INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

It is estimated that 1 to 5% of mechanically ventilated patients repeatedly fail attempts at weaning from mechanical ventilation and face a substantial risk of becoming chronic ventilator-dependent patients who cannot sustain spontaneous breathing for longer than a few hours (1). This proportion increases to as much as 31 to 56% in some long-term ventilator units where difficult-to-wean patients are located (1, 2). Studies of predictors, protocols and specific weaning strategies have been largely confined to patients intubated for shorter periods of time (3). Chronic ventilator dependence is not only a major medical problem but it is also an extremely uncomfortable state for a patient, carrying important social implications (1). Yet few investigators have attempted to determine the pathophysiologic mechanisms underlying weaning failure in long-term ventilator-dependent patients (7, 8). We believe that a better understanding of these mechanisms could help to improve the general management of such patients and ultimately lead to a partial, if not complete, discontinuation of mechanical ventilation, or to the closure of the tracheostomy and implementation of nocturnal noninvasive mechanical ventilation, which may be more acceptable to the patients.

Therefore, we undertook this study to investigate the pathophysiologic mechanisms hindering the liberation from mechanical ventilator in long-term ventilator-dependent patients. Although no predetermined selection was planned, the study population consisted of patients with advanced chronic obstructive pulmonary disease (COPD) and postoperative patients suffering from phrenic nerve lesion after cardiac surgery.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The study was conducted according to the guidelines of the Declaration of Helsinki, and the protocol was approved by the Institutional Ethics Committee. Written informed consent was obtained from each patient before enrollment in the study.

Patients

A total of 59 patients was enrolled in the study. At the time of the study they had to be in reasonably stable clinical condition, i.e., free of acute episodes such as fever, pain, or hemodynamic instability. Patients with neoplasms, body temperature above 38° C, a hemoglobin level below 10 g/dl, severe primary cardiomyopathy, sepsis, or severe renal or hepatic failure were excluded.

Thirty-nine patients were receiving mechanical ventilation through a tracheostomy cannula (tracheal cannula size, 9 ± 1 mm ID). These patients had been admitted to the Chronic Ventilator Unit of our Institution from other hospitals because, after several weaning failures and the execution of the tracheostomy, the caring physicians classified them as difficult-to-wean. Here, the patients were taken in charge by the primary physician of the Unit. Our weaning protocol was considered only when the patients were judged to have reached not only clinical stability, but also their possible best clinical "performance." Then, patients meeting the criteria listed below were asked to provide their informed consent to be enrolled in the study protocol. Because of the historical characteristics of the Institution, which is the reference rehabilitation and chronic care center for a large geographic area in North Italy, patients afflicted predominantly by lung and cardiac diseases are admitted here to follow specific rehabilitation programs. Among other patients, long-term mechanically ventilated patients are also admitted to our Institution to undergo a program of progressive discontinuation of mechanical ventilation and to be discharged to a home program of long-term ventilator assistance if liberation from the ventilator fails.

Twenty-eight patients (eight females and 20 males 67 ± 8 yr of age) had a diagnosis of COPD based on clinical history and previous pulmonary function tests. At the time of the study, these patients had been ventilated for 33 ± 9 d. The remaining 11 patients (6 females and 5 males 67 ± 6 yr of age) underwent either valvular replacement or coronary bypass surgery, but could not be disconnected from the ventilator after the recovery from surgery. At the time of the study they had been ventilated for 22 ± 9 d. Henceforth these patients will be referred to as postcardiac surgery (PCS) patients. All PCS patients showed the following suggestive signs of diaphragmatic dysfunction (9): (1) radiologic findings of an elevated hemidiaphragm; (2) fluoroscopic evidence of hemidiaphragmatic paradoxical motion; and (3) paradoxical abdominal motion in the supine position during spontaneous breathing. Echography excluded the presence of pleural effusion in every instance.

Two additional groups of patients were enrolled in the study. The first one (stable COPD) consisted of nine tracheostomized patients with stable COPD (one female and eight males 65 ± 8 yr of age) who needed mechanical ventilation for acute exacerbations of COPD, but were breathing spontaneously through the tracheostomy at the time of the study. They had not been using the ventilator for 18 ± 7 mo. The tracheostomy was not closed upon the decision of the physicians in charge. These patients were receiving long-term oxygen therapy and were in a rehabilitation program. The second additional group (stable PCS) consisted of 11 spontaneously breathing patients (four female and seven male 63 ± 8 yr of age) who had been successfully disconnected from the ventilator within 48 h after cardiac surgery and were attending a rehabilitation program in our Institution. At the time of the study (19 ± 6 d after surgery) all stable PCS patients met the same criteria for diaphragmatic dysfunction as their mechanically ventilated PCS counterparts (9), and they were in a stable condition without intubation or tracheostomy. Patients' characteristics and lung function data obtained from the clinical records accompanying the patients on admission to the hospital are reported in Table 1.

Measurements

Physiologic signals, namely, flow (), volume (V), esophageal (Ppl), gastric (Pab), and airway opening (Pao) pressures were obtained as previously described (7, 10). To measure flow and volume a heated Lilly pediatric-type pneumotachograph, connected to a differential pressure transducer (Screenmate 701240-015009; Jaeger, Wurzburg, Germany) was inserted between the proximal tip of the tracheal cannula in both mechanically ventilated COPD and PCS patients, whereas it was connected either to the proximal tip of the tracheal cannula (stable COPD) or to a mouthpiece (stable PCS) in spontaneously breathing patients. Transpulmonary (PL) and transdiaphragmatic (Pdi) pressures were obtained by subtracting Pes from Pao and Pga, respectively. The signals were fed into a microcomputer through an A/D board (Compaq 386 equipped with 80387 math coprocessor; Compaq, Houston, TX, , and with DT2801/A A/D board, Data Translation, Marlboro, MA) and processed with appropriate software (Labdat and Anadat; RHT-InfoDat Inc., Montreal, PQ, Canada). Data are reported as mean ± SD unless otherwise specified.

Arterial blood gases and pH were measured with a gas analyzer (ABL300; Radiometer, Copenhagen). Oxygen saturation was continuously monitored by a pulse oximeter (Minolta).

Minute ventilation (E), tidal volume (VT), and inspiratory (TI) and expiratory (TE) time, total cycle duration (Ttot), respiratory frequency (f), mean inspiratory flow (VT/TI), and the "duty cycle" (TI/ Ttot) were calculated from the flow and volume signals. The respiratory frequency was also measured on the Ppl and Pdi records (central frequency: f_c) to detect the occurrence of ineffective inspiratory efforts, i.e., inspiratory muscle contractions occurring during expiration but insufficient to generate inspiratory flow and volume (Figure 1). The f/VT ratio (3) was computed by dividing both f and f_c by VT. Dynamic lung compliance (CL_dyn) and pulmonary resistance at midinspiratory volume (RL) were computed from PL, , and V records as previously described (7, 10). Dynamic intrinsic positive end-expiratory pressure (PEEPi_dyn) was estimated from the decrease in Ppl preceding the start of the inspiratory flow with correction for expiratory muscle activity by subtraction of the negative deflection in Pab (7, 10).

View larger version (28K): [in this window] [in a new window]   Figure 1.   Representative record from a patient showing ineffective inspiratory efforts (IIE) during the spontaneous breathing trial. From top to bottom: flow = airflow; volume = tidal volume; Pao = airway opening pressure; Ppl = esophageal pressure; Pdi = transdiaphragmatic pressure. Arrows indicate esophageal and transdiaphragmatic pressure swings that were not associated with the generation of inspiratory flow and volume. As a result, the breath after an IIE started at a lower end-expiratory lung volume, and the ventilatory frequency (f ) underestimated the central frequency (fc).

The magnitude of the inspiratory muscle effort was estimated from the pressure-time product for the inspiratory muscles (PTPpl) (11) and for the diaphragm (PTPdi) (7, 11). Mean inspiratory Ppl () was computed as the area subtended by Ppl and the chest wall static recoil pressure (Pstw)-time curve divided by the inspiratory effort duration. Mean inspiratory Pdi () was obtained by dividing the area subtended by the Pdi swing per breath by the corresponding time (12).

Maximum inspiratory pressure (MIP), maximum transdiaphragmatic pressure (Pdi_max) and maximum esophageal (Ppl_max) pressures were measured in each subject as previously described (7). Tension-time index of the diaphragm (TTdi) was computed using Pdi_max, according to the method of Bellemare and Grassino (13, 14). Mean inspiratory Ppl and Pdi were also expressed as a fraction of Ppl_max and Pdi_max, respectively.

The neuromuscular drive was estimated by the decrease in airway opening pressure at 0.1 s (P_0.1) after the onset of an inspiratory effort against an occlusion, as previously described (15). We also computed the P_0.1/VT/TI ratio, i.e., the effective inspiratory impedance (16).

Procedure

Mechanically ventilated patients. At the time of the study the ventilator-dependent patients had been ventilated, upon the decision of the primary physicians, in the synchronized intermittent mandatory ventilation mode (SIMV) with a tidal volume of about 8 ml/kg and a frequency of about 12 to 14 breaths/min (Evita ventilator; Dräger, Germany). Following their clinical judgment, the primary physicians decided if the patient was ready to attempt to discontinue mechanical ventilation and to resume spontaneous breathing. The first condition for planning weaning was that the disease process precipitating the need for mechanical ventilation appeared satisfactorily resolved, i.e., there were no laboratory data or chest radiographic findings suggesting either extrapulmonary or respiratory tract infection. In addition, patients had to be in reasonably stable clinical (no fever, pain, or anxiety, etc.) and hemodynamic (no tachycardia or systemic hypotension) condition, without evident signs of respiratory distress (total breathing frequency < 35 breaths/min) and satisfactorily oxygenated (Sa_O2 > 90% with a FI_O2  0.4) without acidosis.

All patients were studied in the semirecumbent position. After topical anesthesia two balloon-tipped polyethylene catheters were inserted through the nose into the stomach and esophagus (7, 10), and the "occlusion test" was used to ensure the correct positioning of the esophageal catheter (17). As soon as the patient was accustomed to the experimental setting and equipment, arterial blood was sampled from the radial artery to measure Pa_O2, Pa_CO2, and pH. Then the weaning trial started, and the patient was disconnected from the ventilator and left to breathe spontaneously while receiving supplemental oxygen to keep Sa_O2 > 93% (18). The circuit included the pneumotachograph, connectors with the side port to sample Pao, and a T-tube (4). Physiologic signals were collected for 3 min from the commencement of unsupported breathing. The last of these 3 min was used for data analysis. Then a unidirectional valve (Model AC1315C; ASEM Milano, Italy) was inserted in the circuit to measure P_0.1, Ppl_max and Pdi_max according to our usual procedure (7). After this, patients continued to breathe spontaneously until one or more of the following events occurred: (1) oxygen saturation of 90% or less at an FI_O2 of 0.5, (2) diaphoresis, (3) evidence of increasing respiratory distress, (4) tachycardia, (5) arrhythmias, or (6) hypotension (4). In these circumstances we collected the physiologic signals for 1 minute and sampled the blood gases before reconnecting the patient to the ventilator. In patients who did not show any sign of respiratory distress and were able to continue to breathe, signals were collected at 60 min from the baseline measurement before the withdrawing of the balloon catheters and the end of the study.

Spontaneously breathing patients. Patients with stable COPD and PCS patients were also studied in the semirecumbent position. Using the same procedure, i.e., connection to the pneumotachograph and positioning of the esophageal and gastric balloons, we collected the physiologic signals of flow, volume, and pressures to compute the physiologic variables mentioned above in the same sequence. Supplemental oxygen was delivered to maintain oxygen saturation above 93%.

Data Analysis

Comparisons between groups were performed using a two-stage approach: first, a Kruskall-Wallis one-way nonparametric analysis of variance was performed to detect an overall difference (p < 0.05); second, if the nonparametric ANOVA showed overall differences, these were investigated by performing Bonferroni-corrected, Mann-Whitney U-tests, again at a global significance level of p < 0.05. Changes in breathing pattern, neuromuscular drive, lung mechanics, inspiratory muscle effort, and arterial blood gases between the start and finish of the weaning trial were compared by means of Student's paired t test. Linear regression by means of the least-squares method was employed to calculate the correlation between physiologic indices (19).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Thirty-one of the 39 patients submitted to the weaning trial did not tolerate unsupported breathing for longer than 40 ± 14 min and had to be reconnected to the ventilator because of excessive dyspnea and respiratory distress. Twenty of those thirty-one patients were mechanically ventilated because of exacerbation of COPD, whereas the other 11 were PCS patients who did not resume spontaneous ventilation after cardiac surgery. In the subsequent period, 27 of the 31 patients remained ventilator-dependent and were discharged to a home mechanical ventilation program; two ventilator-dependent patients were discharged to hospitals closer to their homes; and two patients died from complications (sepsis) 16 and 18 d after the study, respectively. Henceforth, the 31 patients who could not be liberated from mechanical ventilation were labeled as VD (ventilator-dependent) patients with COPD and PCS patients. By contrast, eight patients, all with COPD, were able to sustain spontaneous breathing through the T-tube and did not require reconnection to the ventilator. These patients are labeled as W (successfully weaned) with COPD. No PCS patients could be weaned. All eight patients with W-COPD were able to continue unsupported breathing also in the days after the test until they were discharged from hospital still with the tracheostomy and were enrolled in a rehabilitation program and long-term O₂ therapy. In Table 2 we show the physiologic data collected in the 39 mechanically ventilated patients in the third minute of spontaneous breathing after disconnection from the ventilator and in the two groups of stable, nonventilated patients.

Breathing Pattern

Minute ventilation and tidal volume were significantly lower in VD-COPD than in W-COPD, whereas breathing frequency did not differ between the two groups. W-COPD exhibited a breathing pattern not significantly different from that of patients with COPD in stable conditions. In VD-PCS patients, VT was significantly lower than in stable PCS patients and similar to VD-COPD. However, the breathing frequency was higher than in VD-COPD such that E was higher, too, and not different from W-COPD and the two groups of stable, nonventilated patients. The rapid shallow breathing index (f/VT) averaged 111 ± 98, 99 ± 40, and 50 ± 28 breaths/min/L in VD-COPD, VD-PCS, and W-COPD, respectively. However, in eight patients with VD-COPD, we observed ineffective inspiratory efforts as illustrated in Figure 1. Some Ppl and Pdi pressure swings occurred early during expiration in the unsupported breathing period and were not associated with the generation of inspiratory flow and volume. In these patients ventilatory frequency (f) averaged 20.3 ± 4.5 breaths/min, whereas the central frequency (f_c) was greater, averaging 28.0 ± 5.7 breaths/min. In these patients f/VT ratio averaged 69 ± 29 breaths/min/L, whereas f_c/VT was greater, averaging 96 ± 45 breaths/min/L. Also in the whole group of 20 patients with VD-COPD the f_c/VT ratio (122 ± 96 breaths/min/L) was significantly higher than the f/VT ratio (p = 0.0185). In particular, accepting the value of f/VT > 100 as a poor predictor of successful weaning (3), four patients with VD-COPD shifted from f/VT < 100 to f_c/VT > 100, whereas in the other two patients with VD-COPD the f_c/VT went above 80, which represents the lower bound of that threshold (3); one patient changed f/VT only from 43 to 49 and the last patient had already a f/VT > 100 (Figure 2).

View larger version (17K): [in this window] [in a new window]   Figure 2.   Relationship between flow-based breathing frequency to tidal volume (f/VT) and Pdi-based breathing frequency to tidal volume (fc/ VT) ratios in patients with VD-COPD. Horizontal and vertical bars = threshold of 100 breaths/min/L (drawn from reference 3); oblique line = identity line. For explanations, see text.

Neuromuscular Drive

P_0.1 was higher in both groups of VD patients, being slightly greater in VD-COPD than in VD-PCS (Table 2). Similar to the breathing pattern, no difference was perceivable between W-COPD and stable COPD, though in the latter P_0.1 was higher than in stable PCS patients. The data of effective respiratory impedance (P_0.1/VT/TI) are shown in Figure 3. Essentially, they reflect the behavior of P_0.1, though with a better separation between VD-COPD and W-COPD and stable COPD, i.e., there was virtually no overlap among these groups of patients.

View larger version (12K): [in this window] [in a new window]   Figure 3.   Individual and average values for effective inspiratory impedance (P0.1/VT/TI) at the start of the weaning trial (Groups VD-COPD, W-COPD, VD-PCS) and during spontaneous unsupported breathing in the control groups (stable COPD, stable PCS). Horizontal bars = average values. *p < 0.05, Group VD-COPD versus Group W-COPD and Group stable COPD; #p < 0.05, Group VD-COPD versus group VD-PCS; §p< 0.05, Group VD-PCS versus Group stable PCS.

Lung Mechanics

Data of lung mechanics are reported in Table 2. PEEPi_dyn was highest in patients with VD-COPD, and was significantly greater than in patients with W-COPD and those with stable COPD. Also, five patients of group VD-PCS had some PEEPi, without any history of airway disease, probably because of the high frequency of breathing and the additional flow resistance of the tracheal cannula. Pulmonary resistance was significantly higher in the patients with VD-COPD than in both those with W-COPD and those with stable COPD. For both PEEPi_dyn and RL there was no difference between W-COPD and stable COPD. Also VD-PCS patients had greater pulmonary resistance than did stable patients, though this could be due to the tracheal cannula. Dynamic lung compliance was not significantly different between the groups.

Inspiratory Muscle Strength and Effort

It can be seen in Table 2 that all indices of inspiratory muscle strength were lower in VD-COPD and VD-PCS groups than in the groups of W-COPD, stable COPD or stable PCS. MIP was lower than Ppl_max in patients with PEEPi, as shown in Figure 4. As predictable on the basis of the underlying disorder, Pdi_max was smaller than Ppl_max in PCS patients, particularly in VD-PCS. The magnitude of the inspiratory effort, assessed by means of PTPpl and PTPdi, was only slightly different between the groups. The data of the load/capacity balance, i.e. mean Ppl and Pdi expressed as a fraction of maximum Ppl and Pdi, respectively, are illustrated in Figure 5. In both groups of VD patients, the load/capacity balance was greater than in W or stable patients. The individual overlap was minimum between patients with VD-COPD and the other patients with COPD. Interestingly, the separation among the groups of patients with COPD was around 0.4 (20). The individual values of TTdi can be seen in Figure 6, whereas in Table 2, average values for the different groups are provided. The majority of patients classified in the VD-COPD and VD-PCS groups were close to or above the 0.15 threshold (13, 14), whereas all the weaned and stable patients were below it. The correlation between the effective inspiratory impedance and the load/capacity balance for the inspiratory muscles is shown in Figure 7. The patients with the highest P_0.1/VT/TI ratio were those with the highest inspiratory resistance. The same relationship was found also with the /Pdi_max ratio and with TTdi.

View larger version (22K): [in this window] [in a new window]   Figure 4.   Relationship between maximal inspiratory pressure (MIP) and maximal esophageal pressure (Pplmax) in the five groups of patients. Horizontal bars = average values. The legend of the symbols is provided in the figure. Solid line = identity line; dashed lines = isophlets of MIP expressed as percentage of Pplmax. In patients with VD-COPD MIP significantly underestimated the force- generating capacity of the respiratory muscles if compared to Pplmax measured during the same maximal maneuver (Mueller maneuver).

View larger version (16K): [in this window] [in a new window]   Figure 5.   Individual and average values for Load/Capacity Balance in the five groups of patients. (Top panel ) Ratio between mean and maximal inspiratory esophageal pressures (Ppl/Pplmax). (Bottom panel ) Ratio between mean and maximal inspiratory transdiaphragmatic pressures (Pdi/Pdimax). Horizontal bars = average values; dashed lines = cutoff value of 0.4 (drawn from Reference 20) for both Ppl/Pplmax and Pdi/Pdimax. For other definition of abbreviations, see Figure 1. *p < 0.05, Group VD-COPD versus Group W-COPD and Group stable COPD; #p < 0.05, Group VD-COPD versus group VD-PCS; §p < 0.05, Group VD-PCS versus Group stable PCS. For further explanations, see text.

View larger version (16K): [in this window] [in a new window]   Figure 6.   Tension-time diaphragmatic index (TTdi) at the start of the weaning trial (Groups VD-COPD, W-COPD, VD-PCS) and during spontaneous unsupported breathing in the control groups (stable COPD, stable PCS). TI = inspiratory time; Ttot = total respiratory cycle duration; TI/Ttot = inspiratory time expressed as a fraction of Ttot; Pdi/Pdimax = ratio between mean inspiratory and maximal transdiaphragmatic pressures. The hyperbolic curve represents a TTdi value of 0.15 (cutoff value drawn from references 13 and 14). Other symbols = see the legend in the figure. Most of the patients who failed to wean showed a TTdi > 0.15. Successfully weaned patients and controls had TTdi values below this threshold.

View larger version (17K): [in this window] [in a new window]   Figure 7.   Relationship between effective inspiratory impedance and Ppl/Pplmax ratio of the five groups of patients. P0.1/VT/TI = effective inspiratory impedance; Ppl/Pplmax = ratio between mean and maximal inspiratory esophageal pressures. Solid line = regression line. Other symbols = see the legend in the figure. The regression equation is provided on the top of the figure. For further explanations, see text.

Physiologic measurements were repeated immediately before reconnection to the ventilator in patients who needed it. We did not find any significant change in the breathing pattern, neuromuscular drive, lung mechanics, or inspiratory effort in comparison with the data obtained immediately after disconnection from the ventilator (Table 2). For example, at the end of the unsuccessful weaning trial in patients with COPD and PCS patients, respectively, E averaged 6.5 ± 1.5 and 9.8 ± 3.2 L/min, P_0.1 averaged 5.5 ± 1.7 and 4.0 ± 0.8 cm H₂O, and PTPpl averaged 395 ± 120 and 305 ± 145 cm H₂O/s over 1 min. By contrast, in all the 31 patients who failed to wean (VD-COPD and VD-PCS) Pa_CO2 was higher (Pa_CO2: +12.3 ± 8.0 mm Hg, p < 0.0001) and pH lower (pH: 0.08 ± 0.04, p < 0.0001) than their respective values measured immediately before disconnection from the ventilator (Table 1).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The purpose of this study was to investigate the pathophysiologic mechanisms underlying the inability to sustain spontaneous ventilation in ventilator-dependent patients referred to a special Chronic Care and Rehabilitation Center. In general, such patients are sent to this kind of Unit after a number of days of mechanical ventilation in the Intensive Care Unit (ICU) and after repeated failure of weaning attempts. Several studies on difficult-to-wean patients have been performed in the ICU setting, in patients recovering from acute respiratory failure (6, 21), whereas very few data exist on the pathophysiologic profile of ventilator-dependent patients out of the acute episode, when these patients are potential candidates for chronic, long-term, home ventilatory assistance (7, 8).

This study shows that patients who could not be liberated from the ventilator had some common pathophysiologic characteristics, which became apparent soon after disconnection from the ventilator: (1) small tidal volume; (2) high neuromuscular drive; (3) abnormal lung mechanics; (4) reduced inspiratory muscle strength. As a result, both the load/capacity balance and the effective inspiratory impedance were greater than in patients who continued to breathe spontaneously. Clearly, there were also some differences related to the underlying pathology, namely COPD or PCS.

Breathing Pattern

Compared with patients who were liberated from the ventilator, and to those in stable conditions after their rehabilitation program, all the patients who had to be reconnected to the ventilator had a reduced VT at the start of the unsupported breathing period. However, the breathing frequency and E were higher in the VD-PCS than in the patients with VD-COPD. In the latter group the spontaneous breathing frequency was only slightly greater than that in patients who were successfully weaned or those with stable COPD receiving long-term O₂ therapy (Table 2). This difference between the two groups of VD patients cannot be attributed to different phases of clinical condition because all patients had been ventilated for more than 3 wk before our study. Rather, it could be explained by differences in the lung mechanics characteristics of the two groups.

Patients with advanced COPD have pulmonary hyperinflation caused by the loss of lung recoil and small airway closure early in expiration, aggravated by dynamic hyperinflation because of expiratory flow limitation (8, 24). Under these circumstances, ineffective inspiratory efforts can occur when the pressure generated by the inspiratory muscles is inadequate to counterbalance the elastic recoil of the respiratory system and to produce inspiratory flow. Ineffective inspiratory efforts have been found during assisted mechanical ventilation (7, 25), and during recovery from acute exacerbation in patients with COPD (26). In eight of our 20 patients with VD-COPD the central frequency (f_c), i.e., including the ineffective inspiratory efforts (Figure 1), was 40% greater than the apparent ventilatory frequency (f). However, ineffective efforts cannot contribute to defend minute ventilation, which remained low.

Furthermore, because of pulmonary hyperinflation, patients with severe COPD breathe close to the upper portion of the volume/pressure (V/P) curve, where a nonlinear relationship occurs. A higher breathing frequency would shorten the expiratory time causing a further rise in the end-expiratory lung volume (EELV). Because of the flat configuration of the V/P curve in the upper extreme, any increase in the EELV would further reduce VT. Hence, it may be hypothesized that patients with VD-COPD only slightly increased their breathing frequency also because of the mechanical constraints of the system. This seems to have occurred in at least five of the 20 patients with VD-COPD (25%), in whom the average f was 18.0 ± 3.7 breaths/min, f/VT ratio was 49 ± 20 breaths/min/L, and no ineffective inspiratory efforts were found. Although this argument is perhaps speculative, it might help the clinician to understand what patients do with their breathing patterns in an effort to protect themselves from further hyperinflation.

The lack of rapid shallow breathing in a significant percentage of our patients with VD-COPD (65%) soon after disconnection from the ventilator constitutes an important difference from data obtained in the ICU setting (22). On the one hand, similar to what has been reported for ICU patients (6), a high f/VT ratio, e.g., > 100 breaths/min/L (3), is highly predictive of weaning failure in COPD also in the chronic settings. On the other hand, we found that a low f/VT, e.g., < 80 breaths/min/L (3), was not always associated with a successful disconnection from the ventilator in the chronic care setting. However, our data in Figure 2 show that f/VT ratio, computed on the true central frequency of breathing (f_c/VT), can be greater than the apparent ventilatory frequency (f/VT), because of ineffective inspiratory efforts. As shown in Figure 2, four (33%) of 12 patients with VD-COPD, who had f/VT ratio < 100 breaths/min/L, had a f_c/VT ratio > 100, whereas another two patients went above the lower bounds of f/VT (80 breaths/ min/L) identified by Yang and Tobin (3). These data can explain, at least in part, the low f/VT found in some chronically ventilator-dependent patients for some specific disease categories.

Compared with patients with COPD examined in an ICU setting (22), the patients of our study were ventilated for a longer period, i.e., 33 ± 9 versus 19 ± 31 d, and had much worse lung mechanics. This finding may have important pathophysiologic and clinical implications. First, the high PEEPi suggests that pulmonary hyperinflation may play an important role in the mechanisms underlying chronic ventilator dependence. Second, one should be aware that difficult-to-wean patients with COPD, ventilated for about a month and bearing a tracheostomy, may not necessarily exhibit, when disconnected from the ventilator, the rapid shallow breathing (see above) observed in patients recovering from acute respiratory failure (3, 22). Under these conditions, the severity of the patient's distress could be underestimated by caregivers looking for a shallow but also rapid breathing. Our mechanical interpretation of the shallow, but not rapid, breathing of a significant percentage of patients with VD-COPD is supported by the fact that our VD-PCS patients, who were also out of the acute episode, did increase their frequency to defend minute ventilation. As a matter of fact, at variance from patients with severe COPD, cardiac patients with functional lesion of the phrenic nerve are likely to have a reduced EELV (27), but not flow limitation and dynamic hyperinflation, the major mechanical constraints to increasing f and E.

Neuromuscular Drive

Although with some known limitations (28), the high values of P_0.1 suggest an enhanced neuromuscular drive and indicate that the low VT in VD patients is not due to an insufficient central stimulus, but rather to a poor transformation of that high drive into the ventilatory output. This is clearly reflected by the high values of P_0.1/VT/TI, i.e., the effective inspiratory impedance (16).

It should be mentioned that variables derived from the breathing pattern and drive may be influenced by some factors. Although it is conventionally believed that the breathing pattern may exhibit a considerable degree of interindividual and intraindividual variability, data obtained in normal subjects suggest that the range of variability is relatively narrow, hence making it easy to detect abnormalities in patients (29). In this study we did not address the issue of variability. Further studies might be suggested to determine breath-by-breath and day-to-day variability of the breathing pattern in patients with prolonged mechanical ventilation. Regarding P_0.1, alterations in the end-expiratory lung volume, further development of abnormal muscle length and chest wall distortion may affect its value. It has been reported that the variability of P_0.1 may be greater than that of breathing pattern variables (29). However, the documentation of a high drive to breathe, was predictive of weaning failure in our patients as it was in patients of previous studies (6, 15, 30).

Mechanical Load and Load Capacity Balance

In agreement with other studies on ventilator-dependent patients with COPD (7, 21), lung mechanics was severely abnormal in our patients. In particular, PEEPi_dyn and RL were greater in VD-COPD than in W-COPD or stable COPD. It has to be noted also that static PEEPi, which reflects more closely the actual inspiratory threshold load and the real level of dynamic hyperinflation (31), averaged 9.6 ± 4.7 cm H₂O, whereas PEEPi_dyn amounted to only 5.9 ± 3.0 cm H₂O, i.e, 61% of static PEEPi. Among other factors such as malnutrition (in particular in patients mechanically ventilated for many days), electrolitic disturbances, and steroid therapy instituted to treat the acute exacerbation (24), pulmonary hyperinflation is a cause of the reduced inspiratory muscle strength that was greater in VD-COPD, in line with the finding of high values of PEEPi (32). In this condition, the large inspiratory work load caused by abnormal resistance and PEEPi had to be sustained by the inspiratory muscles with reduced pressure-generating capacity (Table 2); hence, leading to a more unfavorable load-capacity balance (33) compared with W-COPD and stable COPD (Figure 5).

Both the /Pdi_max and /Ppl_max ratios were > 0.4 in VD-COPD. Studies in normal subjects have shown that / Pdi_max > 0.4 was associated with the incapability to sustain spontaneous breathing for long (20). In line with these data, the TTdi, an index of diaphragmatic endurance (13, 14), not only was higher in the VD-COPD compared with the other two groups, but also individual values were close to or above 0.15 in almost all the patients of that group. A series of studies in normal subjects (13) and in patients with stable COPD (14) have suggested that a TTdi > 0.15 was invariably associated with signs of diaphragmatic fatigue and excessive distress during spontaneous ventilation. These data, which are in agreement with previous findings (7, 21), suggest that in patients with VD-COPD the inspiratory muscles are facing an excessive work load in terms of their capability to meet the ventilatory demand. Under these circumstances, the task of maintaining spontaneous ventilation and sufficient CO₂ clearance may become unbearable. This condition may hamper the possibility to withdraw mechanical ventilation without excessive patient distress and CO₂ retention, thus determining chronic ventilator dependence (7, 33). Whether this may be related to inspiratory muscle fatigue remains to be established (21, 23).

The VD-PCS patients had some similarities and some differences compared with the patients with VD-COPD, though the outcome was eventually the same, i.e., chronic ventilator dependence. As previously mentioned, the VD-PCS patients had a low VT but were more able to defend E by increasing frequency. That impairment of the overall inspiratory muscle strength and of the diaphragm in particular was likely the consequence of a functional phrenic nerve lesion (27), which was more severe in VD-PCS patients than in the patients with stable PCS. Although lung mechanics was not severely compromised, the reduction in the diaphragmatic pressure-generating capacity increased the load/capacity balance to a level similar to that of the VD-COPD and close to the 0.4 threshold (20), in the presence of a higher than normal P_0.1.

End of Spontaneous Breathing Trial

At variance with studies in the ICU (3, 22), there was no significant change in the physiologic variables examined in this study between the beginning and the end of the spontaneous breathing trial either in VD patients or in those who were weaned. However, the increase of Pa_CO2 at the end of the spontaneous breathing period indicates that the spontaneous breathing pattern was not efficient in terms of CO₂ elimination and that the ventilatory pump was not coping with the metabolic demand (34). It might be speculated that patients who had to be reconnected to the ventilator were already breathing at their maximum potential and did not have any room for change such that the low VT with increasing CO₂ retention might reflect the behavior of patients with a failing ventilatory pump.

Clinical Implications

The data from this study provide physiologic insight into why some patients cannot be liberated from the ventilator. The demonstration of a marked load-capacity imbalance, resulting from an excessive inspiratory work load and reduced inspiratory muscle pressure generating capacity, may explain why an enhanced drive to breathe results in a poor ventilatory outcome in terms of both CO₂ clearance and capability to sustain spontaneous breathing. In addition, our data show that PCS patients, and some patients with COPD in whom mechanical ventilation had to be reinstituted because of unbearable dyspnea, developed rapid shallow breathing soon after disconnection from the ventilator. Similarly to some studies on large heterogeneous ICU populations (3, 5, 6, 30), a high f/VT ratio > 100 can predict weaning failure also in the chronic care setting. By contrast, in a significant proportion of long-term, ventilator-dependent patients with COPD, ineffective inspiratory efforts and mechanical constraints, presumably, may prevent the development of rapid shallow breathing commonly observed in fail-to-wean patients in the acute setting (3, 5, 22). Therefore, even though the absence of rapid shallow breathing may predict weaning success in heterogeneous acute patients (3, 6), specific disease categories, e.g., long-term ventilated patients with COPD, may fail disconnection from the ventilator though exhibiting a low f/VT ratio. In these patients the documentation of an elevated P_0.1 and P_0.1/VT/TI could help to identify shortly after the start of the unsupported breathing trial those who are poor candidates for imminent discontinuation of mechanical ventilation.

All the 31 VD patients had to be reconnected to the ventilator within 60 min from the start of the unsupported breathing trial, and they could not be liberated from mechanical ventilation during their hospitalization (29 remained ventilator-dependent and two died). This fact might suggest that short periods of unsupported breathing can be very accurate in predicting whether the patient would ever be liberated from the ventilator. Although a recent ICU study suggests that a 30-min weaning trial may be as good as a longer 120-min trial to predict successful weaning (4), another study on a large population pointed out that short trials bear a high risk of reintubation rate (6). Furthermore, a controlled clinical trial showed that, in the ICU setting, patients who received some intervention had a reduced incidence of prolonged mechanical ventilation (5). Similar studies have not been performed yet in the chronic care setting, where long-term mechanically ventilated patients are concentrated. Hence, in our opinion, data from ICU patients cannot be extrapolated, and our observations on short- term spontaneous breathing trials in the chronic care setting, even if very appealing, deserve further investigation.

In summary, we report in this study physiologic measurements in patients who, after prolonged mechanical ventilation after either an episode of acute exacerbation of COPD or a phrenic lesion after cardiac surgery, could not be liberated from the ventilator and had to be enrolled in a home ventilator assistance program. Our data show that the major mechanism underlying this condition is the association between abnormal lung mechanics, in particular PEEPi and lung resistance in patients with COPD, and the reduced pressure-generating capacity of the inspiratory muscles because of either pulmonary hyperinflation (patients with COPD), among other factors, or phrenic nerve lesion (PCS patients). Under these circumstances the central drive is enhanced to defend ventilation but poorly transformed into inspiratory flow by the impaired inspiratory muscles, such that the breathing pattern results in a low tidal volume. The latter is inefficient to meet the metabolic demand and hence to clear CO₂. However, the high drive to breathe imposes the necessity to use a substantial portion (> 40%) of the maximum pressure-generating capacity of the   inspiratory muscles to sustain spontaneous ventilation. This makes unsupported breathing unbearable without excessive dyspnea after a short period, such that the patients have to be reconnected to the ventilator to prevent dramatic failure of the ventilatory pump. Furthermore, our data suggest that noninvasive measurement of P_0.1 and breathing pattern may help to identify why some patients are ventilator-dependent. This identification may have significant consequences: (1) frequent weaning attempts can be avoided, reducing undue stress on the ventilatory pump, and (2) aggressive therapy to reduce the ventilatory work load may be pursued.