Limited predictability of maximal muscular pressure using the difference between peak airway pressure and positive end-expiratory pressure during proportional assist ventilation (PAV)

Background If the proportional assist ventilation (PAV) level is known, muscular effort can be estimated from the difference between peak airway pressure and positive end-expiratory pressure (PEEP) (ΔP) during PAV. We conjectured that deducing muscle pressure from ΔP may be an interesting method to set PAV, and tested this hypothesis using the oesophageal pressure time product calculation. Methods Eleven mechanically ventilated patients with oesophageal pressure monitoring under PAV were enrolled. Patients were randomly assigned to seven assist levels (20–80%, PAV20 means 20% PAV gain) for 15 min. Maximal muscular pressure calculated from oesophageal pressure (Pmus, oes) and from ΔP (Pmus, aw) and inspiratory pressure time product derived from oesophageal pressure (PTPoes) and from ΔP (PTPaw) were determined from the last minute of each level. Pmus, oes and PTPoes with consideration of PEEPi were expressed as Pmus, oes, PEEPi and PTPoes, PEEPi, respectively. Pressure time product was expressed as per minute (PTPoes, PTPoes, PEEPi, PTPaw) and per breath (PTPoes, br, PTPoes, PEEPi, br, PTPaw, br). Results PAV significantly reduced the breathing effort of patients with increasing PAV gain (PTPoes 214.3 ± 80.0 at PAV20 vs. 83.7 ± 49.3 cmH2O•s/min at PAV80, PTPoes, PEEPi 277.3 ± 96.4 at PAV20 vs. 121.4 ± 71.6 cmH2O•s/min at PAV80, p < 0.0001). Pmus, aw overestimates Pmus, oes for low-gain PAV and underestimates Pmus, oes for moderate-gain to high-gain PAV. An optimal Pmus, aw could be achieved in 91% of cases with PAV60. When the PAV gain was adjusted to Pmus, aw of 5–10 cmH2O, there was a 93% probability of PTPoes <224 cmH2O•s/min and 88% probability of PTPoes, PEEPi < 255 cmH2O•s/min. Conclusion Deducing maximal muscular pressure from ΔP during PAV has limited accuracy. The extrapolated pressure time product from ΔP is usually less than the pressure time product calculated from oesophageal pressure tracing. However, when the PAV gain was adjusted to Pmus, aw of 5–10 cmH2O, there was a 90% probability of PTPoes and PTPoes, PEEPi within acceptable ranges. This information should be considered when applying ΔP to set PAV under various gains.


Background
Although mechanical ventilation is a crucial tool in decreasing the respiratory effort required by ventilated patients, diaphragmatic weakness can rapidly develop with complete diaphragmatic inactivity and mechanical ventilation [1]. This type of diaphragmatic powerlessness has been termed ventilator-induced diaphragmatic dysfunction (VIDD) [2]. Controlled mechanical ventilation is a major factor in VIDD, which may be attenuated with assisted ventilation [3,4]. This suggests that maintaining appropriate respiratory effort may be essential to preserving diaphragm function, and the ability to monitor respiratory effort during mechanical ventilation should be an important clinical issue [5].
Pressure applied to the respiratory system is usually assumed to dissipate against resistant and elastic elements. In a mechanically ventilated patient, the applied pressure is shared between the patient and ventilator [6]. This equation is difficult to solve under conventional ventilation because it is challenging to obtain reliable values for respiratory system resistance and elastance. However, in proportional assist ventilation (PAV), obtaining reliable elastance is possible during spontaneous breathing because the end of inspiration can be determined [7][8][9].
PAV with load-adjustable gain factors (PAV+) is a ventilatory mode that delivers assistance in proportion to the instantaneous flow and volume by calculating the instantaneous pressure needed to overcome the elastic and resistive pressures; these are updated several times per minute during PAV ventilation [10]. The proportion assistance is expressed as a percentage of the total pressure assisted (i.e. gain). By using this algorithm, Carteaux et al. [11] proposed a lookup table for estimating peak muscular pressure from peak airway pressure (P aw, peak ) and positive endexpiratory pressure (PEEP) difference (ΔP), thus offering a way to keep the patient in a predefined comfort zone by adjusting the PAV gain. However, this algorithm has not yet been validated [12].
The oesophageal pressure time product (PTP oes ) is a standard reference to assess respiratory muscle pressure. In patients with successful weaning, inspiratory PTP oes is usually <224-255 cmH 2 O · s/min throughout the weaning trial [13]. In addition to possible variability in respiratory elastance and resistance measured during PAV+, respiratory muscular PTP as estimated by Carteaux's method requires several assumptions that may limit its accuracy (e.g. a triangular muscular pressure waveform and a defined inspiratory time based on P aw, peak ) [11]. Thus, the derived muscular PTP may not be equal to the PTP oes . The present study aimed to verify the applicability of Carteaux's method with measured P mus, oes , P mus, oes, PEEPi , PTP oes , and PTP oes, PEEPi under different PAV gain settings.

Methods
From June 2014 to October 2014, all mechanically ventilated patients in our respiratory intensive care unit (10 beds) were screened daily for appropriateness for study inclusion. Patients had to be haemodynamically stable without inotropic agents and had to be ventilated with an inspiratory oxygen fraction <0.5 and PEEP ≤8 cmH 2 O. They also had to agree to oesophageal balloon placement. Exclusion criteria were pregnancy, acute coronary syndrome, aortic dissection as a cause of admission, and nasal or oropharyngeal lesions that prohibited oesophageal balloon placement. We used a single type of ventilator, the Puritan-Bennett 840 with PAV+ mode (Tyco International, Princeton, NJ, USA). The National Cheng Kung University Hospital Ethics Committee (A-BR-102-090) approved this study. The patient's next of kin gave informed consent.
The oesophageal balloon was placed in the lower third of the oesophagus and inflated with 0.5-1 mL of air. Airflow was measured via a pneumotachograph (PN 155362, Hamilton Medical, Bonaduz, Switzerland), while the airway and oesophageal pressures were individually measured using two differential pressure transducers (P/N 113252, Model 1110A, Hans Rudolph, Shawnee, KS, USA). The flow sensor was placed between the endotracheal tube and ventilator Y-piece. Tidal volume was obtained by integration of the flow signal. All signals were sampled and digitalized at 100 Hz, and data were stored in a data-acquisition system (AcqKnowledgement, Biopac MP150, Goleta, CA, USA). All patients were assessed in a 30°supine position with endotracheal suction performed before measurement if clinically required.
For individual patients, seven PAV gain levels (percentage of assistance), namely PAV20 (20% gain), PAV30, PAV40, PAV50, PAV60, PAV70, and PAV80, were randomly applied for 15 min at each level unless the patients showed discomfort. Respiratory mechanics measured by the ventilator during PAV were recorded throughout the course. Passive respiratory mechanics were measured under constant flow at the end of this protocol by increasing the back-up mandatory ventilator rate until all the breathing efforts were suppressed [13,14].

Physiological measurement
Validation of oesophageal pressure measurement Appropriate oesophageal balloon placement was verified by the occlusion test [15]. The ratios of change in oesophageal pressure to the change in airway opening pressure (ΔP oes /ΔP aw ) during three to five spontaneous respiratory efforts against a closed airway were determined to ensure oesophageal pressure measurement reliability.

Respiratory mechanics during PAV and passive mechanical ventilation
The respiratory mechanics (E pav and R pav ) during different PAV levels were recorded as a display on the ventilator screen. The last five E pav and five R pav at each PAV level were used for comparison. The respiratory system mechanics under constant flow and volume-cycled passive mechanical ventilation were determined at the end of the protocol using constant flow and a rapid airway occlusion technique [16,17].
Maximum inspiratory muscular pressure with P oes tracing (P mus, oes ) and inspiratory oesophageal pressure time product per breath (PTP oes, br ) Muscular pressure was calculated by taking into account dynamic E cw , which was obtained as the passive volumeoesophageal pressure slope [13]. P mus, oes was defined as the maximum difference between the passive and active P oes . The inspiratory PTP oes was calculated as the area between the P cw and P oes tracing, starting from the onset of inspiratory effort to the end of inspiratory flow. P cw was obtained by multiplying the tidal volume by dynamic E cw . The onset of inspiratory effort was determined by the rapid descent point from P oes . We calculated PTP oes with and without consideration of the intrinsic PEEP (PEEPi) [13]. Because gastric pressure was not measured, exact amounts of dynamic hyperinflation and expiratory muscle activity were unknown. The PTP oes was thus presented in two forms, the upper bound PTP oes , which attributes the rapid descent of P oes before the onset of inspiratory flow solely to inspiratory muscle activity, and the lower bound PTP oes , which attributes the rapid descent of P oes solely to cessation of expiratory effort [13,14]. PTP oes, PEEPi and PTP oes thus represent the upper and lower bounds of PTP, respectively ( Fig. 1).
Maximum inspiratory pressure from ΔP and PAV gain (P mus, aw ) and inspiratory pressure time product from airway per breath (PTP aw, br ) P mus, aw during PAV was obtained by using the formula adopted by Carteaux [11]: PTP aw, br was calculated under the assumption of a triangular inspiratory path with the end of inspiratory effort at P aw, peak .

Statistical analysis
The results are given as mean ± SD, unless otherwise specified. The Kruskal-Wallis test was used to compare means from different groups. Dunn's multiple comparison test was performed over pairs of groups. Repeated measured analysis of variance (ANOVA) was used to compare the means of E pav and R pav measured by the ventilator during various PAV gain levels. Correlatios between PTP oes, br and P mus, oes , PTP oes, PEEPi, br and P mus, oes, PEEPi , and PTP aw, br and P mus, aw were analysed using the two-tailed Spearman correlation test. Linear regression between PTP oes, br and P mus, oes , PTP oes, PEEP, br and P mus, oes, PEEPi , and PTP aw, br and P mus, aw was analysed with a forced regression line through the origin. Limits of agreement between P mus, aw and P mus, oes were examined using Bland-Altman analysis. All tests were two-sided, and a p value less than .05 was considered statistically significant. P cw, PEEPi P cw P Oes P mus, oes, PEEPi P mus, oes PTP oes PTP oes, PEEPi Fig. 1 Graphic illustration of flow, airway pressure (P aw ), and oesophageal pressure tracing (P oes ) during proportional assist ventilation. Chest wall recoil pressure (P cw ) was calculated from the product of tidal volume and dynamic chest wall elastance. Upper bound oesophageal pressure time product (PTP oes, PEEPi ) was calculated as the integration of the difference between P cw, PEEPi and P oes . Lower bound oesophageal pressure time product (PTP oes ) was calculated as the integration of the difference between P cw and P oes . P mus, oes and P mus, oes, PEEPi represent the maximal difference between passive and active P oes All analyses were performed using Prism version 5 (GraphPad Software, San Diego, CA, USA).

Results
The results of 18 consecutive patients who fulfilled the inclusion criteria were recorded. Two patients were excluded from further analysis because of a low ΔP oes /ΔP aw ratio. One patient was excluded because of a poor oesophageal pressure signal, and four patients were excluded because of an inadequate duration of P oes tracing secondary to the intolerance of the patients to low-gain PAV. Ultimately, 11 patients with an adequate duration of PAV recording at all stages of PAV support were analysed. The clinical demographics and respiratory mechanics of these patients are shown in Table 1. The tidal volume, P aw, peak , E pav , and R pav under various PAV gain levels are shown in Fig. 2. Significantly higher tidal volumes were found with high PAV gains. As predicted, P peak increased with PAV gain. There were no significant changes in R pav , but E pav was significantly higher with a high PAV gain (p < 0.0001).
PTP oes , PTP oes, PEEPi , peak muscular pressure and duration of inspiration (Ti) with different PAV gains and their correlation analysis PTP oes and PTP oes, PEEPi during various PAV gain factors are shown in Fig. 3. Progressive reductions in PTP oes and PTP oes, PEEPi were noted with increasing PAV gain levels. Significant differences were found among those with low-gain and high-gain PAV (p < 0.0001). However, no significant difference in PTP oes or PTP oes, PEEPi was found between PAV20 vs. PAV30, PAV30 vs. PAV40, PAV40 vs. PAV50, or PAV50 vs. PAV60. P mus, aw tended to underestimate P mus, oes or P mus, oes, PEEPi with all levels of PAV gain except PAV20 (Fig. 4a). The minimal difference between P mus, aw and P mus, oes was at the level of PAV30 (Fig. 4a). The T i, aw estimated from the onset of inspiratory effort to P aw, peak was not different from that estimated from flow tracing from PAV20 to PAV50. However, the T i, aw was significantly shortened compared to the Ti estimated from flow tracing within PAV60 to PAV80 (data not shown, p < 0.0001). Spearman correlation analysis revealed significant correlation between P mus, aw and PTP aw, br (r 2 = 0.9341), P mus, oes and PTP oes, br (r 2 = 0.8751), and P mus, oes, PEEPi and PTP oes, PEEPi, br (r 2 = 0.8862). Linear regression analysis disclosed the best-fit slope between PTP aw, br and P mus, aw to be 0.56, between PTP oes, br and P mus, oes to be 0.73, and between PTP oes, PEEPi, br and P mus, oes, PEEPi to be 0.83.
Bland-Altman analysis of P mus between P mus, aw and P mus, oes and selection of optimal P mus There was limited agreement between P mus, aw and P mus, oes as determined by Bland-Altman analysis (Fig. 4b).
The bias was -1.2 cmH 2 O. The 95% confidence interval  Table 2. PAV60 was associated with the highest probability (91%) of optimal P mus according to P mus, aw (5-10 cmH 2 O). However, the best PAV gain for optimal PAV assessed from P mus, oes or P mus, oes, PEEPi was quite diverse and was absent in two patients. The concordance rate for selection of optimal PAV gain was <50% between P mus, aw and P mus, oes and P mus, aw and P mus, oes, PEEPi . P mus,aw within 5-10 cmH 2 O was not present in PAV20 but was present in 11-82% of breaths in other PAV gains. Around 80% of breaths in PAV50 or PAV60 were associated with P mus,aw within 5-10 cmH 2 O. PTP oes <224 cmH 2 O·s/min and PTP oes, PEEPi <255 cmH 2 O·s/min are considered admissible according to Jubran et al. [13]. Despite the limited predictability of P mus, oes or P mus, oes, PEEPi from P mus, aw , patients with P mus, aw between 5 and 10 cmH 2 O are had 93% probability of PTP oes <220 cmH 2 O·s/min and 88% probability of PTP oes, PEEPi <255 cmH 2 O·s/min, regardless of the PAV gain. Only two breaths were associated with PTP oes values <40 cmH 2 O·s/min. When P mus,aw was achieved within 5-10 cmH 2 O, three PAV gain levels (PAV40, PAV50 and PAV60) were associated with >90% probability of admissible PTP oes and PTP oes, PEEPi .

Discussion
Our analyses revealed several interesting findings. First, PTP oes and PTP oes, PEEPi significantly decreased with increasing PAV gain in patients with PAV. Second, the prediction of P mus, oes or P mus, oes, PEEPi from airway pressure tracing had limited accuracy. Third, the deduction of PTP aw from ΔP may underestimate PTP oes or PTP oes, PEEPi . Fourth, an optimal P mus, aw (5-10 cmH 2 O) could be achieved in 91% of patients with PAV60, and despite the lack of accuracy for predicting P mus, oes or P mus, oes, PEEPi from airway pressure tracing, maintaining P mus, aw within 5-10 cmH 2 O was associated with PTP oes <224 cmH 2 O·s/min or PTP oes, PEEPi <255 cmH 2 O·s/min in approximately 90% of breaths.
The significant increase in P aw, peak but minimal difference in tidal volume with increasing gain level indicates substantial adaptation of muscular pressure during PAV [18]. The lower elastance during low assist could be explained by high respiratory drive (i.e. inspiratory muscle activity does not return to zero during the 300 ms occlusion time), which underestimates the elastic recoil pressure at end-inspiration. PEEPi is unlikely to be a cause Fig. 2 Tidal volume, peak airway pressure (P aw, peak ) and respiratory mechanics during proportional assist ventilation (PAV) under different gains. PAV20 indicates a mean gain level of 20%. Significant differences in tidal volume were found between PAV60 vs. PAV20, PAV70 vs. PAV20, PAV70 vs. PAV30, and PAV70 vs. PAV40. Significant differences in P aw, peak were found among individual P aw, peak levels under different gains, except the P aw, peak of PAV20 vs. P aw, peak of PAV30 and P aw, peak of PAV70 vs. P aw, peak of PAV80. For PAV-based patient elastance (E pav ), significant differences were found between PAV20 vs. PAV50, PAV60, PAV70, and PAV80; PAV30 vs. PAV50, PAV60, PAV70, and PAV80; PAV40 vs. PAV50, PAV60, PAV70, and PAV80; PAV50 vs. PAV70 and PAV80; PAV60 vs. PAV80; and PAV70 vs. PAV80. No significant difference was found in PAV-based patient resistance (R pav ) among various gains. For the E pav and R pav comparison, one patient was not included because of insufficient numbers of E pav and R pav in PAV20 and PAV30 because it did not increase with greater PAV assist in the current study [9].
The algorithm proposed by Carteaux et al. [11] is a simple bedside approach to estimate inspiratory muscular pressure (P mus, aw ) in mechanically ventilated patients under PAV. We found it to be of limited value in predicting P mus, oes . P mus, aw tends to overestimate P mus, oes in PAV20 but more commonly underestimates P mus, oes from PAV40 to PAV80. Therefore, the proportion of alleviation of respiratory muscle output was usually incompletely attained as the PAV gain intended it to be. Besides, the wide 95% confidence interval from the Bland-Altman analysis of P mus, oes and P mus, aw implicated that P mus, oes could not be accurately predicted by P mus, aw .
There are several possible explanations for these findings. First, for the unique condition where P mus, oes is usually overestimated in PAV20, a reasonable cause could be the ventilator flow control algorithm. Because respiratory effort is maximal in PAV20, the proportional-integral-derivative algorithm of the flow control system is prone to an airway pressure overshoot by the end of inspiration, which is further exaggerated fourfold in PAV20 for the calculation of P mus, aw [19,20]. Second is a possible discrepancy between PAV+ and CMV measured respiratory mechanics [10]. Although the PAV+ mode was continuously updated, measured respiratory system resistance and elastance may be different from those obtained under CMV [10]. Moreover, the respiratory system resistance measured by PAV+ is not reliable in cases with severe expiratory flow limitations. Third is the presence of PEEPi. In a recently published PAV+ mode bench study [21], the assistance provided by  Fig. 4 a Maximum muscular pressure (P mus ) determined using either oesophageal pressure tracing or airway pressure under different proportional assist ventilation (PAV) gains. Significant differences (p < 0.05) were observed for all gain levels. b Bland-Altman analysis plot showing bias and agreement between maximal muscular pressure calculated from ΔP and PAV gain (P mus, aw ) and maximal muscular pressure calculated from maximum difference between passive and active P oes without consideration of PEEPi (P mus, oes ). The middle dashed line is the mean difference (bias). The outer dashed line is the 95% confidence interval of the difference between P mus, aw and P mus, oes (±1.96 SD) Maximum muscular pressure and intrinsic positive end-expiratory pressure (PEEPi) were calculated as average of 1-minute breaths in each proportional assist ventilation (PAV) gain. Muscular pressures between 5 and 10 cmH 2 O are highlighted. P mus, aw maximal muscular pressure calculated from ΔP and PAV gain, P mus, oes maximal muscular pressure calculated from maximum difference between passive and active P oes without consideration of PEEPi, P mus, oes, PEEPi maximal muscular pressure calculated from maximum difference between passive and active P oes with consideration of PEEPi PAV+ was approximately 25% lower than expected. PEEPi with the associated trigger delay was considered a major factor affecting PAV+ accuracy due to the lack of assist during the initial part of respiratory breath, ultimately resulting in global under-assistance. PTP oes is a better surrogate of respiratory effort in ventilated patients. In this study, the analyses of correlation between P mus, aw and PTP aw , P mus, oes and PTP oes , P mus, oes, PEEPi and PTP oes, PEEPi yielded highly significant results. However, predicting PTP from P mus, aw and P mus, oes differed in the best-fit slope value. The slope value was 0.56 when the linear regression was performed between P mus, aw and PTP aw . The slope increased to 0.73 between PTP oes , br and P mus, oes and to 0.83 between PTP oes , PEEPi , br and P mus, oes, PEEPi . This implicates that the PTP aw should be corrected when projecting into PTP oes . We offer the following explanation for the discrepancy between PTP aw and PTP oes . First, the assumption of a triangular pressuretime product is flawed because respiratory muscle pressure generation is usually exponential [22][23][24]. The integration area above an exponential decay curve is usually larger than the integration area above a triangular line. Second, the inspiratory time is significantly shortened in high-gain PAV. The shortened inspiratory time should result in a smaller PTP aw from the triangular algorithm. A third possible cause is the influence of PEEPi. The algorithm proposed by Cardeaux et al. is also flawed as it does not consider PEEPi. The inclusion of PEEPi led to increases in P mus, oes, PEEPi and PTP oes, PEEPi .
The predefined range of respiratory effort by Carteaux and colleagues [11] needs critical appraisal. Target limits of P mus, aw within 5-10 cmH 2 O or PTP aw between 50 and 150 cmH 2 O·s/min were derived mainly from a desirable inspiratory effort of PTP oes, PEEPi <125 cmH 2 O·s/min [14]. This recommended threshold is arbitrary, not supported by quantitative diaphragm electromyogram, and possibly well below the threshold of threatening diaphragm fatigue [14]. A wider range of PTP oes, PEEPi should be allowable with minimal risk of diaphragm fatigue [13,25,26]. As P mus, aw frequently underestimates P mus, oes in the usual levels of PAV, actual PTP oes, PEEPi values are usually higher than PTP aw . Interestingly, PTP oes, PEEPi measurements were usually <255 cmH 2 O·s/min when P mus, aw were within 5-10 cmH 2 O. This implicates that the recommended grid table for PAV remains a helpful reference for selecting the PAV level, although the newly advocated threshold requires further study for verification.
There are several limitations to the current study. The first is the limited number of patients studied and the fact that all of the patients had started to have weaning trials as reflected by the oxygen fraction and external PEEP level. Thus, our results may not be applicable to acutely ill patients under mechanical ventilation. The second is the lack of gastric pressure measurement, which meant that we could not clarify the contribution of expiratory muscle activity during PAV. However, we did not notice evident abdominal muscle contraction during PAV except in one patient with high PEEPi. Thus, the measured P mus, oes, PEEPi should represent the inspiratory muscle motor outputs for most of our patients.

Conclusions
In summary, our results demonstrate limited accuracy of estimating respiratory effort from airway pressure tracing during PAV. Although P mus, oes decreases with increasing PAV gain, P mus, oes could not be precisely predicted from ΔP under various gain factors. In addition, PTP aw also underestimated PTP oes and PTP oes, PEEPi . However, when the PAV gain was adjusted to a P mus, aw of 5-10 cmH 2 O, there was approximately 90% probability of maintaining the patient within an acceptable PTP range.
Abbreviations CMV: continuous mandatory ventilation; E cw : passive chest wall elastance during CMV; E pav : PAV-based patient elastance; E rs : passive respiratory system elastance during CMV; PAV: proportional assist ventilation; PAV20 to PAV80: 20 to 80% PAV gain; P aw, peak : peak airway pressure during PAV; P cw : chest wall elastic pressure; PEEP: positive end-expiratory pressure; PEEPi: intrinsic PEEP; P mus, aw : maximal muscular pressure calculated from ΔP and PAV gain; P mus, oes : maximal muscular pressure calculated from maximum difference between passive and active P oes without consideration of PEEPi; P mus, oes, PEEPi : maximal muscular pressure calculated from maximum difference between passive and active P oes with consideration of PEEPi; P mus : respiratory muscular pressure; PTP: inspiratory pressure time product; PTP aw, br : inspiratory pressure time product calculated from ΔP and assuming a triangular inspiratory pressure time course per breath; PTP aw : inspiratory pressure time product calculated from ΔP and assuming a triangular inspiratory pressure time course; PTP oes, br : inspiratory pressure time product calculated from the difference between the oesophageal pressure and the relaxed chest wall elastance curve per breath; PTP oes : inspiratory pressure time product calculated from the difference between the oesophageal pressure and the relaxed chest wall elastance curve; PTP oes, PEEPi, br : inspiratory pressure time product calculated from the difference between the oesophageal pressure and the relaxed chest wall elastance curve per breath with consideration of PEEPi; PTP oes, PEEPi : PTP oes with consideration of PEEPi; R max : passive maximum inspiratory resistance during CMV; R min : passive minimum (airway) inspiratory resistance during CMV; R pav : PAV-based patient resistance; T i : duration of the inspiratory time determined from flow tracing during various PAV gains without consideration of PEEPi; T i, aw : duration of the inspiratory time determined from the peak airway pressure during various PAV gains; VIDD: ventilator-induced diaphragmatic dysfunction; ΔP: peak airway pressure and PEEP difference