Prediction Of Weaning

Management of the difficult-to-wean patient is one of the most challenging problems in critical care medicine (2, 3). Overall, approximately 40% of the time that a patient spends receiving mechanical ventilation is devoted to the weaning process, and in certain disease states, such as COPD, the weaning process accounts for approximately 60% of ventilator time (162). Deciding the best time to initiate weaning is especially important. If weaning onset is postponed, the patient is placed at an increased risk of life-threatening, ventilator-induced complications. if weaning is commenced prematurely, the patient may suffer severe cardiopulmonary or psychological decompensation, which sets the patient back in his or her clinical course. Although careful clinical assessment is necessary in deciding when to discontinue mechanical ventilation, this alone is not sufficient, as demonstrated by Stroetz and Hubmayr (163). They studied 31 patients being weaned by gradual reductions in the level of pressure support. The physician in charge of each patient was asked to predict a patient's ability to sustain unassisted breathing without distress for 1 hour. of 22 patients whom the physicians thought likely to fail a weaning trial, 11 were successfully weaned; of nine patients thought likely to be successfully weaned, three failed the trial. This study provides clear evidence of the need to employ objective tests, i.e., predictive indices, when deciding if a patient can tolerate the discontinuation of mechanical ventilation.

Several tests have been described to guide the timing and pace of the weaning process. Because an imbalance between respiratory mechanical load and respiratory muscle capability appears to be the major cause of weaning failure (18, 19), many of the weaning indices directly or indirectly relate to respiratory muscle function. Before assessing respiratory muscle function in a ventilator-supported patient, it is imperative to show that the patient is not likely to develop hazardous hy-poxemia. Weaning should not be contemplated in a patient with a Pao2 of less than 55 mm Hg at an inspired oxygen fraction (FiO2) of 0.40 or higher, because weaning failure is very likely in such patients. However, many patients with satisfactory oxygenation fail weaning attempts, and, except in the patient with marked hypoxemia, indices of oxygenation, such as the PaO2/FiO2 ratio, are unreliable predictors of weaning outcome (164, 165).

Maximum inspiratory pressure, a measurement of inspira-tory muscle strength (see Volitional Tests of Respiratory Muscle Strength in Section 2 of this Statement), is a standard weaning index. In an early study, Sahn and Lakshminarayan (61) found that all patients with a Pi,max value more negative than —30 cm H2O were successfully weaned while all patients with a Pi,max less negative than —20 cm H2O failed a weaning trial. Subsequent investigators have not found Pi,max to provide such clear separation between weaning success and weaning failure patients (13, 18, 106, 164-169). Studies with accessible data on the accuracy of Pi,max are summarized in Table 1. Because the studies differ in design (prospective, retrospective), method of performing Pi,max (best of three attempts, sustained occlusion for 20 seconds), method of weaning (trials of spontaneous breathing, intermittent mandatory ventilation, pressure support), and definitions of weaning success and failure, it is not surprising that accuracy of Pi,max varied considerably among the studies. Nevertheless, it is possible to arrive at some general conclusions. Sensitivity was approximately 0.80, meaning that approximately 80% of patients who succeeded in a weaning trial had a Pi,max value that predicted success (i.e., more negative than —30 cm H2O). However, specificity was approximately 0.25, meaning that only a minority (25%) of patients who failed a weaning trial had a Pi,max that predicted weaning failure (i.e., less negative than —30 cm H2o). Moreover, the ability to predict outcome was not improved by employing a standardized method of measuring Pi,max (18, 106, 165). Based on these data, Pi,max measurements appear to be more helpful in understanding the reason why a particular patient failed a weaning trial rather than in deciding whether to attempt a weaning trial (see additional discussion of Pi,max in this section).

A minute ventilation of less than 10 L/minute, indicating acceptable ventilatory requirements, is another standard weaning index (61). However, in most studies, minute ventilation has been shown to be little better than chance in predicting outcome (165, 166). More helpful is the partitioning of minute ventilation into its Vt and fr components, so as to formulate the ratio of fr to Vt, which can be used to quantitate the extent of rapid shallow breathing. The fr/Vt ratio, measured with a simple handheld spirometer attached to the endotra-cheal tube while the patient spontaneously breathes room air for 1 minute, has been proposed as a reliable method of predicting weaning outcome (165). In an initial "training data set" in 36 patients, an fr/Vt ratio of 105 breaths/minute/L provided the best separation between patients who were successfully weaned and those in whom weaning failed. This threshold value was then prospectively evaluated in 64 patients who constituted the "validation data set." The positive and negative predictive values were 0.78 and 0.95, respectively (165). Analyzing the data with receiver-operating-characteristic (ROC) curves, the area under the curve for the fr/Vt ratio (0.89) was the highest of 10 weaning indices evaluated in the study. Of note, the area under the ROC curves of conventional weaning indices such as minute ventilation (0.40), the PaO2/Pao2 ratio (0.48), and Pi,max (0.61) were not significantly greater than that of an arbitrary test that is expected a priori to have no discriminating value (Figure 7). Jaechke and coworkers (170) reanalyzed the data of Yang and Tobin (165) in terms of likelihood ratios (171). The likelihood ratio for an fr/ Vt ratio of less than 80 breaths/minute/L is 7.53 (Table 2). In other words, an fr/Vt of less than 80 is 7.53 times more likely to occur in a patient who is subsequently successfully weaned than to occur in a patient who will fail a weaning trial; a likelihood ratio of 5-10 is considered to generate moderate and usually useful shifts in pretest to posttest probability (171). Conversely, an fr/Vt ratio greater than 100 is only 0.04 as likely to occur in a patient who will be successfully weaned as in a patient who subsequently fails a weaning trial; a likelihood ratio of < 0.10 is considered to generate large and often conclusive changes in the probability of a given diagnosis (in this case weaning failure).

In patients being weaned from mechanical ventilation, Sas-soon and Mahutte (168) found that the fr/Vt ratio had a sensitivity of 0.97 and a specificity of 0.40. The product of fr/Vt and airway occlusion pressure (P0.1) had the same sensitivity, but specificity increased to 0.60. Disappointingly, the combination of fr/Vt and P0.1 did not have a higher area under the ROC curve than that achieved by fr/Vt alone.

Epstein (172) studied 84 patients with an fr/Vt ratio < 100 breaths/minute/L and observed a positive-predictive value of 0.83. Fourteen patients required reintubation, and all but one of these had developed a new problem such as upper airway obstruction or congestive heart failure. Ten patients with an fr/Vt > 100 were extubated, and four required reintubation (negative predictive value 0.4).

Chatila and coworkers (169) measured fr/Vt with a handheld spirometer in 100 medical-cardiac patients being weaned from mechanical ventilation. The fr/Vt ratio was measured during the first minute of spontaneous breathing, and the patients were then weaned using T-tube trials, CPAP, or pressure-support ventilation. After 30-60 minutes, the fr/Vt ratio was again measured. The area under the ROC curve was higher for the fr/Vt ratio measured at 30-60 minutes than during the first minute of spontaneous breathing, 0.92 ± 0.03 and 0.74 ± 0.05, respectively. Interestingly, the area under the ROC curve for 30-60 minutes is similar to that reported by Yang and Tobin (0.89) (165). This study by Chatila and co-workers (169) emphasizes the importance of not measuring the fr/Vt ratio during the first minute of spontaneous breathing, when respiratory drive may still be suppressed. It does not necessarily mean that one needs to wait 30-60 minutes to reach a steady-state value that truly represents the patient's clinical status.

In a study of 218 extubated patients in a medical ICU, Epstein and Ciubotaru (173) showed that women have a higher fr/Vt ratio than men, which could not be explained by body size, and that fr/Vt was further increased in women with a narrow endotracheal tube. Consequently, the false-negative rate for fr/Vt in women with a narrow endotracheal tube (« 7 mm internal diameter) was especially high. When comparing data from different studies of weaning indices, the pretest probability of a successful outcome becomes very important. Epstein and Ciubotaru (173) deliberately selected patients with a very high pretest probability of successful outcome (0.84), because every patient tolerated a weaning trial and was extu-bated; in contrast, outcome was in considerable doubt (pretest probability 0.56) in the patients studied by Yang and Tobin (165), who were trying to identify the earliest point in time that a patient could resume spontaneous ventilation. For identical false-negative (0.03) and true-negative rates (0.64) (166), the posttest probability of a successful outcome for fr/Vt > 100 breaths/minute/L, calculated on the basis of Bayes' theorem, is 20% for the patients of Epstein and Ciubotaru (173) versus 5% for the patients of Yang and Tobin (165). In other words, fr/Vt is less helpful in cases when the physician strongly suspects that the patient can tolerate weaning and ex-tubation than when the physician is very doubtful about a patient's outcome.

Recently, Ely and coworkers (129) investigated whether the combination of predictive indices followed by a trial of spontaneous breathing would hasten the pace of weaning. Over a 9-month period, ventilator-supported patients in their medical ICU and coronary care units were screened each morning for five factors: PaO2/Fio2 ratio > 200; PEEP « 5 cm H2O; fr/Vt « 105 breaths/minute/L; intact cough on suction-ing; and absence of infusions of sedative or vasopressor agents.

TABLE 1. ACCURACY OF MAXIMAL INSPIRATORY PRESSURE AS A PREDICTOR OF WEANING OUTCOME

P imax Threshold (cm H2O)

Sensitivity

Specificity

Positive Predictive Value

Negative Predictive Value

Reference Number

« -30

0.68

0

0.56

0

166

« -30

1.00

0

0.67

0

167

« -20

NA

NA

0.91

0.22

164

« -30

NA

NA

0.92

0.21

164

« -15

1.00

0.11

0.59

1.00

165

« -20

1.00

0.14

0.60

1.00

165

« -30

0.86

0.21

0.58

0.55

165

« -20

0.91

0.30

0.82

0.55

168

« -20

0.90

0.26

0.67

0.60

169

« -30

0.67

0.69

0.78

0.55

169

Definition of abbreviations: NA = not available; Pi,max = maximum Inspiratory pressure.

Definition of abbreviations: NA = not available; Pi,max = maximum Inspiratory pressure.

False Positive Rate False Positive Rate

False Positive Rate False Positive Rate

Figure 7. Receiver operating characteristic (ROC) curves for frequency/tidal volume (f/Vr) ratio, CROP index (an index that integrates measurements of dynamic compliance, respiratory rate, arterial oxygenation, and maximal inspiratory pressure), maximal inspiratory pressure (Pi,max), and minute ventilation (Ve) in 36 patients who were successfully weaned and 28 patients who failed a weaning trial. The ROC curve is generated by plotting the proportion of true positive results against the proportion of false positive results for each value of a test. The curve for an arbitrary test that is expected a priori to have no discriminatory value appears as a diagonal line, whereas a useful test has an ROC curve that rises rapidly and reaches a plateau. The area under the curve (shaded) is expressed (boxed) as a proportion of the total area. Reprinted by permission from Reference 165.

Patients in the intervention group (n = 149) who met all five criteria underwent a 2-hour trial of spontaneous breathing that same morning. If the patient did not develop clinical signs of distress, employing objective criteria developed in earlier studies of weaning techniques (127, 128), the trial was considered successful, and the patient's physician was notified of this result. The control group (n = 151) had daily screening but did not undergo the spontaneous breathing trial. Although patients in the intervention group had more severe disease, with higher APACHE II and acute-lung-injury scores, their median duration of mechanical ventilation was 1.5 days less than the control group (p = 0.003), and they had lower rates of complications (p = 0.001) and reintubation (p = 0.04) and lower ICu charges (p = 0.03). This study demonstrates that the systematic use of weaning indices results in better patient outcomes than reliance on the clinical judgment of the attending physician.

In the original description of the fr/Vt ratio, the measurement was obtained with a hand-held spirometer while the patient was disconnected from the ventilator and breathed room air for 1 minute (165). Some investigators have measured fr/ Vt during partial ventilator assistance, such as pressure support. For example, Lee and coworkers (174) reported that eight of nine patients who failed a weaning trial had an fr/Vt < 105 breaths/minute/L, and they concluded that fr/Vt was unhelpful. Because pressure support is well known to decrease fr and increase Vt, it is not surprising that the fr/Vt threshold value developed during unassisted breathing would not apply during pressure support ventilation.

Finally, the definition of the outcome event is very important in assessing the accuracy of a diagnostic test. Accumulating data suggest that the fr/Vt ratio is more helpful in predicting the earliest point in time that a patient can breathe spontaneously without assistance from a ventilator than in predicting the need for reintubation after the patient has been extubated (129, 172, 173, 175). That is, fr/Vt appears to be more helpful in deciding when to initiate the weaning process, rather than deciding when to terminate it and extubate the patient.

0 0

Post a comment