Increase of Oxygen Consumption during a Progressive Decrease of Ventilatory Support Is Lower in Patients Failing the Trial in Comparison with Those Who Succeed

• ❖ An elevated oxygen consumption of respiratory muscles is associated with weaning failure

• ❖ Patients failing a weaning trial, in comparison with those who succeed, have a higher baseline oxygen consumption and are less able to increase it when ventilatory assistance is decreased

AFTER the acute phase of respiratory failure, when ventilatory assistance is reduced (for example, by reducing pressure support [PS] level) or discontinued, some of the patients are unable to sustain the necessary work of breathing and develop signs of fatigue, prompting the clinician to reinstitute ventilatory assistance. This occurrence, which has been termed “weaning failure,” can be caused by an increased mechanical load (reduced compliance, increased resistance, intrinsic positive end-expiratory pressure), or a decreased ability of the patient to generate pressure, or lack of endurance.1,2 

Whereas in normal subjects, the oxygen consumption (V̇o²) of respiratory muscles (V̇o^{2, resp}) does not exceed 5% of the total V̇o²of the body,3several studies demonstrated that the V̇o^{2, resp}in the patients being weaned from mechanical ventilation can be considerably higher,4–7and many authors, rather than focusing on the mechanical work of breathing, evaluated the role of V̇o^{2, resp}as a predictor of weaning success.5,8–11 

In several studies,5,6,8,12,13V̇o^{2, resp}has been computed as the difference between the V̇o²measured during controlled mechanical ventilation and during spontaneous (assisted) breathing. This approach assumes that, during controlled mechanical ventilation, the respiratory muscles of the patient, being fully relaxed, are passively displaced and that the changes in the body V̇o²during the transition from controlled to assisted breathing are caused solely by changes in respiratory muscle V̇o². However, such a condition might not be easily satisfied in the clinical setting, particularly because the changes in sedation level normally implemented during the transition from controlled ventilation would affect, per se , the total V̇o²of the body.

We therefore chose to follow a different approach, and we hypothesized that during a trial of decremental PS levels, a V̇o²increase would be more pronounced in patients eventually unable to sustain the PS decrease, indicating a greater V̇o^{2, resp}in these patients. We did not aim to develop an index able to predict a patient's weaning success (or failure); rather, we aimed to assess whether an increased V̇o^{2, resp}played any significant role in weaning failure, testing the hypothesis that when the PS level is reduced, V̇o²increases more in those patients unable to sustain the decrease in ventilatory assistance. To estimate a patient's V̇o², we used indirect calorimetry, a noninvasive and reliable method.14 

The protocol was approved by our institution's ethical committee (San Gerardo Hospital, Monza [MB], Italy); informed consent was obtained or deferred according to the committee recommendations. The study was performed in the eight-bed general Intensive Care Unit of a university hospital.

Patients were enrolled when being ventilated in PS for more than 24 h for acute respiratory failure and judged eligible for a test of weaning from mechanical ventilation by the attending physician.

Exclusion criteria were the following:

• Absence of resolution of the primitive disease(s)

• Hemodynamic instability requiring vasopressors

• Gas exchange impairment requiring positive end-expiratory pressure more than 10 cm H²O and/or fractional inspired oxygen tension more than 50% to obtain a Pao²of at least 80 mmHg

• PS level equal or greater than 20 cm H²O

• Core body temperature more than 38°C

After enrollment, patients were connected to an Engstrom Carestation (General Electric, Madison, WI) with respiratory parameters unmodified from those set by the attending physician. This ventilator is equipped with the M-CAiOVX “metabolic module”14(General Electric); the module consists of a fast differential paramagnetic oxygen analyzer, an infrared analyzer for carbon dioxide, and a pneumotachograph to measure inspired and expired volumes. The pneumotachograph and gas sampling ports are housed in a disposable connector, placed close to the patient, between the Y-piece of the ventilatory circuit and the endotracheal tube. The signals from the pneumotachograph and the gas analyzers are synchronized to allow breath-by-breath estimations of gas exchange. The device computes online the patient's V̇o²and carbon dioxide production.

The decremental PS trial (fig. 1) was performed by sequentially applying the following PS levels above positive end-expiratory pressure: 20, 16, 12, 8, 4, and 0 cm H²O and keeping each level for 10 min. At the end of the 10-min period, P0.1 (a validated index of respiratory drive15) was measured in triplicate. The decremental PS trial was stopped, and the PS was raised again to allow the patients to rest whenever they developed one of the following signs of respiratory distress:

• Respiratory rate more than 35 breaths/min

• Oxygen saturation measured by pulse oximetry less than 90%

• Heart rate more than 140 beats/min or variation more than 30% from baseline

• Systolic blood pressure more than 180 mmHg

• Increase in the end-tidal carbon dioxide more than 8 mmHg

• Dyaphoresis or anxiety

On the opposite, the PS level was immediately decreased (before the end of the 10-min period) to the next (lower) level if the patient developed signs of “overassistance” such as cough, tidal volume more than 15 ml/kg, or apnea for more than 15 s. At the end of the period at PS of 0 cm H²O (i.e. , continuous positive airway pressure) or after the patient had met the criteria of respiratory distress, the PS was switched to the level at which the lowest V̇o²had been observed, which was maintained for 10 min (“resting phase”). At the end of this phase, end-expiratory lung volume was measured by a 20% variation in inspired oxygen fraction.16Finally, the PS was set at the lowest level that the patient had been able to tolerate in the first phase, and this level was kept for 1 h, unless the patient developed signs of respiratory distress; in other words, the patient was kept on continuous positive airway pressure if this level had been tolerated without distress in the first phase (to verify, over a longer time span, the ability of the patient to breathe without ventilatory assistance) or at the level of PS above the one at which the patient had developed distress.

The trial was considered successful (success group) if a patient could breathe at continuous positive airway pressure for 1 h and failed (failure group) if, at any point, the patient developed any of the aforementioned signs of respiratory distress.

At the end of the trial, patients were switched to a volume assist/control ventilation for 10 min to determine V̇o²in such condition and to measure respiratory system compliance and resistance by means of end-expiratory and end-inspiratory pauses,17analyzing only the traces of the occlusions with a flat plateau on the airway pressure.

The Engstrom Carestation was connected to a personal computer, which recorded two types of data streams continuously.

• Continuous waveforms of airway pressure, flow, volume, carbon dioxide, and oxygen concentrations

• Breath-by-breath data on tidal volume, respiratory rate, V̇o²

Data from the last 3 min of each step were averaged, and we determined the lowest V̇o²observed during the trial (V̇o^{2, min}), and the corresponding PS level was indicated as PS^REST. Expiratory and inspiratory pressure time product18were calculated as described elsewhere.19Briefly, the pressure generated by respiratory muscles was calculated at any instant as the difference between the measured airway pressure and the theoretical airway pressure of a passive system:

Paw = positive end-expiratory pressure + flow*resistance + volume/compliance

where flow and volume are those actually measured at any instant, while resistance and compliance are measured during the phase of assist/control mechanical ventilation.

Data are indicated as mean ± SD or median (interquartile range). Variables between the two groups (failure or success) were compared by means of unpaired Student t  test or (for nonparametric variables, i.e. , Simplified Acute Physiology Score II, days on mechanical ventilation and PS^REST) Mann–Whitney U test or (for categorical variables i.e. , gender and mortality in the intensive care unit) chi-square test. Variables tested over different levels of PS were analyzed by a two-way ANOVA, having “group” (failure or success) and PS level as factors. If either the group or interaction effect resulted statistically significant we performed a post hoc  analysis comparing, at each PS level, the two groups (Holm–Sidak method). Association between two variables was assessed by linear regression. A level of P < 0.05 (two-tailed) was considered as statistically significant. Statistical analyses were performed by SPSS 16.0 for Windows (SPSS, Inc., Chicago, IL) and by SigmaPlot 11.0 (Systat Software, Inc., Chicago, IL).

We enrolled 28 patients, aged 67 ± 15 yr and ventilated for 30 ± 35 days. Initial causes of respiratory failure are reported in table 1.

Sixteen patients were able to complete the weaning trial successfully, and 12 failed it. Criteria determining failure of the trial were tachypnea in seven patients, excessive agitation in four patients, and increase of end-tidal carbon dioxide in one. Four patients failed the trial at a PS level of 8 cm H²O, two failed at 4 cm H²O, and the rest of the patients (i.e. , six) failed during the continuous positive airway pressure phase. Table 2reports the main clinical parameters of the two groups, collected in the baseline phase. No difference between the two groups could be observed except for a trend toward a lower compliance in the failure group.

Figure 2reports some examples of V̇o²as a function of PS level (see also fig. 1, Supplemental Digital Content 1, which is a figure showing the entire family of curves, https://links.lww.com/ALN/A597). In most patients, PS^RESTdid not correspond to the highest PS level; in other words, in these patients, a decreasing PS level correlated with an initial decrease in V̇o²and then an increase. Figure 3reports the frequency distribution of PS^RESTin the two groups of patients: PS^RESTwas higher in the failure than in the success group (17 ± 3.5 vs.  13.7 ± 4.2 cm H²O; P < 0.05).

Measurement of V̇o²was reproducible. A tight correlation was found between V̇o^{2, min}(i.e. , measured during the decremental PS trial) and V̇o²measured during the “resting phase,” at the end of the decremental PS trial (r²= 0.81, P < 0.001; slope 0.85; see fig. 2, Supplemental Digital Content 2, which is a figure showing the correlation between these two variables, https://links.lww.com/ALN/A598).

At PS^REST, V̇o^2,minwas higher in the failure than in the success group (215 ± 53 vs.  174 ± 44 ml/min, respectively; P < 0.05); on the contrary, no difference was observed in the maximum V̇o²obtained in the patients of the failure or success groups (271 ± 58 and 256 ± 65 ml/min, respectively). Accordingly, the absolute increase in oxygen consumption from V̇o^{2, min}to the maximum V̇o²observed was greater in the success than in the failure group (94 ± 71 vs.  52 ± 24 ml/min; P < 0.05).

Figure 4displays the variations of respiratory rate, tidal volume, and P0.1 after the reduction of PS: tidal volume did not differ between the failure and the success group but significantly decreased at decreasing PS levels (ANOVA: group effect, P = 0.062; PS effect, P < 0.001; interaction, P = 0.235). Respiratory rate was higher the failure group than in the success group and at lower PS levels (ANOVA: group effect, P < 0.001; PS effect, P < 0.001; interaction, P = 0.954); similar results were found for P0.1 (ANOVA: group effect, P < 0.001; PS effect, P < 0.001; interaction, P < 0.001).

The work of breathing, estimated from pressure-time product, increased more in the failure than in the success group, for a given decrease in PS (see fig. 3, Supplemental Digital Content 3, which is a figure showing the values of pressure-time product at different PS levels in the failure and success groups, https://links.lww.com/ALN/A599).

When the PS level was decreased below PS^REST, the V̇o²increase was more pronounced in the success group than the failure group (fig. 5; ANOVA: group effect, P < 0.001; PS effect, P < 0.001, interaction, P = 0.386), indicating that patients able to complete the weaning trial were those who reacted to the decrease of ventilatory assistance with a greater increase in V̇o².

This finding was confirmed when V̇o²was plotted as a function of P0.1 at different PS levels (fig. 6): the relationship between these two variables was steeper for the success than for the failure group, indicating that for the same increase in P0.1 the V̇o²would increase more in the success than in the failure group.

V̇o²recorded in the phase of controlled mechanical ventilation was not significantly different from V̇o^{2, min}for patients in the success group (197 ± 58 vs.  174 ± 44 ml/min, P = not significant), or in the failure group (196 ± 72 vs.  216 ± 53 ml/min, P = not significant) (see fig. 4, Supplemental Digital Content 4, which is a figure showing the comparison between V̇o²recorded in the phase of controlled mechanical ventilation and V̇o^{2, min}in the two groups of patients, https://links.lww.com/ALN/A600).

End-expiratory lung volume was similar between the success and failure groups (1,509 ± 803 vs.  1,665 ± 403 ml; P = not significant).

The main findings of this work can be summarized as follows: patients able to complete a weaning trial have a baseline V̇o²lower than patients failing the trial, and react to a decrease of ventilatory assistance (i.e. , to an increased load) with a proportionally greater increase in V̇o². The finding that V̇o²increased more in patients able to sustain the weaning trial than in those who failed contradicts our initial hypothesis and conflicts with some previous observations. A number of studies aimed at assessing the role of V̇o²during weaning from mechanical ventilation reported that an increased V̇o^{2, resp}is associated with weaning failure8,9,11,13,20; other studies did not find the measurement of V̇o^{2, resp}of any use in predicting the outcome of the weaning trial.4,5We offer the following explanation for the discrepancy between ours and previous findings. In the presence of an intact neuromuscular function, when the ventilatory assistance is reduced, V̇o²increases more in those patients who, because of a higher dead space and/or to higher resistance and elastance and/or to a greater minute ventilation, will develop a higher work of breathing. In this condition, thus, a higher V̇o^{2, resp}reflects a higher work required to breathe, and not surprisingly patients required to perform a higher work of breathing may fail their weaning trial. On the other hand, it has to be considered that if a patient is required to perform an increased workload we could assimilate the decrementing PS to a treadmill exercise: if the treadmill runs faster (i.e. , the PS is decreased) the patient will have to increase his/her V̇o²to be able to cope with the increased demand, otherwise he/she will fall from the treadmill (or fail the weaning trial). As a matter of fact, patients in the failure group, in spite of a large respiratory drive, increased as suggested by the P0.1 (they tried to run faster as the treadmill was running faster) could not respond to the P0.1 increase with a high enough increase in their V̇o², most likely for a deterioration in muscle function. At variance for the same P0.1 increase, success patients could develop a substantial increase in V̇o². This interpretation is somehow supported by sports medicine: highly trained subjects are characterized by a maximal V̇o²by far greater than normal subjects. Conversely, it has been shown that in patients with failing hearts the ability of consuming oxygen is greatly impaired in comparison with matched controls and that in such patients there is a close relationship between the maximal V̇o²and the muscle mass21,22; this finding holds true when referring specifically to respiratory muscles: maximum inspiratory strength is related to maximum V̇o²23. Adequate training results in an increase in V̇o²of healthy,24diseased subjects,25and even of isolated muscles.26In our study, respiratory muscles of patients failing a weaning trial might have reached their “lactate threshold,” shifting to anaerobic metabolism and hence becoming unable to sustain the effort for a prolonged period of time.27 

Most of the studies showing an increased V̇o²in the subjects who failed weaning were conducted on patients ventilated for a short period of time (often just 1 or 2 days): in these patients, likely to have an intact respiratory muscle function, a high V̇o^{2, resp}simply reflects a higher mechanical work of breathing while, probably, our patients ventilated for days had a decreased muscular efficiency and were not able to generate the necessary mechanical work. Interestingly, among the studies not confirming the finding that a higher V̇o^{2, resp}is associated with a weaning failure there is the paper by Hubmayr et al .,5which was conducted in a population of patients ventilated for several days. It should also be noted that patients in the failure group had a V̇o^{2, min}(i.e. , the minimal V̇o²observed at any PS level) considerably higher than that of the success group, which might have limited the possibility of further increasing the V̇o². Another factor that might aid to explain the discrepancy between ours and previous results is the different approach used: in a decremental PS trial we identified the level of support associated with the lowest V̇o², and we took this as a reference; noticeably, although in the failure group the V̇o²during assist/control mechanical ventilation (usually taken as a reference in previous works) decreased further, this was not the case for the failure group.

This study has a clinical implication because so far, a higher V̇o^{2, resp}has been described in the literature as a predictor of weaning failure, mainly as an indicator of an excessive work required from the patient to breathe (mirroring, in the example previously used the speed of the treadmill on which the patient is “running”). Far from challenging this solid evidence, our study proposes an additional mechanism for weaning failure: the inability of the patient to increase his/her V̇o²to generate the required work of breathing in response to the increased demand (mirroring the inability of the patient to run faster on the treadmill).

Another finding of the study was that, in most patients V̇o²increased when ventilatory assistance was increased above a certain level: increasing PS to a level higher than the patient's need can lead to an activation of expiratory muscles, to control excessive inflation and/or to promote exhalation. This was indeed confirmed by the presence of relevant expiratory pressure-time products at high pressure support levels; this was especially true in the success patients, characterized by more favorable respiratory mechanics and, likely, muscle function. It is, however, unknown whether this finding is caused by the specific design of this study and whether it can be translated to the general population of patients undergoing spontaneous assisted ventilation.

The study has some limitations. The population is relatively limited, and we did not perform a power analysis before conducting the study. Rather, also based on previous reports on this matter, we assumed that approximately 30 patients would have constituted an adequate sample size for a physiologic study that did not aim at assessing predictors of weaning, but simply at describing one of the mechanisms of weaning failure. In any case, the nonstatistically significant differences observed should be interpreted cautiously because of the possible lack of power of our study. We used a commercial device, based on the principle of indirect calorimetry to measure V̇o²in our patients; this system has the advantage of being noninvasive (requiring only a small connector at the circuit Y) and of providing continuous measurements of V̇o². In our study we found a good reproducibility of the V̇o²measurements obtained during two different study phases; the system has been validated mainly by comparisons with the previous Deltatrac monitor,14,28,29showing a good reliability. Although the different inspiratory flow profiles have not been found to affect measurement reproducibility,30no data are available on the potential effect of different tidal volumes or respiratory rates. Moreover, reports concerning comparisons with the reverse Fick methods, although promising, are limited.31The choice of maintaining each PS level for 10 min only was mandated by the necessity of keeping the protocol within a reasonable time, avoiding natural drifts of V̇o². Actual capability of the patients to breathe without assistance was, however, demonstrated over a longer time span (1 h). Patients were not instrumented with an esophageal balloon, but pressure-time product was derived from airway flow and pressure tracings.19Not surprisingly, baseline characteristics of the patients in the failure and success groups were not perfectly balanced, with patients in the failure group having a longer (although not significantly) duration of mechanical ventilation, a trend toward a lower compliance, and a higher baseline heart rate; this imbalance might help explain why patients in the failure group were unable to increase their V̇o²in response to the increased demand. Impending cardiac failure is a frequent cause of weaning failure, and it might indeed have contributed to the impossibility of the failure group to raise the V̇o²adequately; unfortunately advanced hemodynamic monitoring (such as Swan–Ganz or PICCO catheters) was not available in our patients to address this issue.

In this study we have shown that patients who fail a weaning trial have a higher baseline V̇o²and are less able to increase their V̇o²when reacting to a decrease of ventilatory assistance.