Failure to wean (FTW) from mechanical ventilation (MV) is a significant clinical and economic problem. In 2003, approximately 300,00 patients required MV support for more than 96 hours in the USA and the estimated cost of these episodes was $16 billion [1]. The number of patients requiring long-term MV support is increasing five times as rapidly as the number of hospital admissions [2] and many of these patients experience FTW.

The etiology of FTW is often complex, but an imbalance in the demand placed on the inspiratory muscles used to generate inspiratory pressure during tidal breathing and their maximal pressure generating capability (Pibr/Pimax) has been implicated as a major contributor to this problem [35]. Numerous animal studies have documented ventilator-induced diaphragm dysfunction following as little as six hours of controlled MV [68], but less data examining the effects of MV on the human diaphragm are available. Knisely et al. [9] studied two children who had been ventilated for 7 and 45 days and qualitatively found profound atrophy of diaphragm muscle fibers following prolonged MV support. Levine et al. [10] documented approximately 55% atrophy in human diaphragms following 19 to 56 hours of controlled MV. Hermans et al. [11] recently reported marked reductions in magnetically stimulated transdiaphragmatic pressure in humans in the first week of MV support. Hussain et al. documented upregulation of catabolic process in human diaphragms following 15 to 276 hours of controlled MV [12], and Jaber et al. documented a 32% reduction in endotracheal tube pressure following magnetic diaphragm stimulation in humans following six days of MV support [13].

As an elevated Pibr/Pimax ratio is thought to be a major contributor to weaning failure [4, 5] and MV has been shown to rapidly cause diaphragm weakness in humans, strength training the inspiratory muscles emerges as a possible treatment for FTW. Preoperative inspiratory muscle strength training (IMST) has been shown to reduce the incidence of postoperative respiratory complications in high-risk cardiac surgery patients [14] and has also been demonstrated to preserve postoperative inspiratory muscle strength following major abdominal surgery [15].

We [16] and others [17, 18] have published successful case series and Caruso et al. published an unsuccessful [19] trial examining the effect of IMST on weaning outcome in FTW patients, but to date no adequately powered, randomized trial examining the effect of IMST on weaning outcome exists. We hypothesized that an IMST program, grounded in accepted principles of muscle strength training [20], coupled with progressively lengthening breathing trials (BT) would improve weaning outcome compared with the SHAM condition.

Materials and methods

After approval from the University of Florida Health Center Institutional Review Board (Federal wide Assurance FWA00005790), written informed consent was obtained from the patients or their legally designated surrogates. The trial was registered on Clinical Trials number NCT00419458. Patients were recruited from the adult medical, general surgical and burn ICUs of Shands Hospital at the University of Florida. Censuses of patients who were supported with MV were regularly queried and patients who had FTW with usual care were identified. Subjects were considered a FTW case when the patient failed to wean with usual care. Entry and exclusion criteria are shown in Table 1.

Table 1 Entry and exclusionary criteria

Subjects were studied from February 2004 until February 2009. The protocol was a single-blinded design with SHAM treatment. Subjects were blinded to their group assignment. Randomization was performed with a computerized random number generator and group assignments were sealed in opaque envelopes. Subjects were not randomized until they failed the initial BT.

Maximal inspiratory pressure measurement

Maximal inspiratory pressure (MIP) was measured on the first day of participation, every Monday and on days when the subjects attempted a 12-hour aerosol tracheotomy collar (ATC) trial. MIP was measured using the method of Caruso et al. [21]. Briefly, a one-way valve was attached to the patient's tracheostomy tube that allowed exhalation but blocked inspiration. The valve was connected to an electronic recording manometer and the patient was vigorously encouraged to inhale and exhale as forcefully as possible for 20 seconds. MIP measurements were repeated three times with a two-minute rest period with MV support between each attempt; the most negative value was recorded.

Inspiratory muscle strength training

IMST was performed five days per week (Monday to Friday) with a threshold inspiratory muscle trainer (Threshold PEP; Respironics Inc; Murrysville, PA, USA), which provided a threshold inspiratory pressure load between -4 and -20 cmH2O. The Threshold PEP device is marketed as an expiratory positive pressure device, but can provide an inspiratory threshold load if one inspires through the exhalation port. An inspiratory threshold training device is commercially available (Threshold IMT; Respironics Inc; Murrysville, PA, USA), but we found that many patients were unable to open the poppet valve at the lowest pressure setting (8 cmH2O) on the Threshold IMT device. When performing IMST, the subjects were disconnected from the MV and the IMST device was attached to their tracheostomy tube with the cuff inflated. Subjects breathed room air during IMST. Subjects performed four sets of 6 to 10 breaths per day, with two minutes of rest with MV support between each set. The training device was set to the highest pressure setting that the subject could consistently open during inspiration, and was progressed daily as tolerated. Subjects were instructed to inhale and exhale as forcefully as possible during the IMST breaths. The IMST training program was based on clinical experience obtained prior to initiating this trial. Respiratory pressures at the tracheostomy tube were monitored during IMST and SHAM training with CO2SMO Plus respiratory monitors with Analysis Plus software (Respironics Inc; Murrysville, PA, USA) interfaced to a laptop computer.

SHAM training

The SHAM group used a resistive inspiratory muscle training device (Pflex; Respironics Inc; Murrysville, PA, USA) set on the largest opening. The Pflex device had a 3 mm hole drilled into the body, which further reduced the pressure required to generate airflow. Subjects performed SHAM training by being removed from the ventilator circuit and the training device was attached to the tracheostomy tube. Subjects breathed room air during SHAM treatment. Subjects performed four sets of 6 to 10 breaths, five days per week, and were instructed to breathe with long, slow inspiratory and expiratory efforts during training. SHAM subjects were given two minutes of rest with MV support between each set. IMST and SHAM treatments were normally conducted between 07.30 am and 09.00 am, Monday through Friday.

Breathing trials

All subjects participated in progressively lengthening BTs with reduced or no MV support. Three types of BT were used: ATC, continuous positive airway pressure (CPAP) and reduced pressure support trials. Trials were conducted seven days per week, usually commencing around 09.00 am and only one trial per day was attempted. The initial BT was an ATC trial, and patients were allowed to breathe without MV support as long as tolerated. Subjects who tolerated this initial ATC trial for 72 hours were considered weaned and were not studied. Criteria for terminating BT included: 30 beats/min or more increase in heart rate, systolic blood pressure above 180 mmHg or below 90 mmHg, oxygen-hemoglobin saturation (SPO2) below 90% for five minutes, respiratory rate above 35 breaths/min for five minutes, serious dysrhythmias, if the patient requested to be returned to MV support or there was clinical evidence of respiratory distress (substernal retraction and sternocleidomastoid retraction, paradoxical breathing, or diaphoresis).

The daily progression for the ATC trials was: one, two, three, four, six, nine, and twelve hours. The second ATC trial was targeted for the step below the duration the patient tolerated on their first ATC trial, not to exceed six hours. For example, if a patient tolerated four hours on the initial ATC trial, the second ATC trial duration was three hours, the next four hours and so on. When a subject failed an ATC trial, the next trial was the same duration. If a subject was unable to participate in ATC trials for several days, the ATC trial target duration was decreased by the number of steps equal to the number of days missed. When subjects successfully completed a 12-hour ATC trial, the next day they progressed to breathing without MV support as tolerated. If they tolerated the ATC trial for 72 hours, they were classified as weaned.

If the subject was unable to complete at least one hour on the initial ATC trial, the next day a one-hour CPAP trial was attempted. CPAP trials were progressed by one hour per day until reaching three hours and then the patient began the ATC trial schedule as above. If the patient was unable to complete the initial one-hour CPAP trial, the next day they attempted a one-hour reduced pressure support trial (no synchronized intermittent mandatory ventilation breaths, about 50% of their baseline pressure support and baseline positive end expiratory pressure (PEEP)). If successful, the reduced pressure support trial duration was increased by one hour per day until reaching three hours whereupon they then began the CPAP and ATC trial progressions as detailed above.

Patients received usual nursing care during BT, but rehabilitation activities were withheld during BT until the patients could tolerate a six-hour ATC BT. Once patients could tolerate a six-hour BT, rehabilitation activity during BT was begun but reduced to approximately 50% of the normal duration and intensity until weaning. Breathing data during BT were monitored with ICU clinical bedside monitors and with a CO2SMO Plus respiratory monitor with Analysis Plus software (Respironics Inc; Murrysville, PA, USA) interfaced to a laptop computer. Prior to commencing the first and final BT, dynamic compliance and inspired and expired airway resistance were measured with the CO2SMO Plus respiratory monitors while the patients received their baseline level of MV support.

Statistical analysis

Categorical variables were analyzed with Chi-square tests. Between groups tests on continuous variables were analyzed with independent samples Student t tests. Within-group variables were analyzed with t tests for paired measures. Repeated measures analysis of variance (ANOVA) tests were used for variables with group, time factors, and group × times interactions. Cell means contrasts were used to explore differences when significant interactions were present in ANOVA. Statistical significance was set at P< 0.05.


The flow of subjects from evaluation to participation is shown in the CONSORT diagram (Figure 1). The randomization process resulted in groups that were equivalent on demographic factors, reasons for respiratory failure, treatment with renal replacement therapy, duration of MV prior to starting study intervention, duration of the initial ATC trial to failure, MIP, and other prognostic variables (Tables 2 and 3). Additionally, both groups experienced similar comorbidities during hospitalization before intervention (Table 4), received similar pharmacologic management during study intervention (Table 5), experienced similar complications during the study (Table 6), and underwent similar diagnostic and therapeutic procedures during study intervention (Table 7). Of note, 43% of the IMST subjects and 29% of SHAM subjects were dialysis dependent. Dialysis dependency has been associated with a reduced wean rate [22, 23].

Figure 1
figure 1

CONSORT diagram.

Table 2 Primary admission medical and surgical diagnoses
Table 3 Demographic and medical data
Table 4 Comorbidities between hospital admission and entering study
Table 5 Drug use during intervention by group
Table 6 Complications occurring during intervention period
Table 7 Diagnostic and therapeutic procedures performed during study

Six subjects did not fail during the initial ATC trial and were weaned without further intervention. These subjects were not randomized to treatment groups and were not included in the analysis. Three IMST subjects died during the 28-day treatment period, one withdrew from the study and two patients were transferred to other facilities before completing 28 days of treatment. These six subjects were classified as weaning failures. Three subjects in the SHAM group died during the 28-day treatment period and three subjects were transferred to other facilities before completing 28 days. These six subjects were also classified as weaning failures.

Excluding the initial BT, the IMST group performed 330 trials and the SHAM group performed 382 trials. The IMST and SHAM groups successfully completed 77.0% and 73.0% of the BT, respectively (P = 0.23).

The IMST and SHAM groups participated in 9.7 ± 4.0 and 11.0 ± 4.8 strength and SHAM training sessions, respectively (P = 0.09). The mean training pressure setting on the IMST device was 7.2 ± 2.6 vs. 12.8 ± 3.6 cmH2O for the initial and final training bouts, respectively (P< 0.0001, Table 3). The SHAM group's modified training device was set at the largest orifice (lowest resistance setting) for all sessions. The IMST group developed -9.54 ± 3.70 and -14.52 ± 4.59 cmH2O of inspiratory pressure at the tracheotomy tube during the initial and final IMST bouts (P = 0.0004). Corresponding training pressure values for the SHAM group were -3.10 ± 1.54 and -3.36 ± 2.08 cmH2O (P = 0.86). The treatment × group interaction for pressure developed during training was significant (P< 0.0001). The SHAM group's pre to post-training MIP change was not significant (-43.5 ± 17.8 vs. -45.1 ± 19.5 cmH2O, P = 0.39), while the IMST group's MIP increased (-44.4 ± 18.4 vs. -54.1 ± 17.8, cmH2O, P< 0.0001). There were no adverse events observed during IMST or SHAM treatments.

Twenty-five of 35 IMST subjects weaned (71%, 95% confidence interval (CI) = 55% to 84%), while 16 of 34 (47%, 95% CI = 31% to 63%) SHAM subjects weaned (P = 0.039). The number of patients needed to be treated for effect was 4 (95% CI = 2 to 80).

In order to further explore the role of MIP changes in weaning outcome, we performed a post-hoc analysis on MIP using weaning outcome as the independent measure. The pre- and post-training MIP measures for the weaning success (n = 41) and failure (n = 28) groups were respectively; -44.0 ± 20.2 and -53.5 ± 20.7 cmH2O versus -43.9 ± 14.8 and -43.9 ± 15.0 cmH2O. A repeated measures ANOVA revealed a significant outcome × time interaction and the change in MIP for the successfully weaned group was significantly greater than the failure to wean group (P< 0.0001).


Our primary findings were that the IMST rehabilitation program rapidly improved MIP and improved weaning outcome compared with the SHAM condition. The weaning rate (47%) achieved by the SHAM group was comparable with usual care conditions as reported in observational studies examining comparable FTW patients [2426].

Other workers have shown that MIP is a poor predictor of extubation success [2731]. Several differences between this study and the studies that found MIP to be a poor predictor of extubation outcome must be acknowledged: 1) studies that have shown MIP to be a poor predictor of extubation outcome examined intubated patients in the acute phase of MV support [28, 31], whereas our subjects had received MV support for approximately six weeks prior to starting intervention and all of our patients had tracheotomies; 2) our selection criteria identified patients who had FTW because of inspiratory muscle weakness that was amenable to strength training; and 3) none of the studies that have evaluated MIP as an extubation predictor used any type of strength training program to increase MIP. Investigators have found that higher values of MIP are associated with improved weaning outcome in chronic FTW patients. Yang [31] reported in a cross-sectional study that the Pibr/Pimax ratio of successfully weaned patients was lower than FTW patients. Carlucci et al. [5] have recently shown in an observational study with a group of long-term FTW patients similar to ours, that patients who eventually weaned, improved their MIP, and lowered the Pibr/Pimax ratio, while those who FTW did not. Our findings also support a role for increased MIP in improved weaning outcomes.

We propose that respiratory muscle weakness is a greater contributor to failed weaning than fatigue. During failed BTs in FTW patients, respiratory distress is often clinically described as "fatigue". However, several authors have reported heightened respiratory muscle activity during failed BTs compared with stable respiratory muscle activity among patients who successfully completed BTs. For example, Teixeira et al. [32] measured a 50% increase in the work of breathing of FTW patients over the course of failed BTs, whereas successful patients maintained a constant work of breathing during the trials. Jubran et al. [33] reported similar findings and an absence of low-frequency fatigue during failed BT.

Alternatively, we hypothesize that inspiratory muscle weakness initiates a high proportional ventilatory drive requirement during unassisted BT, when weakened inspiratory muscles must generate increased muscle tension in order to adequately ventilate the lungs. During MV support, a relatively low motor drive elicits large, ventilator-assisted tidal volume breaths. When an unsupported BT begins, a discrepancy between the elevated respiratory drive and afferent lung volume feedback can lead to an increased awareness of respiratory effort [34]. Perceived feedback errors will be addressed by further increases in respiratory motor drive, but the feedback discrepancy cannot be corrected by a highly-driven, weakened inspiratory pump that generates insufficient volume feedback [35].

Ongoing efferent-afferent feedback errors propel a positive feedback loop, resulting in the progressively higher levels of respiratory drive, inspiratory esophageal pressure, and work of breathing reported by others, and it may lead to clinical respiratory distress [34, 36]. If this positive feedback cycle progresses to high levels of inspiratory muscle work, reflex sympathetic activation can occur, with shunting of blood from the periphery to the working respiratory muscles [37, 38]. Elevated sympathetic activity is a probable cause of the tachycardia, hypertension, and diaphoresis frequently observed during failed BT in FTW patients. IMST has been shown to attenuate the sympathetic activation induced by high intensity inspiratory muscle work [39].

Strengthening the inspiratory muscles theoretically could correct the feedback discrepancy between respiratory drive and lung/chest expansion and may result in a lower perception of breathing effort. The perception of breathing effort has been experimentally altered by manipulations of inspiratory muscle strength. Campbell et al. [40] studied the perception of inspiring against standard inspiratory resistive loads before and after weakening the inspiratory muscles to about 30% of baseline with neuromuscular blockade. In the weakened state, subjects rated the effort of loaded breathing higher than in the unblocked condition. We [41] studied the effects of strengthening the inspiratory muscles on perception of inspiratory effort and respiratory drive in healthy subjects. Both the respiratory drive and the effort of breathing against standard inspiratory resistive loads were lower following a 50% improvement in MIP. These findings support the hypothesis that the perception of inspiratory effort and respiratory drive are inversely proportional to inspiratory muscle strength and may help explain why an increased MIP contributed to weaning.

Whenever severely debilitated patients undergo muscle strength training, the possibility of exercise-induced muscle damage must be considered. Human [42, 43] studies have documented that long-term, high resistance inspiratory loading can induce diaphragm muscle fiber damage. Although we did not examine diaphragm samples for training-induced damage, we think that it is unlikely that the IMST program induced muscle damage for the following reasons: 1) the duration of muscle loading during each IMST training session was approximately one minute per day. In contrast, animal and human studies have documented diaphragm damage with prolonged, high resistance loads, lasting 1.5 [44, 45] to 96 hours [46]. 2) Our IMST patients were able to inspire against increasing inspiratory loads on a daily basis. If the patients had been experiencing muscle soreness and contractile fiber damage from IMST, one would have expected diminished muscle performance, rather than increasing performance.

Our results are encouraging, but limitations must be acknowledged. The weaning results were significant, but this was a single site study with a relatively small sample size. Our IMST method is not suitable for all FTW patients. Patients must be sufficiently alert to cooperate with IMST, and patients whose FTW etiology is not the result of treatable inspiratory muscle weakness are unlikely to benefit from IMST. Our subjects were recruited primarily from surgical ICUs, with approximately 22% of the subjects treated in the medical ICU.


In conclusion, we found an improved MIP and weaning outcome with IMST compared with SHAM training in medically complex, long-term FTW patients. IMST is a clinically practical and safe method to improve weaning outcome in selected FTW patients.

Key messages

  • IMST can rapidly increase MIP in medically complex, long-term FTW patients.

  • IMST, in conjunction with BT, can increase the number of FTW patients weaned versus SHAM training plus BT.