1999 Donald F Egan Scientific Lecture

Weaning from Mechanical Ventilation: What Have We Learned?

Martin J Tobin MD

It is a pleasure and honor for me to give the 26th Annual tients fail weaning trials, we need to delve into the under-
Donald F Egan Scientific Lecture at the American Asso- lying pathophysiological mechanisms. Four anatomical
ciation for Respiratory Care. Much of my research career sites or functions may be involved: respiratory centers,
has focused on weaning from mechanical ventilation, and respiratory muscles, lung mechanics, and gas exchange
a major stimulus for my interest in this subject was an function of the lung.4 I’ll discuss data pertaining to each
outstanding review article published in RESPIRATORY CARE site and indicate how each is important in understanding
by its current editor, Dave Pierson, in the early 1980s.1 In why patients fail weaning trials.
that article, Dave made the subject of weaning exciting. We begin with the respiratory centers. A depressed re-
More importantly, he pointed out many areas about which spiratory center drive at the start of the weaning trial will
we knew nothing. To someone starting an academic ca- cause hypoventilation, making weaning failure inevitable.
reer, the subject of weaning appeared particularly ripe for Another possibility is for the drive to be normal at the start
research.
of the trial, but then to fall during the course of the trial.
While preparing for this morning’s lecture, I took my
It’s been suggested that it would be clever for the body to
copy of Egan’s Fundamentals of Respiratory Therapy from
decrease respiratory center output as a way of avoiding
the shelf. This was the first textbook directed primarily at
contractile fatigue of the respiratory muscles. Such a strat-
respiratory therapists. It was first published in 1969, and
the copy I own is the third edition, published in 1977.2 In egy has even been called “central wisdom.”5 Figure 1
the 1977 edition, Dr Egan wrote that “The separation of a shows the total pressure generated by inspiratory muscles,
patient from his ventilator is very nearly pure art.” Wean- expressed as pressure-time product, measured by Amal
ing is still an art in 1999. But over the next half hour, I Jubran in 17 patients who failed a weaning trial.6 All but
hope I can show you that the approach to weaning has a one patient showed an increase in pressure generation be-
more scientific basis than was the case 20 years ago. tween the beginning and end of the T-tube trial. As such,
downregulation of respiratory motor output is not common
Pathophysiology of Weaning Failure in patients who fail a trial of weaning.
Next, we move to the respiratory muscles. We used to
A patient failing a weaning trial exhibits the physical think that maximum inspiratory pressure, which reflects
signs of respiratory distress. We see heightened activity of inspiratory muscle strength, was helpful in predicting which
the sternomastoid muscles, recession of the suprasternal patients could come off the ventilator. In 100 patients
fossa, recession of the intercostal spaces, paradoxical mo- undergoing a weaning trial, we found no difference in
tion of the abdomen, tachypnea, and sometimes cyanosis.3 maximum inspiratory pressure between weaning success
These physical signs tell us the patient is not able to sus- and weaning failure patients.7 As such, respiratory muscle
tain spontaneous ventilation. But to understand why pa- weakness doesn’t appear to be a common cause for failure
to wean. Might the respiratory muscles deteriorate be-
tween the beginning and end of a weaning trial? Yes, if
Martin J Tobin MD is affiliated with the Division of Pulmonary and
Critical Care Medicine, Loyola University of Chicago Stritch School of they develop respiratory muscle fatigue.8
Medicine and Hines Veterans Affairs Hospital, Maywood, Illinois. Is it important to know whether these patients develop
muscle fatigue? It’s extremely important. Darlene Reid9
This article is based on a transcript of the 26th Annual Donald F Egan
Scientific Lecture delivered by Martin J Tobin MD at the 45th Interna-
has demonstrated electron-microscopic evidence of severe
tional Respiratory Congress of the American Association for Respiratory muscle destruction in hamsters who developed diaphrag-
Care, Las Vegas, Nevada, December 14, 1999. matic fatigue (Fig. 2). The same process may happen in
Correspondence: Martin J Tobin MD, Division of Pulmonary and Critical
weaning failure patients. Patients who fail a weaning trial
Care Medicine, Loyola University Medical Center, 2160 South First already have problems before they commence the trial.
Avenue, Maywood IL 60153. Then, as they fail the trial, they may be developing a new

RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4 417

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

This patient would have developed the type of structural

injury I showed you in the hamster model. Of course, the
data are from only a single patient. Franco is now studying
a larger group of patients to determine the frequency of
fatigue in patients undergoing weaning trials.
The data in Figures 5 and 6 show the stress on the
respiratory muscles in weaning failure patients. This stress
is related to the work the muscles perform. The informa-
tion in Figure 5 represents a huge amount of compressed
data.6 The tracings are ensemble averages of several breaths
from each individual patient. The failure patients had much
larger swings in esophageal pressure by the end of the trial
than at the beginning. Also, the swings in pressure were
much greater in the failure patients than in the success
patients.
Why is work of breathing (WOB) increased in patients
who fail a weaning trial? To answer this question, Amal
Jubran measured inspiratory resistance, dynamic elastance
(which is the inverse of compliance), and auto positive
Fig. 1. Values of inspiratory pressure-time product at the start and end-expiratory pressure (auto-PEEP) (see Figure 6).6 She
end of an unsuccessful trial of weaning in 17 patients with chronic found that the values for each variable were much higher
obstructive pulmonary disease. All but one patient showed an in failure patients than in success patients over the course
increase in pressure generation between the onset and end of the of a trial. Each variable also deteriorated over time in the
trial. (Based on data from Reference 6.)
failure patients. That is, patients who fail a weaning trial
display a progressive worsening of their pulmonary me-
separate problem—that is, structural damage resulting from chanics, resulting in large increases in their WOB.
contractile muscle fatigue. Might the pulmonary mechanics be more severely de-
To determine whether muscle fatigue is likely in such ranged in the failure patients even before they come off the
patients, Amal Jubran measured the tension-time index of ventilator? Could you tell, on the basis of mechanics, that
the inspiratory muscles (Figure 3).6 Tension-time index is weaning failure is going to be inevitable? To address this
the product of two fractions: the mean pressure per breath question, Amal Jubran looked at passive lung mechanics
over maximum inspiratory pressure, and the time of inspi- before taking patients off the ventilator.10 Measurements
ration over total respiratory cycle time. She made mea- of airway pressure, transpulmonary pressure, and esopha-
surements at the start of a T-piece trial and at its end, about geal pressure, combined with the end-inspiratory occlu-
45 minutes later. None of the weaning success patients sion method, allow you to respectively characterize the
developed a tension-time index above 0.15—the value that overall respiratory system, the lung itself, and the chest
has been linked with muscle fatigue.8 Five of the failure wall. You can also divide respiratory resistance into the
patients, however, had a tension-time index of 0.15 or component resulting from ohmic resistance, reflecting air-
higher by the end of the trial. This observation suggests way resistance, and the component arising from stress in-
that these five patients may have developed inspiratory homogeneities in the system, consequent to pendelluft and
muscle fatigue. Tension-time index is an indirect index, viscoelastic forces.
and it doesn’t provide concrete evidence that fatigue ac- Before performing the T-piece trials, she passively ven-
tually occurred. tilated the patients. Respiratory system resistance was
To convincingly detect fatigue, you need to stimulate equivalent in the weaning success and weaning failure
the phrenic nerves and measure the contractile response of patients (Figure 7).10 Moreover, partitioning of resistance
the diaphragm.8 Figure 4 shows measurements obtained into the components reflecting airway resistance and stress
by Franco Laghi in a patient who failed a weaning trial. At inhomogeneity revealed no difference between the groups.
the start of the trial, the patient’s twitch transdiaphrag- That the respiratory mechanics were similar in the two
matic pressure was around 30 cm H2O, which is normal. groups before the start of a trial implies that something in
The patient then underwent a T-piece trial lasting a half the act of spontaneous breathing causes the weaning fail-
hour. Franco repeated the measurements 15 min after com- ure patients to deteriorate over the course of the trial. We
pleting the trial, and again at 30 and 60 min. The twitch can speculate about mechanisms by which spontaneous
pressures fell considerably compared with baseline. These breathing could worsen respiratory mechanics, but to find
data provide conclusive evidence of diaphragmatic fatigue. the real reason we need further research.

418 RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

Fig. 2. Electron micrograph of the diaphragm in a control hamster (upper) and in a

hamster that had breathed through a resistive load for 6 days (lower). Loading was
achieved by tightening a polyvinyl band around the trachea until swings in esophageal
pressure were ⬃20% of maximal inspiratory pressure; pulmonary resistance was in-
creased 6.5 fold. Compared with the normal structure, the loaded animals developed
sarcomere disruption with loss of distinct A bands and I bands and development of Z
line streaming. (From Reference 9, with permission.)

The fourth, and final, aspect of the pathophysiology is ure 8). The PCO2 rose because the patients developed rapid
gas exchange. Some patients fail a weaning trial with no shallow breathing—with inevitable increase in dead-space
change in their arterial blood gases, whereas others de- ventilation. Alveolar ventilation went down but overall
velop increases in carbon dioxide tension (PCO2) or de- minute ventilation didn’t change.
creases in oxygen tension. The term hypoventilation is The hypoxemia that occurs in some patients failing a
used synonymously with hypercapnia. But when you see weaning trial is usually associated with an increase in
an increase in PCO2, it doesn’t mean that minute ventilation venous admixture. A further factor contributing to the hy-
has necessarily fallen. In a group of patients failing a poxemia is a decrease in mixed venous oxygen saturation
weaning trial, we found no relationship between PCO2 and (Figure 9).12 The fall in mixed venous oxygen saturation is
minute ventilation.11 Instead, we found that more than partly the result of the considerable cardiovascular de-
80% of the variance in PCO2 could be explained by the mand experienced by weaning failure patients, as first
patients’ tidal volume (VT) and respiratory frequency (Fig- shown by François Lemaire.13 In a classic study, François

RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4 419

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

Fig. 3. The relationship between the ratio of mean esophageal pressure to maximum inspiratory pressure
(Pēs/PImax) and duty cycle (TI/TTOT) in 17 ventilator-supported patients with chronic obstructive pulmonary
disease who failed a trial of spontaneous breathing and 14 patients who tolerated the trial. Circles and
triangles represent values at the start and end of the trial, respectively; closed symbols indicate patients who
developed an increase in PaCO2 during the trial. Five of the 17 patients in the failure group developed a
tension-time index of ⬎ 0.15 (indicated by the isopleth), suggesting respiratory muscle fatigue. N represents
the value in a normal subject. (From Reference 6, with permission.)

Fig. 4. Recordings of transdiaphragmatic twitch pressure (Pdi) in a patient with a C4 spinal cord injury
10 min before a trial of spontaneous breathing and at several intervals after the end of the failed trial
that lasted 30 min. The nadir in twitch pressure was reached 30 min after the end of the trial, and at 60
min twitch pressure was still less than that recorded 10 min before the trial. This finding indicates the
development of contractile muscle fatigue.

420 RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

Fig. 5. Ensemble average plots of flow and esophageal pressure (Pes) at the start and end of a trial of spontaneous breathing in 17
ventilator-dependent patients with chronic obstructive pulmonary disease who failed the trial and 14 patients who tolerated the trial
and were extubated. At the start of the trial, the inspiratory excursion in Pes was greater in the failure group, and it showed a further
increase by the end of the trial. To generate these plots, flow and Pes tracings were divided into 25 equal time intervals over a single
respiratory cycle for each of the 5 breaths for each patient in the two groups. For a given patient, the 5 breaths from the start of the
trial were then superimposed and aligned with respect to time, and the average at each time point was calculated. The group mean
tracings were then generated by ensemble averaging of the individual mean from each patient. The same procedure was performed
for breaths at the end of the trial. (From Reference 6, with permission.)

showed that weaning failure patients develop an increase off the ventilator. They are useful, however, in our assess-
in their pulmonary artery wedge pressure and left-ventric- ment of the patient who has already failed a trial—to un-
ular end-diastolic volume. The increased stress on the car- derstand why that patient failed.
diovascular system probably resulted from the increased In the past, it was felt that the gestalt of an experienced
WOB. When intrathoracic pressure becomes more nega- clinician at the bedside was better at predicting weaning
tive, the afterload of the left ventricle increases, which, in outcome than physiologic indices. The accuracy of this
turn, makes it more difficult to maintain cardiac output.12,13 gestalt had never been studied until recently. Randy Stro-
etz and Rolf Hubmayr14 asked attending physicians in the
Prediction of Weaning Outcome intensive care units of the Mayo Clinic to predict whether
their patients were likely to succeed in a weaning trial. Of
I will now discuss how good we are at predicting wean- the 31 patients in the study, the physicians predicted that
ing outcome. The predictive indices listed by Dr Egan in 22 would fail the trial. Yet, half of the 22 patients were
1977 are similar to those you see listed today.1,3 One dif- successfully weaned. This doesn’t mean that clinical as-
ference is his inclusion of a dead space-to-VT ratio of less sessment is useless. It remains necessary, but it’s not suf-
than 0.60 as a helpful predictor of weaning success; few ficient. We need something in addition to clinical assess-
people today would recommend this measurement. Over ment.
the last 20 years, we have found that the classic variables, Some people say you can dispense with weaning pre-
such as maximum inspiratory pressure, minute ventilation, dictors completely, and go directly to some weaning
and vital capacity have very high rates of false positives method, such as a T-tube trial or pressure support. But to
and false negatives.3,7 Many indices don’t help in telling use any weaning approach you have to first think of the
us whether or not an individual patient is likely to come possibility that the patient might tolerate it. In the study

RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4 421

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

hand-held spirometer— over one minute.7 The measure-

ments were made while patients were disconnected from
the ventilator and breathing room air. We combined the
measurements into an index of rapid shallow breathing,
the frequency-to-VT ratio. The higher the ratio, the more
severe the rapid shallow breathing, and the greater the
likelihood that the patient would fail a weaning trial. We
tested the accuracy of this index in 100 patients, and found
a ratio of 100 breaths/min per liter gave the best separation
of the groups—a value that’s easy to remember.
One of the best ways of evaluating the accuracy of any
diagnostic test is to use receiver operating characteristic
curves.3 These curves are created by taking multiple val-
ues of a test measurement, and plotting the true positive
rate against the false positive rate (Figure 11). You then
measure the area under the curve, and this tells you the
overall accuracy of the test. A perfect test has an area
under the curve of 1.0. A test that’s no better than chance
has an area under the curve of 0.50. For our patients, the
area under the curve for minute ventilation was 0.40, mean-
Fig. 6. Inspiratory resistance of the lung (Rinsp,L), dynamic lung ing that minute ventilation was worse than flipping a coin
elastance (Edyn,L), and intrinsic positive end-expiratory pressure at the patient’s bedside in predicting weaning outcome.
(PEEPi) in 17 weaning failure patients and 14 weaning success The other classic index, maximum inspiratory pressure,
patients. Data were obtained during the second and last minute of
had an area of 0.61; it’s slightly better than chance in
the trial, and at one third and two thirds of the trial duration. Be-
tween the onset and end of the trial, the failure group developed predicting outcome. The CROP index, which integrates a
increases in Rinsp,L (p ⬍ 0.009), Edyn,L (p ⬍ 0.0001), and PEEPi (p ⬍ number of physiologic variables, was substantially better,
0.0001), and the success group developed increases in Edyn,L (p ⬍ with an area of 0.78. The frequency-to-VT ratio had an
0.006) and PEEPi (p ⬍ 0.02). Over the course of the trial, the failure
area of 0.89. This simple index turned out to be the most
group had higher values of Rinsp,L (p ⬍ 0.003), Edyn,L (p ⬍ 0.006),
and PEEPi (p ⬍ 0.009) than the success group. (From Reference 6, accurate predictor.
with permission.) We answer research questions by making measurements
in groups of patients. But when we leave research and go
back to clinical practice, our focus shifts to a single pa-
from the Mayo Clinic, we see that half the patients that tient. That relationship between one patient and one clini-
physicians thought not ready for weaning actually suc- cian is the soul of clinical medicine. What you really want
ceeded.14 As such, the systematic use of predictors alerts to answer is what’s the likelihood that the patient in front
us to the possibility that some of the patients we think not of you can come off the ventilator? Let’s take a situation
ready for weaning are in fact much better than they appear.
where you have no clue whether a patient is likely to come
From my discussion of the pathophysiology of weaning
off the ventilator. In the language of statisticians, this is a
failure, it’s clear that patients failing a weaning trial ex-
pre-test probability of 50 per cent.15 If you measure the
perience huge stresses on their respiratory muscles and
frequency-to-VT ratio and the value is above 100, you
cardiovascular system. These stresses might damage their
draw a line on Figure 12 between the pre-test probability,
heart or respiratory muscles and also cause considerable
anxiety and distress. The measurement of predictive indi- 50 per cent, and the likelihood ratio, which we know is
ces—provided they are reasonably reliable—avoids sub- 0.04 for a frequency-to-VT ratio above 100.16 Then, you
jecting patients prematurely to such stresses before they continue the line to get the post-test probability, and you
are able to cope with them. find it’s less than 5 per cent. Here you have a patient about
Several years ago, we noted that patients who went on whom you’re in total doubt as to clinical outcome; if you
to fail a weaning trial developed an increase in respiratory find that the frequency-to-VT ratio is above 100, the in-
frequency and a fall in VT as soon as we took them off the formation changes your post-test probability to a less than
ventilator (Figure 10).11 We reasoned that measuring these 5 per cent likelihood that the patient will come off the
changes might be useful in forecasting weaning outcome. ventilator. If the frequency-to-VT ratio is 80, this has a
We subsequently undertook a study, where we mea- likelihood ratio of 7.5,16 which changes the post-test prob-
sured frequency and VT with simple instrumentation—a ability to nearly 95 per cent. This example illustrates the

422 RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

Fig. 7. Maximal resistance (overall column height) of the respiratory system (Rmax,rs), lung (Rmax,L), and
chest wall (Rmax,w) in weaning failure (F) and weaning success (S) patients during passive ventilation;
the clear portions of the columns represent minimum resistance (Rmin) while the shaded portions
represent additional resistance (⌬R). No differences in Rmax,rs, Rmin,rs, or ⌬Rrs were observed between
the groups, nor between the lung and chest wall components. Upward directed bars represent ⫾ SE
(standard error) of Rmin, while downward directed bars represent ⫾ SE of ⌬R. (From Reference 10, with
permission.)

Fig. 8. Relationship between tidal volume (VT) and respiratory frequency with carbon dioxide tension
(PaCO2) in seven patients who failed a spontaneous breathing trial. PaCO2 was significantly correlated
with VT (r ⫽ 0.84, p ⬍ 0.025) and frequency (r ⫽ 0.87, p ⬍ 0.025); 81% of the variance in PaCO2 could
be explained by the changes in these two variables (From Reference 11, with permission.)

RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4 423

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

Fig. 9. Ensemble averages of interpolated values of mixed venous oxygen saturation (Sv៮ O2) during
mechanical ventilation and a trial of spontaneous breathing in patients who succeeded in the trial (open
symbols) and patients who failed the trial (closed symbols). During mechanical ventilation, Sv៮ O2 was
similar in the two groups (p ⫽ 0.28). Between the onset and end of the trial, Sv៮ O2 decreased in the failure
group (p ⬍ 0.01), whereas it remained unchanged in the success group (p ⫽ 0.48). Over the course of
the trial, Sv៮ O2 was lower in the failure group than in the success group (p ⬍ 0.02). Bars represent
standard errors. (From Reference 12, with permission.)

value of combining your clinical judgment, which is your

pre-test probability, with a test, in this case the frequency-
to-VT ratio, and seeing how it alters your post-test prob-
ability.

Weaning Techniques

For the remaining portion of my presentation, I’ll focus

on the different methods used for weaning. We have four
approaches. With pressure support and intermittent man-
datory ventilation (IMV), you decrease the support from
the ventilator and force the patient to undertake more of
the work needed for a given minute ventilation. The third
and oldest approach is to perform T-piece trials several
times a day. Dr Egan described how this approach was
being used in 1977: “Some experts advocate removing the
patient from his ventilator for a fixed short period of time
and gradually shorten the intervals between. As an exam-
ple, this might mean letting the patient breathe unassisted Fig. 10. Breath-by-breath plot of respiratory frequency and tidal
for 2 minutes of an hour, then 2 minutes every half-hour, volume (VT) in a patient who failed a weaning trial. The arrow
indicates the point of resuming spontaneous breathing. Rapid,
quarter-hour, and so on, until mechanical ventilation is shallow breathing developed almost immediately, suggesting the
discontinued.” That approach involves a huge amount of prompt establishment of a new steady state. (From Reference 11,
work for the intensive care unit staff, and it’s not hard to with permission.)

424 RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

Fig. 11. Receiving-operating-characteristic (ROC) curves for frequency-to-tidal volume ratio (f/VT), CROP
index (acronym for compliance, rate, oxygenation, and pressure, which integrates factors associated
with risk of respiratory failure), maximum inspiratory pressure (PImax), and minute ventilation (V̇E) in
weaning success and weaning failure patients. 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 (in box) as a proportion of the total area. (From Reference 7, with permission.)

see why multiple T-piece trials became very unpopular. controlled ventilation in baboons causes a decrease in con-
The fourth approach, and the one I personally prefer, is to tractility of the diaphragm, suggesting the development of
perform a T-piece trial once a day. If patients are breathing muscle atrophy. With mechanical ventilation, we want to
comfortably after a half hour, they’re extubated. Patients achieve rest without causing muscle atrophy. The ideal
who fail go back on the ventilator for at least 24 hours amount of rest needed by a patient has never been studied.
before we make another weaning attempt. Franco Laghi19 addressed this issue in healthy human vol-
Studies from our lab show that WOB is enormous in unteers (Figure 13). He induced muscle fatigue by having
patients who fail a weaning trial.6 A major goal of me- the subjects breathe through a resistive load. He stimulated
chanical ventilation is to decrease this work.17 But if the the phrenic nerves and measured transdiaphragmatic twitch
respiratory muscles get too much rest might they develop pressure (Pdi). The subjects started off with a baseline
atrophy? Antonio Anzueto18 has shown that 11 days of twitch value in the high 30s, which is normal. When the

RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4 425

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

fatigue and it’ll require at least 24 hours to recover from

that.
When introduced, IMV looked like the ideal way to
wean patients from the ventilator. It took into account the
need for rest to avoid muscle fatigue and also the idea that
too much rest might cause atrophy.20 By allowing the pa-
tient to take some spontaneous breaths, IMV should pre-
vent muscle atrophy. By resting the patient during the
mandatory breaths, fatigue should be avoided. The balance
between these two factors makes it theoretically possible
to customize the approach for each individual patient.
Unfortunately, IMV doesn’t work according to plan.
Figure 14 shows measurements in a single patient.21 Ac-
cording to the theory, we’d expect a decrease in diaphrag-
matic and sternomastoid activity during the assisted breaths.
But we can’t tell these tracings from the spontaneous
breaths. That the effort performed by the patient is the
same for the mandatory and spontaneous breaths was first
pointed out by John Marini.22 A patient doesn’t know
whether the ventilator is going to provide assistance on the
next breath. As such, the patient fires his respiratory cen-
ters at the onset of the breath. When the ventilator starts to
assist him, he’s unable to switch off his respiratory cen-
ters. As a result, the effort he performs is the same for the
ventilator breaths as for the spontaneous breaths.
Pressure support is the other commonly used method of
weaning.23 Pressure support was popularized as a means
of overcoming the resistance of the endotracheal tube. The
story goes that if patients are able to breathe comfortably
at that level of pressure support, they should be able to
breathe without difficulty following extubation. The prob-
Fig. 12. The effect of frequency-to-VT ratio measurements on clin-
lem is to figure out what’s the level of pressure support
ical equipoise, ie, a pre-test probability of 50 per cent. A frequen- that overcomes the resistance of the endotracheal tube.
cy-to-VT ratio of ⬎ 100 breaths/min per liter has a likelihood ratio Various levels, such as 6 or 8 cm H2O, have been sug-
of 0.04. The post-test probability is obtained by drawing a line gested.
between the pre-test probability, 50 per cent, and the likelihood The people who proposed the addition of pressure sup-
ratio, 0.04, and then extending the line; this results in a post-test
probability value of less than 5%. A frequency-to-VT ratio of less
port to overcome the resistance of the endotracheal tube
than 80, which is known to have a likelihood ratio of 7.5, results in appear to have forgotten that when a tube is in the airway
a post-test probability approaching 95 per cent. (Modified from for some time it causes inflammation and edema. When
Reference 15.) the tube is removed, the resistance of the upper airway will
be higher than normal. This point was nicely shown by
subjects breathed through a resister, the twitch pressure Christian Strauss.24 He found that WOB in patients fol-
fell to about 25 cm H2O. Several investigators had previ- lowing extubation was virtually identical to what it had
ously shown that twitch pressures fall after resistive load- been while they breathed on a T-piece. Any amount of
ing. What’s new here is that Franco measured the change pressure support causes you to underestimate the work a
in the contractile properties of the diaphragm over the patient will have to perform following extubation—which
subsequent 24 hours.19 He observed some recovery over is what you’re trying to forecast.
the first 8 hours. Between 8 and 24 hours, there was no Another problem arises with pressure support in pa-
further recovery. Compared with baseline, the twitch pres- tients with chronic obstructive pulmonary disease—the
sures at 24 hours were significantly depressed. This ob- most challenging group to wean from the ventilator. This
servation tells us that a considerable period of rest is needed problem relates to the off-cycling of the time of inflation.
for recovery from diaphragmatic fatigue. That’s the reason Mechanical inflation is switched off when inspiratory flow
I prefer performing a T-piece trial just once a day. When falls to some value, such as 25 per cent of the peak value.23
patients fail, I fear that they may have respiratory muscle Patients with chronic obstructive pulmonary disease have

426 RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

Fig. 13. Induction of diaphragmatic fatigue (stippled bar) produced a significant fall in transdiaphrag-
matic twitch pressure (Pdi) elicited by twitch stimulation of both phrenic nerves. Significant recovery
of twitch pressure was noted in the first 8 hours after completion of the fatigue protocol; no further
change was observed between 8 and 24 hours, and the 24-hour value was significantly lower than
baseline. The delay in reaching the nadir of twitch Pdi probably results from twitch potentiation,
induced by repeated contractions, which was present at the end of the protocol. Values are mean ⫾
standard error. * Significant difference compared with baseline value, p ⬍ 0.01. (From Reference 19,
with permission.)

neurons in the brainstem become impatient. They’re say-

ing, “The ventilator is still pumping gas into the lungs, but
we think it’s time to breathe out.” The expiratory neurons
get switched on, and the patient fights the ventilator. This
is not something you want to do in a patient with preex-
isting weaning difficulties.
Amal Jubran investigated this issue in critically ill pa-
tients. The interrupted tracing in Figure 15 represents the
chest wall recoil.26 Halfway during the period of mechan-
ical inflation, we see that esophageal pressure was higher
than the chest wall recoil. This means that the patient had
switched on his expiratory muscles while the ventilator
was still pumping gas into the lungs. The measurements in
Fig. 14. Electromyograms of the diaphragm (EMGdi) and the ster- her study were based on a number of assumptions, partic-
nocleidomastoid muscles (EMGscm) in a patient receiving synchro- ularly the positioning of the chest wall recoil line. One of
nized intermittent mandatory ventilation. Intensity and duration of our fellows, Sai Parthasarathy, readdressed the question by
electrical activity is similar during assisted (A) and spontaneous (S)
breaths. Paw ⫽ airway pressure. Pes ⫽ esophageal pressure. (From
inserting needle electrodes into the transversus abdomi-
Reference 21, with permission.) nis—the major muscle of expiration.27 Again, about half-
way during the period of mechanical inflation he found
increases in resistance and compliance. The product of that the patients recruited their abdominal muscles. This
these two variables is the time constant of the respiratory problem with pressure support arises because of the algo-
system.25 An increase in the time constant means it’ll take rithm used for cycling off the inflation phase. As a result,
longer for air to move in and out of the bronchi. Specifi- most patients with chronic obstructive pulmonary disease
cally, it’ll take longer for flow to drop from its peak down receiving a high level of pressure support will be forced to
to 25 per cent of that value. As a result, the expiratory fight the ventilator.

RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4 427

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

Fig. 15. Esophageal pressure (continuous line) in a patient with chronic obstructive pulmonary disease
receiving pressure support of 20 cm H2O. The interrupted line represents the estimated recoil pressure
of the chest wall. The tracings have been superimposed so that chest wall recoil pressure is equal to
esophageal pressure at the onset of the rapid fall in esophageal pressure in late expiration (right of
figure). Times at which esophageal pressure is higher than chest wall pressure signify a minimal
estimate (lower bound) of expiratory effort. The expiratory muscles become active about halfway
during the period of mechanical inflation. (From Reference 26, with permission.)

In patients who fail a weaning trial, we want to rest their level of ventilator support. In recent randomized, controlled
respiratory muscles. Clinicians often assume that simply trials of weaning techniques, however, 70 to 80 per cent of
connecting a patient to a ventilator is sufficient to achieve patients tolerated their first T-piece trial.29,30 Patients went
rest. But patients can have difficulty even in triggering the from full ventilator support, consisting of assist-control
machine (Figure 16). One of our fellows, Phil Leung, found ventilation, to a T-piece trial, without a gradual decrease in
that up to 30 per cent of attempts made by patients fail to the level of support.
trigger the ventilator.28 Why do patients have difficulties A major milestone in weaning research was the first
in triggering? To understand this phenomenon, Phil looked randomized, controlled trial carried out by Laurent Bro-
at the characteristics of the breaths that immediately pre- chard.29 He compared three different methods: IMV, T-
ceded the triggering and nontriggering attempts. The breaths pieces, and pressure support. Before this study, most com-
before nontriggering attempts had a higher VT and a lower mentators said it really didn’t matter what technique you
expiratory time. When you inhale a large VT, the elastic used for weaning—that they’re all the same. Laurent
recoil pressure at the peak of inspiration will be high. If the showed it clearly matters. For the first time, he showed
time for exhalation is also shorter, the pressure in your that one technique, IMV, was markedly inferior to the
system—the elastic recoil pressure—will be above normal other weaning approaches. People often misinterpret the
when you finish trying to exhale. We quantify this pres- results of Laurent’s study, and say that he showed that
sure in terms of auto-PEEP. And Phil found that auto- pressure support was better than T-pieces. Pressure sup-
PEEP was higher before attempts that failed to trigger the port was better than the combination of the T-piece group
ventilator than for the attempts that triggered the machine. and the IMV group. There was no difference between
That is, the real trigger sensitivity—not the set sensitivi- pressure support and T-pieces, when the T-piece group
ty—is much higher in patients who fail to trigger the ma- was analyzed separately from the IMV group.
chine. The following year we published a randomized con-
A problem in talking about the subject of weaning is the trolled trial conducted with collaborators in Spain.30 We
word itself. “Weaning” implies a gradual reduction in the looked at the four approaches I mentioned earlier: single

428 RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

Fig. 16. Recordings of tidal volume, flow, airway pressure (Paw), and esophageal pressure (Pes) in a
patient with chronic obstructive pulmonary disease receiving pressure support ventilation. Approxi-
mately half of the patient’s inspiratory efforts do not succeed in triggering the ventilator. Triggering
occurred only when the patient generated a Pes more negative than – 8 cm H2O (indicated by the
interrupted horizontal line), which was equal in magnitude to the opposing elastic recoil pressure. Each
ineffective triggering attempt is signalled by a braking of expiratory flow, whereby flow returns to zero
due to the action of the inspiratory muscles. Thus, monitoring of expiratory flow provides a more
accurate measurement of the patient’s intrinsic respiratory rate than the number of machine cycles
displayed on the bedside monitor. (From Reference 4, with permission.)

daily trials of spontaneous breathing, multiple trials of trials.30 He studied 300 patients who underwent a daily
spontaneous breathing, pressure support, and IMV. Like screen by respiratory therapists. The daily screen consisted
Laurent Brochard, we found that IMV had the worse out- of looking at the patient’s oxygenation, the level of PEEP,
come. Using a Cox proportional-hazards regression model, the absence of rapid, shallow breathing (a frequency-to-VT
we found that the single daily trial of spontaneous breath- ratio of less than 105),7 the presence of a good cough on
ing resulted in a three-fold increase in the rate of success- suctioning, and lack of infusions of pressors or sedatives.
ful weaning compared with IMV, and a two-fold increase Patients passing the screen were randomized to an inter-
in the rate of successful weaning compared with pressure vention group and a control group. The control group was
support. managed in the usual manner by the attending physicians,
Wes Ely31 subsequently undertook a study that com- largely consisting of pressure support or IMV. Patients in
bined two aspects of our previous research: the use of the intervention group underwent a two-hour trial of spon-
weaning predictors7 and the use of spontaneous breathing taneous breathing,30 without getting permission from the

RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4 429

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

attending physician. The attending physicians of patients is likely that many of them would have required reintuba-
passing the two-hour trial were contacted verbally and a tion. To properly answer the question, you’d need to take
note to that effect was also written in the chart. a group of patients, measure the predictive indices, and
Although the patients in the intervention group were then extubate every patient irrespective of whether or not
sicker, with higher acute physiology and chronic health they tolerated a weaning trial.
evaluation and lung injury scores, they were weaned twice Reintubation represents a major new frontier for re-
as fast as the control group. That is, a two-step strategy, search. We need to find out what exactly is going on in
consisting of the systematic measurement of weaning pre- these patients. To date, we don’t have a single study prob-
dictors7 combined with a spontaneous breathing trial,30 ing the pathophysiology of reintubation.
achieved a better outcome. Looking at the details, 59 per In summary, the major reason that patients fail weaning
cent of the patients tolerated the trial. In general, about 10 trials is their enormous respiratory work load. We’re still
to 15 per cent of extubated patients require reintubation. If unsure whether these patients develop respiratory muscle
the investigators had been aggressive and extubated every fatigue. We need to answer this question because it has
patient who passed the spontaneous breathing trial, you’d major implications for patient management. In deciding
expect about 50 per cent of patients to have tolerated ex- the right time to take a patient off the ventilator, we’ve
tubation. In contrast, 32 per cent were actually extubated. learned that the judgment of an experienced clinician is
Despite this nonaggressive approach, the rate of successful not enough. You need weaning predictors. And when
extubation was more than double that in the control group. they’re measured systematically, predictors result in more
In early 1999, we published a study conducted with effective management. Of the weaning techniques avail-
collaborators in Spain to determine if patient outcome was able, a number of randomized, controlled trials have shown
different for a spontaneous breathing trial lasting a half that one of the most ingrained approaches, IMV, is the
hour versus two hours.32 To emphasize how thinking has least effective. A single daily trial of spontaneous breath-
changed about the right length for a T-piece trial, I refer to ing appears to be the most expeditious weaning technique.
what Dr Egan wrote in 1977: “When the patient can breathe When looking at the story of weaning research over the
unassisted around the clock, and is moving a reasonable last 20 years, one is reminded of the saying of the French
amount of air without undue effort, and can walk for short essayist, Michel de Montaigne:
distances consistent with his general physical condition,
and when ventilation is satisfactory and stable by blood Whenever a new discovery is reported to the sci-
gas values, it is time to consider removal of the endotra- entific world, they say first, “it is probably not true.”
Thereafter, when . . . demonstrated beyond ques-
cheal tube.” When we work in a field, we often don’t
tion, they say “yes, it may be true, but it is not
notice how much it advances. In another 20 years, I expect important.” Finally, when a sufficient time has
people will think some of my statements today as strange— elapsed, they say “Yes, surely it is important, but it
probably much sooner than 20 years! Returning to our is no longer new.
recent study, patient outcome was the same for spontane-
ous breathing trials lasting for two hours or a half hour.32
Contrasted with the previous recommendation that T-piece That statement was made more than 400 years ago—plus
trials should last 24 hours, being able to make a decision ça change, plus c’est la même chose.
within a half hour frees up time for staff to take care of I will finish by returning to Dr Egan’s book. He pointed
other tasks and simplifies the approach to weaning. out that weaning is very nearly a pure art. As critical care
In our recent study, the intensive care unit mortality was shifts increasingly toward a focus on technology, more
5 per cent in patients who succeeded in a trial and didn’t than ever is there a need for the personal interaction be-
require reintubation.32 In contrast, patients who succeeded tween two human beings: the clinician and the patient.
in the trial, were extubated, but then required reintubation Respiratory therapists play a key role at the bedside of the
had a mortality rate of 33 per cent—a similar experience patient who’s frightened by the process of weaning. This
has been reported by Scott Epstein.33 We found that re- interplay between one human being and another will re-
spiratory frequency was high in the patients who failed the main the dominant factor in determining which patients
spontaneous breathing trial, but the values were similar in are successfully weaned from the ventilator.
the patients who were successfully extubated and in those
Thank you.
requiring reintubation. A superficial assessment of these
data might lead you to conclude that weaning indices do
not predict the need for reintubation. The design of our REFERENCES
study, however, was not adequate to reach a conclusion on 1. Pierson DJ. Weaning from mechanical ventilation in acute respira-
this issue. The patients in our study who failed the wean- tory failure: concepts, indications, and techniques. Respir Care 1983;
ing trial had a much higher frequency, and, if extubated, it 28(5):646–662.

430 RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4

 WEANING FROM MECHANICAL VENTILATION: WHAT HAVE WE LEARNED?

2. Egan DF. Fundamentals of respiratory therapy, 3rd edition. St Louis: 19. Laghi F, D’Alfonso N, Tobin MJ. Pattern of recovery from diaphrag-
CV Mosby; 1977. matic fatigue over 24 hours. J Appl Physiol 1995;79(2):539–546.
3. Tobin MJ, Alex CG. Discontinuation of mechanical ventilation. In: 20. Sassoon CSH. Intermittent mandatory ventilation. In: Tobin MJ,
Tobin MJ, editor. Principles and practice of mechanical ventilation. editor. Principles and practice of mechanical ventilation. New York:
New York: McGraw-Hill; 1994: 1177–1206. McGraw-Hill; 1994: 221–237.
4. Tobin MJ, Jubran A, Hines E Jr. Pathophysiology of failure to wean 21. Imsand C, Feihl F, Perret C, Fitting JW. Regulation of inspiratory
from mechanical ventilation. Schweiz Med Wochensch 1994;124(47): neuromuscular output during synchronized intermittent mechanical
2139–2145. ventilation. Anesthesiology 1994;80(1):13–22.
5. Tobin, MJ, Gardner WN. Monitoring of the control of ventilation. In: 22. Marini JJ, Smith TC, Lamb VJ. External work output and force
Tobin MJ, editor. Principles and practice of intensive care monitor- generation during synchronized intermittent mechanical ventilation:
ing. New York: McGraw-Hill; 1998: 415–464.
effect of machine assistance on breathing effort. Am Rev Respir Dis
6. Jubran A, Tobin MJ. Pathophysiological basis of acute respiratory
1988;138(5):1169–1179.
distress in patients who fail a trial of weaning from mechanical
23. Brochard L. Pressure support ventilation. In: Tobin MJ, editor. Prin-
ventilation. Am J Respir Crit Care Med 1997;155(3):906–915.
ciples and practice of mechanical ventilation. New York: McGraw-
7. Yang KL, Tobin MJ. A prospective study of indexes predicting the
Hill, 1994: 239–257.
outcome of trials of weaning from mechanical ventilation. N Engl
J Med 1991;324(21):1445–1450. 24. Strauss C, Louis B, Isabey D, Lemaire F, Harf A, Brochard L.
8. Tobin MJ, Laghi F. Monitoring respiratory muscle function. In: To- Contribution of the endotracheal tube and the upper airway to breath-
bin MJ, editor. Principles and practice of intensive care monitoring. ing workload. Am J Respir Crit Care Med 1998;157(1):23–30.
New York: McGraw-Hill; 1998: 497–544. 25. Tobin MJ. Respiratory mechanics in spontaneously-breathing pa-
9. Reid WD, Huang J, Bryson S, Walker DC, Belcastro AN. Dia- tients. In: Tobin MJ, editor. Principles and practice of intensive care
phragm injury and myofibrillar structure induced by resistive load- monitoring. New York: McGraw-Hill; 1998: 617–654.
ing. J Appl Physiol 1994;76(1):176–184. 26. Jubran A, Van de Graaff WB, Tobin MJ. Variability of patient-
10. Jubran A, Tobin MJ. Passive mechanics of lung and chest wall in ventilator interaction with pressure support ventilation in patients
patients who failed or succeeded in trials of weaning. Am J Respir with chronic obstructive pulmonary disease. Am J Respir Crit Care
Crit Care Med 1997;155(3):916–921. Med 1995;152(1):129–136.
11. Tobin MJ, Perez W, Guenther SM, Semmes BJ, Mador MJ, Allen SJ, 27. Parthasarathy S, Jubran A, Tobin MJ. Cycling of inspiratory and
et al. The pattern of breathing during successful and unsuccessful expiratory muscle groups with the ventilator in airflow limitation.
trials of weaning from mechanical ventilation. Am Rev Respir Dis Am J Respir Crit Care Med 1998;158(5 Pt 1):1471–1478.
1986;134(6):1111–1118. 28. Leung P, Jubran A, Tobin MJ. Comparison of assisted ventilator
12. Jubran A, Mathru M, Dries D, Tobin MJ. Continuous recordings of modes on triggering, patient effort, and dyspnea. Am J Respir Crit
mixed venous oxygen saturation during weaning from mechanical Care Med 1997;155(6):1940–1948.
ventilation and the ramifications thereof. Am J Respir Crit Care Med 29. Brochard L, Rauss A, Benito S, Conti G, Mancebo J, Rekik N, et al.
1998;158(6):1763–1769. Comparison of three methods of gradual withdrawal from ventilatory
13. Lemaire F, Teboul JL, Cinotti L, Giotto G, Abrouk F, Steg G, et al. support during weaning from mechanical ventilation. Am J Respir
Acute left ventricular dysfunction during unsuccessful weaning from Crit Care Med 1994;150(4):896–903.
mechanical ventilation. Anesthesiology 1988;69(2):171–179.
30. Esteban A, Frutos F, Tobin MJ, Alı́a I, Solsona JF, Vallverdu I, et al.
14. Stroetz RW, Hubmayr R. Tidal volume maintenance during weaning
A comparison of four methods of weaning patients from mechanical
with pressure support. Am J Respir Crit Care Med 1995;152(3):
ventilation. Spanish Lung Failure Collaborative Group. N Engl J Med
1034–1040.
1995;332(6):345–350.
15. Meade MO, Cook DJ. A framework for decision making. In: Tobin
31. Ely EW, Baker AM, Dunagan DP, Burke HL, Smith AC, Kelly PT,
MJ, editor. Principles and practice of intensive care monitoring. New
York: McGraw-Hill; 1998: 141–147. et al. Effect on the duration of mechanical ventilation of identifying
16. Jaeschke RZ, Meade MO, Guyatt GH, Keenan SP, Cook DJ. How to patients capable of breathing spontaneously. N Engl J Med 1996;
use diagnostic test articles in the intensive care unit: diagnosing 335(25):1864–1869.
weanability using f/VT. Crit Care Med 1997;25(9):1514–1521. 32. Esteban A, Alı́a I, Tobin MJ, Gil A, Gordo F, Vallverdu I. Effect of
17. Tobin MJ. Mechanical ventilation. N Engl J Med 1994;330(15): spontaneous breathing trial duration on outcome of attempts to dis-
1056–1061. continue mechanical ventilation. Spanish Lung Failure Collaborative
18. Anzueto A, Peters JI, Tobin MJ, de los Santos R, Seidenfeld JJ, Group. Am J Respir Crit Care Med 1999;159(2):512–518.
Moore G, et al. Effects of prolonged mechanical ventilation on di- 33. Epstein SK, Ciubotaru RL, Wong JB. Effect of failed extubation
aphragmatic function in healthy adult baboons. Crit Care Med 1997; on the outcome of mechanical ventilation. Chest 1997;112(1):
25(7):1187–1190. 186–192.

RESPIRATORY CARE • APRIL 2000 VOL 45 NO 4 431