on a trial of unassisted breathing followed by successful extubation. Their measure had a sensitivity of 100% and a specificity of 56% (LR for a positive result, 2.3; and LR for a negative result, 0).

Another single study64 of 230 patients evaluated predictors for successful extubation within 24 h of the patient undergoing cardiovascular surgery. The authors presented their results as differences in means and SDs in those patients who successfully underwent extubation by 24 h after surgery and those who did not. Successfully extubated patients had a statistically significant larger vital capacity, a shorter operating room time, and a higher PaCO2 level, but the differences between groups were small. Patients who were successfully extubated had a mean American Society of Anesthesia surgical risk score of 1.5, while those patients who were not extubated successfully by 24 h after surgery had a surgical risk score of 3.3.

A separate group of 10 studies (Table 15 15A )65 66 67 68 69 70 71 72 73 74 evaluated the ability of variables to predict the duration of mechanical ventilation following cardiac surgery. The predictor variables considered included those related to preoperative morbidity (eg, prior myocardial infarction), pre-ICU respiratory mechanics (eg, FEV1 percent predicted), surgical issues (eg, second cardiac surgery procedures), and postoperative events (eg, new Q waves on ECG). In general, this table provides information about which variables might be the most important to consider, although the relative importance of each variable, and the threshold values of importance for each variable, were not available. The variables with the greatest potential include just one preoperative variable (preoperative length of stay), one intraoperative variable (fentanyl dose), and a number of postoperative variables (duration of mechanical ventilation prior to weaning, maximum expiratory pressure, presence of new Q waves, degree of bleeding and RBC transfusion, and decreased cardiac output). The only variable that was examined in more than one study was whether patients had undergone coronary artery bypass surgery. The results suggest only a small decrease in the probability of successful extubation (LR, 0.42; 95% CI, 0.24 to 0.75) for patients who had undergone a procedure other than coronary artery bypass grafting (CABG).


View this table:
[in this window]
[in a new window]

  		Table 15. Predictors of Duration of Mechanical Ventilation Following Cardiac Surgery*

 

View this table:
[in this window]
[in a new window]

  		Table 15A. Continued

 

   Conclusion
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Conclusion
References
 
Studies have evaluated an extraordinarily diverse collection of variables for their ability to predict successful weaning and/or duration of mechanical ventilation. Many of these physiologic predictors already have provided great insights into the mechanisms of the failure of liberation. However, from a clinical point of view, the results are disappointing. First, a large number of predictors were found to be of no use in predicting the results of weaning. We found few predictors (1) that had been studied in > 50 patients and (2) for which investigators presented data that allowed estimates of the predictive power, and (3) had, at least in some studies, appreciable predictive power. Of these predictors, none are extremely powerful, and their results are not consistent across studies.

Only twice, after pooling, did we observe an LR of > 10 or < 0.1. The P0.1/PImax ratio was highly predictive of trials of unassisted breathing and extubation in two studies, with a pooled LR of 16.3 (95% CI, 2.35 to 113). Most of the remaining tests did not bear results that are very helpful in increasing or decreasing the probability of success. We did not observe any pooled LRs between 5 and 10, although we did observe five variables with LRs < 0.2, indicating that a negative test result is associated with a moderate reduction in the probability of weaning. These variables for the combined end point of a successful trial of unassisted breathing followed by successful extubation included the following: an RSBI < 100 breaths/min/L for trials of unassisted breathing; the compliance/rate/oxygenation/pressure (CROP) index for trials of extubation and an RR of > 38 breaths/min; tidal volume standardized to body weight; and NIF of < 20 to 25 cm H2O. Therefore, on balance, the best results achieved with any of these tests were moderate reductions in the probability of successful weaning in association with a negative test result.

The virtual absence of any tests with high LRs (thereby markedly increasing the probability of successful weaning) and the less infrequent occurrence of tests with LRs substantially < 1 (thereby appreciably decreasing the likelihood of successful weaning) corresponds to tests with high sensitivity (ie, > 90%) but unimpressive specificity. Again, this corresponds to positive test results that do not increase the likelihood of success substantially and to negative test results that sometimes decrease the probability of success appreciably. For example, assuming a pretest probability of success of 50%, a high RR (ie, > 38 breaths/min; LR, 0.32) will decrease the probability of success in reducing mechanical ventilation support from 50% to approximately 25%, the probability of success in a trial of unassisted breathing (LR, approximately 0.2) to approximately 20%, and the probability of success in a trial of extubation (LR, approximately 0.55) to approximately 33%.

The most frequently studied test, and one of the most powerful, is the RSBI. Pooled results for this test consistently show that a positive result (ie, a breathing patternthat is neither rapid nor shallow) is minimally helpful in increasing the probability of successful weaning. LRs from individual studies are usually < 2, meaning that the pretest probability of 50% will rise no higher than 66%. Considering the pooled data, the LR for the RSBI at predicting successful trials of unassisted breathing was 1.7, the LR for predicting successful extubation was between 1.3 and 1.8 (the latter value occurs when the variable is indexed to body weight), and the LR for predicting successful trials of unassisted breathing and extubation was as high as 2.8.

LRs associated with a negative result (ie, breathing that tends to be rapid and shallow) were 0.11 for predicting unassisted breathing, 0.39 for successful extubation, and 0.22 for the combined end point of unassisted breathing and extubation. These LRs correspond to decreases in the probability of success from 50% to 10%, 28%, and 18%, respectively.

Another observation about these studies is that measurement techniques often have differed across studies; large coefficients of variations have been demonstrated when different investigators make these measurements.77 An additional challenge is the absence of objective criteria to determine the tolerance for a trial of discontinuation or extubation, and the variation across studies.

Why do most of these tests perform so poorly, and why do so few provide helpful information? The likely explanation is that clinicians already have considered the results when they choose patients for trials of weaning. For instance, clinicians may seldom test patients who have very high RRs, who are capable of generating only very low pressures, or patients whose tidal volumes are very low for their ability to wean. Similarly, clinicians may not wait until the RR, tidal volume, or pressure generation is normal before they undertake weaning, for this would lead to excessive time spent receiving mechanical ventilation. Thus, the range of results is relatively narrow. The more narrow the range of results, the less likely that a test can discriminate between patients destined to fail a weaning trial and those destined to succeed.

Furthermore, when results of a single test are more extreme, it is likely that physicians are attempting to wean the patient only because other observations suggest the limited impact of an isolated aberrant finding. For instance, adequate tidal volume and pressure generation may indicate to a clinician that an elevated RR is due largely to patient anxiety and does not indicate that the patient will be unable to be weaned from mechanical ventilation.

In essence, this means that the predictive power of the tests is "used up" by the time that investigators formally test their properties in patients that clinicians already have decided are candidates for weaning. Thus, it is unrealistic to expect physiologic tests to be highly predictive in patients in whom clinicians judge to have an intermediate probability of weaning success.

Future Research
LRs provide the best format for presenting the results of weaning predictors, and future research should consider this presentation metric. Sensitivity and specificity provide common, but less easily applied, measures of predictive power. Reporting only means and measures of variance for groups that have undergone successful and unsuccessful weaning, or reporting regression coefficients and p values, is far less useful in terms of clinical application.

The results of these studies would be more helpful to clinicians if data were reported related to multiple cut points for a given variable, rather than a single cut point. For instance, rather than reporting success rates in patients with RRs of > 36 breaths/min and < 36 breaths/min, investigators should report success rates in patients with RRs of < 20, 21 to 28, 29 to 36, 36 to 44, and > 44 breaths/min. These cut points are obviously somewhat arbitrary. The point is that since extreme results may be highly predictive, intermediate results may be somewhat predictive, and results at the margin may not be predictive at all. The use of a single cut point or threshold obscures this important information.

Having said this, investigators and clinicians should not expect any test to be particularly powerful. The findings to date validate the clinical intuition. Once clinicians have decided that a patient is likely but not certain to be weaned from mechanical ventilation, a formal examination of physiologic tests that the clinician has in some way considered in making the decision about pretest probability is unlikely to be very helpful.

As we point out elsewhere in this supplement, formal weaning protocols may perform better than usual clinical care. When the predictors of weaning are incorporated in such protocols, they retain their full predictive power, because clinicians have not already used them to select a subgroup of patients whom they are considering for weaning. We believe that, at least in clinical research, further testing of formal weaning protocols represents the best step forward, rather than focusing exclusively on testing physiologically predictive information to optimize the weaning process.

The data included in this systematic review and a more comprehensive discussion of the original articles are included in an Evidence Report of the Agency for Healthcare Research and Quality.78


   Footnotes
 
Abbreviations: APACHE = acute physiology and chronic health evaluation; CABG = coronary artery bypass grafting; CI = confidence interval; CVICU = cardiovascular ICU; FIO2 = fraction of inspired oxygen; LR = likelihood ratio; NIF = negative inspiratory force; OR = odds ratio; P0.1 = airway pressure 0.1 s after the occlusion of the inspiratory port of a unidirectional balloon occlusion valve; PImax = maximal inspiratory pressure; ROC = receiver operating characteristic; RR = respiratory rate; RSBI = rapid shallow breathing index

This article is based on work performed by the McMaster University Evidence-based Practice Center, under contract to the Agency for Healthcare Research and Quality (Contract No. 290-97-0017), Rockville, MD.


   References
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Conclusion
 References
 

  1. Laghi, F, D’Alfonso, N, Tobin, MJ (1995) Pattern of recovery from diaphragmatic fatigue over 24 hours. J Appl Physiol 79,539-546[Abstract/Free Full Text]
  2. Reid, WD, Huang, J, Bryson, S, et al (1994) Diaphragm injury and myofibrillar structure inducted by resistive loading. J Appl Physiol 76,176-184[Abstract/Free Full Text]
  3. Capdevila, X, Perrigault, PF, Ramonatxo, M, et al (1998) Changes in breathing pattern and respiratory muscle performance parameters during difficult weaning. Crit Care Med 26,79-87[CrossRef][ISI][Medline]
  4. Cohen, CA, Zagelbau, D, Gross, D, et al (1982) Clinical manifestations of inspiratory muscle fatigue. Am J Med 73,308-316[CrossRef][ISI][Medline]
  5. Jubran, A, Tobin, MJ (1997) Pathophysiologic basis of acute respiratory distress in patients who fail a trial of weaning from mechanical ventilation. Am J Respir Crit Care Med 155,906-915[Abstract]
  6. Vassilakopoulos, T, Zakynthinos, S, Roussos, C (1998) The tension-time index and the frequency/tidal volume ratio are the major pathophysiologic determinants of weaning failure and success. Am J Respir Crit Care Med 158,378-385[Abstract/Free Full Text]
  7. Epstein, SK, Ciubotaru, RL, Wong, JB (1997) Effect of failed extubation on the outcome of mechanical ventilation. Chest 112,186-192[Abstract/Free Full Text]
  8. Esteban, A, Alia, I, Gordo, F, et al (1997) Extubation outcome after spontaneous breathing trials with t-tube or pressure support ventilation. Am J Respir Crit Care Med 156,459-465[Abstract/Free Full Text]
  9. Irwig, L, Tosteson, A, Gatsonis, C, et al (1994) Guidelines for meta-analyses evaluating diagnostic tests. Ann Intern Med 120,667-676[Abstract/Free Full Text]
  10. Simel, D, Samsa, G, Matchar, D (1991) Likelihood ratios with confidence: sample size estimation for diagnostic test studies. J Clin Epidemiol 44,763-770[CrossRef][ISI][Medline]
  11. Littenberg, B, Moses, L (1993) Estimating diagnostic accuracy from multiple conflicting reports: a new meta-analytic method. Med Decis Making 13,313-321
  12. Linton, DM, Potgieter, PD, Davis, S, et al (1994) Automatic weaning from mechanical ventilation using an adaptive lung ventilation controller. Chest 106,1843-1850[Abstract/Free Full Text]
  13. Rivera, L, Weissman, C (1997) Dynamic ventilatory characteristics during weaning in postoperative critically ill patients. Anesth Analg 84,1250-1255[Abstract]
  14. Stroetz, RW, Hubmayr, RD (1995) Tidal volume maintenance during weaning with pressure support. Am J Respir Crit Care Med 152,1034-1040[Abstract]
  15. Oh, TE, Bhatt, S, Lin, ES, et al (1991) Plasma catecholamines and oxygen consumption during weaning from mechanical ventilation. Intensive Care Med 17,199-203[CrossRef][ISI][Medline]
  16. Frutos, F, Alia, I, Esteban, A, et al (1995) Clinical changes during a T-tube weaning trial. Med Intensiva 19,343-348
  17. Del Rosario, N, Sassoon, CS, Chetty, KG, et al (1997) Breathing pattern during acute respiratory failure and recovery. Eur Respir J 10,2560-2565[Abstract]
  18. Dojat, M, Harf, A, Touchard, D, et al (1996) Evaluation of a knowledge-based system providing ventilatory management and decision for extubation. Am J Respir Crit Care Med 153,997-1004[Abstract]
  19. Chatila, W, Jacob, B, Guaglionone, D, et al (1996) The unassisted respiratory rate-tidal volume ratio accurately predicts weaning outcome. Am J Med 101,61-67[CrossRef][ISI][Medline]
  20. Sassoon, CS, Mahutte, CK (1993) Airway occlusion pressure and breathing pattern as predictors of weaning outcome. Am Rev Respir Dis 148,860-866[ISI][Medline]
  21. Jabour, ER, Rabil, DM, Truwit, JD, et al (1991) Evaluation of a new weaning index based on ventilatory endurance and the efficiency of gas exchange. Am Rev Respir Dis 144,531-537[ISI][Medline]
  22. Saura, P, Blanch, L, Mestre, J, et al (1996) Clinical consequences of the implementation of a weaning protocol. Intensive Care Med 22,1052-1056[ISI][Medline]
  23. Kennedy, SK, Weintraub, RM, Skillman, JJ (1977) Cardiorespiratory and sympathoadrenal responses during weaning from controlled ventilation. Surgery 82,233-240[ISI][Medline]
  24. Tahvanainen, J, Salmenpera, M, Nikki, P (1983) Extubation criteria after weaning from intermittent mandatory ventilation and continuous positive airway pressure. Crit Care Med 11,702-707[ISI][Medline]
  25. Lee, KH, Hui, KP, Chan, TB, et al (1994) Rapid shallow breathing (frequency-tidal volume ratio) did not predict extubation outcome. Chest 105,540-543[Abstract/Free Full Text]
  26. Khan, N, Brown, A, Venkataraman, ST (1996) Predictors of extubation success and failure in mechanically ventilated infants and children. Crit Care Med 24,1568-1579[CrossRef][ISI][Medline]
  27. Krieger, BP, Isber, J, Breitenbucher, A, et al (1997) Serial measurements of the rapid-shallow-breathing index as a predictor of weaning outcome in elderly medical patients. Chest 112,1029-1034[Abstract/Free Full Text]
  28. Afessa, B, Hogans, L, Murphy, R (1999) Predicting 3-day and 7-day outcomes of weaning from mechanical ventilation. Chest 116,456-461[Abstract/Free Full Text]
  29. Leitch, EA, Moran, JL, Grealy, B (1996) Weaning and extubation in the intensive care unit: clinical or index-driven approach? Intensive Care Med 22,752-759[ISI][Medline]
  30. Mergoni, M, Costa, A, Primavera, S, et al (1996) Assessment of various new predictive parameters of the outcome of mechanical ventilation weaning. Minerva Anestesiol 62,153-164[Medline]
  31. Yang, KL (1993) Inspiratory pressure/maximal inspiratory pressure ratio: a predictive index of weaning outcome. Intensive Care Med 19,204-208[CrossRef][ISI][Medline]
  32. Gologorskii, VA, Gelfand, BR, Stamov, VI, et al (1997) Cessation of prolonged artificial ventilation of the lungs and transition to spontaneous respiration of surgical patients. Anesteziol Reanimatol 00,4-10
  33. Kline, JL, Zimnicki, GL, Antonenko, DR, et al (1987) The use of calculated relative inspiratory effort as a predictor of outcome in mechanical ventilation weaning trials. Respir Care 32,870-876
  34. Ochiai, R, Shimada, M, Takeda, J, et al (1993) Contribution of rib cage and abdominal movement to ventilation for successful weaning from mechanical ventilation. Acta Anaesthesiol Scand 37,131-136[ISI][Medline]
  35. DeHaven, CB, Kirton, OC, Morgan, JP, et al (1996) Breathing measurement reduces false-negative classification of tachypneic preextubation trial failures. Crit Care Med 24,976-980[CrossRef][ISI][Medline]
  36. Epstein, SK (1995) Etiology of extubation failure and the predictive value of the rapid shallow breathing index. Am J Respir Crit Care Med 152,545-549[Abstract]
  37. Epstein, SK, Ciubotaru, RL (1996) Influence of gender and endotracheal tube size on preextubation breathing pattern [published erratum appears in Am J Respir Crit Care Med 1996; 155: 2115] Am J Respir Crit Care Med 154,1647-1652[Abstract]
  38. Baumeister, BL, el-Khatib, M, Smith, PG, et al (1997) Evaluation of predictors of weaning from mechanical ventilation in pediatric patients. Pediatr Pulmonol 24,344-352[CrossRef][ISI][Medline]
  39. el-Khatib, MF, Baumeister, B, Smith, PG, et al (1996) Inspiratory pressure/maximal inspiratory pressure: does it predict successful extubation in critically ill infants and children? Intensive Care Med 22,264-268[CrossRef][ISI][Medline]
  40. Fisher, MM, Raper, RF (1992) The "cuff-leak" test for extubation. Anaesthesia 47,10-12[ISI][Medline]
  41. Adderley, RJ, Mullins, GC (1987) When to extubate the croup patient: the "leak" test. Can J Anaesth 34,304-306[Abstract/Free Full Text]
  42. Ely, EW, Baker, AM, Dunagan, DP, et al (1996) Effect on the duration of mechanical ventilation of identifying patients capable of breathing spontaneously. N Engl J Med 335,1864-1869[Abstract/Free Full Text]
  43. Krieger, BP, Ershowsky, PF, Becker, DA, et al (1989) Evaluation of conventional criteria for predicting successful weaning from mechanical ventilatory support in elderly patients. Crit Care Med 17,858-861[ISI][Medline]
  44. Mohsenifar, Z, Hay, A, Hay, J, et al (1993) Gastric intramural pH as a predictor of success or failure in weaning patients from mechanical ventilation. Ann Intern Med 119,794-798[Abstract/Free Full Text]
  45. Capdevila, XJ, Perrigault, PF, Perey, PJ, et al (1995) Occlusion pressure and its ratio to maximum inspiratory pressure are useful predictors for successful extubation following T-piece weaning trial. Chest 108,482-489[Abstract/Free Full Text]
  46. Farias, JA, Alia, I, Esteban, A, et al (1998) Weaning from mechanical ventilation in pediatric intensive care patients. Intensive Care Med 24,1070-1075[CrossRef][ISI][Medline]
  47. Vallverdu, I, Calaf, N, Subirana, M, et al (1998) Clinical characteristics, respiratory functional parameters, and outcome of a two-hour T-piece trial in patients weaning from mechanical ventilation. Am J Respir Crit Care Med 158,1855-1862[Abstract/Free Full Text]
  48. Ashutosh, K, Lee, H, Mohan, CK, et al (1992) Prediction criteria for successful weaning from respiratory support: statistical and connectionist analyses [published erratum appears in Crit Care Med 1994; 22:183] Crit Care Med 20,1295-1301[ISI][Medline]
  49. Sahn, SA, Lakshminarayan, S (1973) Bedside criteria for discontinuation of mechanical ventilation. Chest 63,1002-1005[Abstract/Free Full Text]
  50. Jacob, B, Chatila, W, Manthous, CA (1997) The unassisted respiratory rate/tidal volume ratio accurately predicts weaning outcome in postoperative patients. Crit Care Med 25,253-257[CrossRef][ISI][Medline]
  51. Gandia, F, Blanco, J (1992) Evaluation of indexes predicting the outcome of ventilator weaning and value of adding supplemental inspiratory load. Intensive Care Med 18,327-333[CrossRef][ISI][Medline]
  52. Yang, KL, Tobin, MJ (1991) A prospective study of indexes predicting the outcome of trials of weaning from mechanical ventilation. N Engl J Med 324,1445-1450[Abstract]
  53. Hilbert, G, Gruson, D, Portel, L, et al (1998) Airway occlusion pressure at 0.1 s (P0.1) after extubation: an early indicator of postextubation hypercapnic respiratory insufficiency Intensive Care Med 24,1277-1282[CrossRef][ISI][Medline]
  54. Bouachour, G, Guiraud, MP, Gouello, JP, et al (1996) Gastric intramucosal pH: an indicator of weaning outcome from mechanical ventilation in COPD patients. Eur Respir J 9,1868-1873[Abstract]
  55. Jubran, A, Tobin, MJ (1997) Pathophysiologic basis of acute respiratory distress in patients who fail a trial of weaning from mechanical ventilation. Am J Respir Crit Care Med 155,906-915
  56. Menzies, R, Gibbons, W, Goldberg, P (1989) Determinants of weaning