INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Bedside measurement of pulmonary mechanics during mechanical ventilation is important in assessing the status of the disease and in choosing appropriate ventilator settings. This was initially suggested by a retrospective analysis on 468 ventilated infants showing that pulmonary function testing during mechanical ventilation reduced common acute ventilation- associated complications such as pulmonary air leaks (1). Thereafter, randomized studies were performed in 53 ventilated adults (2) and 245 ventilated neonates (3). Amato and coworkers (2) provided the first demonstration that setting the ventilator according to the measured mechanics of the respiratory system can have an impact on the clinical outcome of adults with acute respiratory distress syndrome (ARDS) by improving survival rate and weaning rate, and by lowering the rate of barotrauma at 28 d. Stenson and coworkers (3) failed to show any substantial benefit in terms of adverse clinical events, duration of intubation, or duration of O₂ administration associated with regular measurement of static respiratory compliance when the whole population was analyzed. However, post hoc analysis found significant benefit in terms of reduced duration of ventilation (40%) in surviving neonates.

In spite of its potential role both for monitoring and for diagnosis, lung function testing is rarely used as a clinical routine in early childhood. Among possible reasons for this condition are the difficulty in measuring flow and pressure accurately, the complexity of the methods used for assessing the mechanical properties of the respiratory system, the lack of commercially available, well-validated systems, and the limited reference data (4, 5).

Ventilation-induced lung injury was until recently synonymous with clinical barotrauma. More recently, alterations in lung fluid balance, increase of endothelial and epithelial permeability, and severe tissue damage have been observed in mechanically ventilated animals. Increased tidal volume (VT) is responsible for ventilator-induced pulmonary edema (6) leading to the concept of volutrauma or high lung volume trauma (7). Low lung volume ventilation can also contribute to ventilation-induced lung injury by increasing shear stress in terminal units that are repeatedly opened and closed (8, 9).

Under conditions of muscle paralysis, the elastic properties of the respiratory system may be assessed by constructing the volume-pressure (V-P) curve obtained under conditions in which airflow is absent (static V-P curve). The static V-P curve of the respiratory system demonstrates a concave lower inflection point (LIP) in the initial segment of the inspiratory limb of the V-P curve at which shear stress occurs, followed by a straight segment of maximal compliance and a convex upper inflection point (UIP), at large lung volume above which static compliance decreases in response to alveolar overdistension (OD).

ARDS is characterized by a low respiratory system compliance (Crs) and reduced lung volume (10, 11). In patients with ARDS, a flattening of the V-P curve indicating OD occurs at much lower volume than in normal patients (12).

The static occlusion technique is the method of reference (13). As it is based on a large number of occluded breaths, recording and analysis are time-consuming. Recently Servillo and coworkers (17) developed an automated low-flow inflation (LFI) technique, which is relatively easy to apply at the bedside and allows one to build a quasi-static V-P curve.

To overcome the need to interrupt the regular ventilation for such maneuvers when one aims to detect OD, dynamic V-P curves obtained during standard ventilation have been analyzed (18). In adults, Ranieri and coworkers (18) showed that dynamic V-P curves obtained during constant flow inflation can provide the same information as the static V-P curve regarding the elastic properties of the respiratory system; in addition, they demonstrated that inspection of the dynamic V-P curve during period of constant flow allows one to detect OD. They described the shape of the curve by the nonlinear coefficient of a second- order polynomial equation (SOPE) fitted to the dynamic V-P curve (constant flow technique) and to the static V-P curve. A negative nonlinear coefficient indicated a decreasing slope with increasing volume, i.e., OD. In children, Fisher and coworkers (19) also used dynamic V-P loops to identify OD and estimated the decreasing changes in volume with progressive increase in pressure by calculating a ratio of the compliance calculated during the last 20% of inspiration (C20) to the compliance value for the entire inspiration (C). Children with OD had C20/C ratio less than 0.80.

The aims of the study were, first, to apply to a pediatric population the LFI technique allowing one to build quasi-static V-P curves; and second, to compare the detection of OD by dynamic and quasi-static V-P curves using two methods of analysis: the fitting of a SOPE describing the shape of the dynamic V-P curves compared with the fitting of a SOPE describing the shape of the quasi-static V-P curves (LFI technique); and the detection of OD by the dynamic C20/C compared with the UIP determined on the quasi-static V-P curve (LFI technique).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

During a 15-mo period, all patients admitted in the pediatric intensive care unit of our university hospital and requiring mechanical ventilation were possible candidates. Some patients were not included for the following reasons: ethical reasons, ventilation in assisted mode (inspiratory pressure support), presence of a drainage tube, or contraindication for sedation and paralysis with vecuronium. Permission was obtained from the parents.

Twenty-one children were included in the study but 20 children are presented: in one the analysis of the quasi-static V-P curve showed small inspiratory efforts, demonstrating that paralysis was not reached. The general characteristics of the patients and the ventilator settings are given in Tables 1 and 2.

Equipment

All measurements and calculations were performed using a portable computerized device for monitoring of respiratory parameters in mechanically ventilated children (Eolia; Dufour, Villeneuve d'Asq, France [23]).

Flows () were obtained by a grid Fleisch pneumotachograph (PNT), either a No. 00 (dead space 2 ml) or a No. 0 (dead space 4.7 ml), or a No. 1 (dead space 15 ml) that was added between the Y piece of the ventilator circuit and the endotracheal tube (ETT). The PNT was connected to a piezoelectric differential pressure transducer (± 12.7 cm H₂O, 163PCO1D36; Honeywell, Minneapolis, MN) whose sensitivity was reduced to ± 2 cm H₂O by analog amplifiers (TL 074 type; Texas Instruments, Dallas, TX) mounted to yield variable offset and gain. A first-order filter with a cutoff frequency at 100 Hz was added as an anti-aliasing filter to reduce signal-noise ratio. Volume variations were obtained by numerical integration of .

Pressure measurements were obtained at the pressure port placed just proximal to the PNT, and connected to a piezoelectric P transducer (± 70.3 cm H₂O; 162PC01D). and P signals were digitized at 256 Hz and synchronized with 0.0625° phase shift at 10 Hz (17.36 µs relative delay). Signals were "smoothed" using a second-order numerical filter. The V channel was calibrated using a calibrated syringe (Hans Rudolph, Kansas City, MO). The P channel was calibrated using an electronic manometer (Pic 400 Premier; Metratec, Houston, TX).

With tidal and P measured at the airway opening, -V and P-V loops were recorded during mechanical ventilation. They were frozen on the monitor. Crs was determined off-line using the inflation technique by the positioning of cursors to determine slopes on the frozen loops (24). The , P, and V data were automatically transferred to a table in spreadsheet program (Excel 5.0; Microsoft, Redmond, WA). This enabled display of the -time curve, the -V and V-P loops obtained in either dynamic conditions or quasi-static conditions (LFI procedure), to fit a SOPE to the data, to calculate the C20/C ratio, and to determine the UIP.

Procedure

All infants were intubated and ventilated in the volume-controlled mode with a Servo ventilator 300 (Siemens Elema AB, Solna, Sweden). Twelve children were intubated with a cuffed ETT. In these children the cuff of the ETT was inflated during measurements to prevent leakage. For all children, the absence of leaks in the circuit was ensured by verification of a stable airway pressure toward the end of a postinspiratory pause.

The ventilator settings, including the level of positive end-expiratory pressure (PEEP) chosen according to clinical requirements, were determined by the attending physician and were not modified before the beginning of the procedure. Patients were supine, sedated by a continuous infusion of midazolam. Suctioning was performed before the study, 15 min before recording.

After an injection of muscle relaxant (vecuronium 0.2 mg/kg), dynamic V-P loops were first obtained in the regular ventilation conditions. A total of 2,000 data points of , V, and P were recorded and Crs was calculated by the monitoring system.

Then, the LFI procedure was performed as described by Servillo and coworkers (17). The LFI technique is a low-flow insufflation performed in the volume-controlled mode set with a respiratory duty cycle (TI/ Ttot) of 0.5, a prolonged low-flow inspiration of 6 s (obtained by setting the minimal value of frequency of the ventilator, i.e., 5 breaths/min), and a VT chosen so as to result in an expected peak insufflation pressure of approximately 40 cm H₂O. For example, in a child with a compliance of 5 ml/cm H₂O, VT would be 5 × 40 = 200 ml and would be delivered at a flow rate of 200 ml/6 s = 33 ml/s. The relatively low flow rate reduces the resistive pressure component during insufflation. In some of the smallest children, when the first try did not reach the expected pressure and no UIP was detectable, we increased the LFI volume determined following the original technique in order to reach the expected peak insufflation of 40 cm H₂O.

To limit the waste of time (without ventilation) during the change from the volume-controlled mode of ventilation to the LFI ventilation, we used an original procedure: first we changed the ventilation mode for a pressure-controlled mode. Then we preset the ventilator in volume-controlled mode with the VT calculated as described previously (VT = Crs × peak inflation pressure of 40 cm H₂O), with the minimal value of frequency (5 breaths/min) and a TI/Ttot of 0.5 required by the technique. The upper pressure limit of the ventilator was set at 50 cm H₂O.

For recording of the V-P curve, at the end of an ordinary inspiration, the ventilator was switched on volume-controlled mode with the settings for LFI and PEEP brought to zero. This resulted in a prolonged expiration followed by a prolonged inspiration. A total of 2,000 data points of , V, and P were recorded during the insufflation and transferred to a table in Excel. After another prolonged expiration, volume and frequency were returned to previous levels.

Calculations

All calculations were performed on the , V, and P data transferred to Excel. Figure 1 illustrates the results of the analysis of dynamic and LFI V-P curves for Patient 7. 

View larger version (22K): [in this window] [in a new window]   Figure 1.   Dynamic (A and C ) and LFI (B and D) recordings of V-P curve obtained in Patient 7 that were analyzed by two techniques. (Panel A) The negative value of the dynamic nonlinear c coefficient of the SOPE fitted to the V-P curve during period of constant flow was indicative of OD. (Panel B ) The LFI c coefficient of the SOPE fitted to the V-P curve data points that were obtained within end-expiratory lung volume and baseline VT was also negative, indicating OD. (Panel C) The dynamic C20/ C ratio lower than 0.80 was indicative of OD. (Panel D) The volume of the UIP on the LFI V-P curve was 5.75 ml · kg1, i.e., included within the VT range for that child during regular ventilation, indicating OD.

Comparison of the shape of the dynamic V-P curve to the shape of the quasi-static V-P curve inferred from the nonlinear c coefficient of a SOPE fitted to the experimental data. For the dynamic V-P curve, a SOPE was fitted to the experimental data points obtained during constant flow inflation (steady-state portion of the dynamic V-P curve [18]) (Figure 2):  V = a + b P + c P², where coefficients a, b, and c are constants. Constant c is defined as dynamic nonlinear coefficient. A negative value of c indicates a convex shape with a decrease in slope with increasing inflation volume, i.e., OD as shown in Figure 1 (panel A) for Patient 7. A positive value of c indicates a concave shape, i.e., a progressive increase in slope with increasing inflation volume.

View larger version (13K): [in this window] [in a new window]   Figure 2.   Dynamic recording of flow-time curve (upper panel  ) and V-P curve (lower panel  ) during constant flow (dotted lines) obtained in Patient 14. The dynamic nonlinear c coefficient of the SOPE fitted to the V-P curve was negative (0.0187), indicating OD.

For the quasi-static V-P curve, a SOPE was fitted to the experimental data points obtained within end-expiratory lung volume (EELV) and baseline VT (18):  V = a + b Pdist + c Pdist², where Pdist is the recoil pressure in distal lung compartments, i.e., below the tip of the ETT and constant c is defined as quasi-static nonlinear coefficient or LFI c. For Patient 7 (as shown in Figure 1, panel B), LFI c was negative, indicating OD.

Comparison of the C20/C determined on the dynamic V-P loop to the determination of the UIP on the quasi-static V-P curve obtained with LFI technique. For analysis of the dynamic V-P curve, the C20/C ratio was calculated on the inspiratory part of the V-P loop obtained during mechanical ventilation as described by Fisher and coworkers (19). This is the ratio of the compliance calculated from the last 20% of the inspiratory V-P curve (C20) to the total compliance calculated from the entire slope of the inspiratory curve (C). C20 was calculated as follows: C20 = (VT  V_0.80Pmax)/(Pmax  0.80 Pmax), where VT is inspiratory tidal volume, V_0.80Pmax the volume at 80% of maximal inspiratory pressure, Pmax airway opening pressure at the zero point flow corresponding to end inspiration, 0.80 Pmax 80% of maximal inspiratory pressure, and C = VT/Pmax  Pmin where Pmin is airway opening pressure at the zero flow point corresponding to the beginning of inspiration. In Patient 7 (Figure 1, panel C ), C20/C was lower than 0.80 and was indicative of OD.

The analysis of the quasi-static V-P curve obtained with the LFI technique was performed as described by Servillo and coworkers (17) and allowed derivation of a curve of Pdist versus volume changes from the start of insufflation. ETT resistance (RETT) was calculated from the equation RETT = K₁ + K₂ where K₁ and K₂ are constants obtained from tabulated values for pediatric tubes (25). Product between airway flow (aw) and RETT was subtracted from each measured value of pressure to obtain tracheal pressure (Ptr). Respiratory system resistance (Rrs) was then calculated from normal breath as the quotient between area of the P-V loop and area of the -V loop. The product between Rrs and aw was then sample by sample subtracted from Ptr to give Pdist. The V-Pdist curve was plotted from data obtained from the low inflation breath. A straight line was drawn through the linear segment of the V-Pdist curve. To define the UIP, a straight line was drawn through the data above the linear segment of the V-Pdist curve. The two lines were extrapolated and crossed at a point defining the UIP. For Patient 7, as shown in Figure 1, panel D, the UIP determined on the LFI curve was 5.75 ml · kg¹, i.e., it was included in the VT range of that child, indicating OD.

Analysis

The analysis of dynamic V-P loops indicated evidence of OD when dynamic nonlinear coefficient c had a negative value (18) or when the C20/C ratio was < 0.80 (19). The analysis of the V-Pdist curve obtained with the LFI technique indicated evidence of OD either when static nonlinear coefficient c had a negative value (18) or when the UIP could be detected and the VT of the child included this UIP. The quasi-static V-P curve (LFI technique) was chosen as reference.

Results obtained from dynamic V-P loops were compared with the results of the LFI V-P curves. Qualitative agreement between results of different techniques was described by  values, and strength of agreement was evaluated following Landis and Koch (26). Spearman correlation coefficient between dynamic and quasi-static nonlinear coefficients was calculated (SPSS Base 8.0; Birmingham, UK). A p value  0.05 was considered as significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In the ARDS group of children, Crs ranged from 0.14 to 0.61 ml · kg¹ · cm H₂O¹. In the non-ARDS group, the Crs ranged from 0.13 to 1.31 ml · kg¹ · cm H₂O¹ (Table 2).

Detection of OD by the Dynamic versus LFI Nonlinear Coefficient c

Dynamic nonlinear coefficient c ranged from 0.0187 to +0.0233. LFI nonlinear coefficient c ranged from 0.0187 to +0.1404. In 18 children, both LFI and dynamic nonlinear coefficient c were obtained. Individual values of dynamic and LFI c are shown in Figure 3. Five children had negative dynamic and LFI nonlinear coefficients c indicative of OD and 11 children had positive dynamic and LFI nonlinear coefficients c indicative of the absence of OD. In two children (numbers 5 and 20) dynamic c was negative whereas LFI c was positive. In children numbers 8 and 15, dynamic and/or LFI nonlinear c coefficient could not be calculated because of a technical software problem. Detection of OD by dynamic nonlinear coefficient c agreed substantially with detection of OD by LFI nonlinear coefficient c on this sample of 18 children ( = 0.75). Shaded areas of the figure correspond to agreement of both techniques. Correlation between LFI and dynamic nonlinear coefficient was significant (r = 0.94, p < 0.01).

View larger version (40K): [in this window] [in a new window]   Figure 3.   Individual values of dynamic c and LFI c. Shaded areas correspond to individual values for which both analyses agreed. Qualitative agreement was substantial as indicated by the high  value. Dynamic c and LFI c disagreed in two cases.

Comparison of the Detection of OD by the C20/C Ratio versus the Detection of the UIP on the Quasi-static V-P Curve

Values of the C20/C ranged from 0.71 to 2.75. Individual values of dynamic C20/C and the determination of the UIP on the LFI curve are shown in Figure 4. A C20/C ratio < 0.80 was observed in two children (number 7 illustrated in Figure 1, panel C, and number 20). A C20/C ratio between 0.81 and 0.99 was observed in five children. It was  1.00 in 12 children. C20/C could not be evaluated in Patient 15 because of a technical software problem.

View larger version (18K): [in this window] [in a new window]   Figure 4.   Individual values of dynamic C20/C compared with the determination of the UIP on the LFI curve. Shaded areas correspond to individual values for which both analyses agreed. Qualitative agreement was poor, as indicated by the low  value. Dynamic C20/C and LFI analysis disagreed in four cases. C20/C: [compliance calculated from the last 20% of inspiratory V-P curve (C20) to the total compliance calculated from the entire inspiratory curve (C)].

An UIP was detected in 17 of 20 children (Table 3). In all but one ARDS patients the UIP could be detected. In these patients, the volume corresponding to this UIP ranged from 4.94 to 11.2 ml · kg¹. The Pdist corresponding to this UIP ranged from 13.09 to 35.06 cm H₂O (total pressure [Ptot] ranged from 14.98 to 44.87 cm H₂O). The UIP was included in the VT measured during the regular tidal ventilation in child number 2 whose C20/C was > 0.80 but < 1.00 (0.84), and in child number 6 whose C20/C was 1.10. 

The UIP could be detected in 12 of 14 non-ARDS patients. In these patients the volume corresponding to this UIP ranged from 5.75 to 18.49 ml · kg¹. The Pdist corresponding to this UIP ranged from 9.67 to 35.51 cm H₂O (Ptot ranged from 11.66 to 38.57 cm H₂O). The UIP was included in the VT measured during the regular ventilation in two children (child number 7 whose C20/C was < 0.8 (0.71) as illustrated in Figure 1, panel D, and child number 14 whose C20/C was > 0.80 but < 1.00 (0.83)). Dynamic detection of OD by the ratio C20/C agreed fairly with quasi-static detection of OD in the 19 children for whom we had both results ( = 0.22). Shaded areas of Figure 4 correspond to agreement of both techniques.

Comparison of the Analysis of the Quasi-static V-P Curve by Either the Graphic Determination of the UIP or the Fitting of a SOPE to the Data

Data were available for both analyses in 19 cases. As shown in Figure 5, the determination of the UIP and the LFI c agreed in all but one case ( = 0.85 indicating an almost perfect agreement). Indeed, in child number 3 the graphic analysis detecting the UIP did not indicate OD whereas the negative value of the nonlinear c coefficient was indicative of OD.

View larger version (13K): [in this window] [in a new window]   Figure 5.   Individual values of LFI c compared with the determination of the UIP on the LFI curve. Shaded areas correspond to individual values for which both analyses agreed. Qualitative agreement was almost perfect, as indicated by the very high  value. LFI c and the determination of the UIP disagreed in one case.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

These results show that (1) quasi-static V-P curves can easily be obtained in children with the LFI technique; (2) OD is not uncommon (suspected in 35% of children of this group) when ventilator settings are determined on a clinical basis; (3) the analysis of dynamic V-P curves by the fitting of a SOPE to the V-P data during constant flow inflation permits detection of OD and showed a good agreement with the results of the analysis of the quasi-static V-P curve; (4) finally, they confirm previous observations by Kano and coworkers (20) that C20/C index is probably an inadequate measure of OD.

The LFI technique was originally developed to build a quasi-static V-P curve in an adult population. These results show that it can be applied to a pediatric population and has the advantage to be standardized to the child's compliance. In this series, the LFI maneuver was well tolerated and no pneumothorax was induced by the maneuver.

However, some adaptations had to be made for the youngest children with lowest compliance. In these children, a complete V-P curve reaching sufficiently high airway pressure could not initially be obtained. This could be explained by the compressibility of gas in the connecting tubes acting as a proximal compliant compartment between the ventilator and the child (27). It could also be related to characteristics of respiratory mechanics in children. In infancy, chest wall compliance is higher than lung compliance and insufficient to passively maintain the EELV (28). Young children dynamically maintain their EELV above that volume determined by the mechanical properties of the system by using laryngeal adductor and diaphragmatic activities to retard expiratory flow (29). However, these strategies aiming at maintaining the EELV were probably lost with muscle paralysis. Furthermore, PEEP levels might have been too low to maintain EELV in some non-ARDS paralyzed children. Indeed, Thorsteinsson and coworkers (30) studying healthy children under anesthesia and paralysis observed that maximal Crs occurred at 5.6 ± 1.4 cm H₂O of alveolar pressure in infants and at 12.2 ± 1.4 cm H₂O in older children. Thus, as our PEEP levels were lower than those in some non-ARDS children, their EELV may have decreased below the LIP. In that case, their compliance and their LFI volume may have been underestimated and hence no UIP was detected. This could be assumed for three children younger than 1.5 yr of age (0.42, 1.42, 0.25 yr of age), and particularly for child number 12 who was ventilated without PEEP and in whom no UIP could be detected. Whatever the reason, if a suitable flow for LFI cannot easily be determined, there could be a major limitation in using LFI in small children with low compliance and particularly in neonates. Using a single inflation to trace the P-V curve may be important in order to standardize the events immediately preceding the recording as they may influence the state of the lung when the P-V curve is obtained (31).

Several parameters determined from the V-P curve have been proposed to guide the adjustment of ventilator settings. Attention has focused on the determination of the LIP. Pulmonary lesion may be aggravated if this LIP lies within the VT by repeated collapse and reopening of terminal units. These lesions may be lessened by stabilizing terminal units with PEEP above the LIP (8, 9).

The identification of the UIP as a marker of OD is another information obtained by the analysis of the V-P curve. As the aim of the study was to detect OD in the regular ventilation conditions determined by the attending physician, dynamic V-P loops were first obtained in these ventilation conditions. The LFI V-P curves were then recorded using for all children the procedure described by Servillo and coworkers (17), i.e., from zero end-expiratory pressure (ZEEP). Jonson and coworkers (32) have shown that in acute lung injury a single expiration to ZEEP leads to lung collapse. During the following LFI from ZEEP lung recruitment happens far above the LIP of the P-V curve and is associated with a high compliance. As dynamic V-P curves were obtained under the best clinically determined PEEP, and LFI V-P curves under ZEEP, there could be discrepancy in lung volume or maximal compliance as described by Jonson and coworkers (32). But our study aimed at determining the UIP and at detecting OD. Jonson and coworkers have clearly shown that LFI V-P curves recorded from ZEEP and from PEEP tend to merge at high pressure at the same UIP and that the analysis of the UIP should be valid.

Rodriguez and coworkers (33) obtained V-P curves by the LFI technique using the standard sensors of the Cesar ventilator for P and V measurements and collecting them on the Cesar ventilator screen, thereby reducing the time required for the maneuver. This attractive way of obtaining V-P curves cannot be applied to the Servo 300 ventilator whose screen does not provide a sufficient accuracy of reading. Moreover, in small children measurements must be performed at the Y piece of the circuit because the volume measured by the sensors of the ventilator can be overestimated compared with volume measured at the airway opening (34).

We measured the V-P curve of the respiratory system and not of the lung alone. As a result, our V-P curves necessarily reflect the contribution of the chest wall. As the patients had no chest wall abnormalities and were paralyzed, the role of the chest wall when measuring total respiratory system was minimized. Moreover, Suter and coworkers (12) and Roupie and coworkers (35) by measuring respectively lung or chest wall compliance together with total respiratory compliance reported that the reduction of total respiratory compliance measured at high VT was fully attributable to change in lung compliance, i.e., that the determination of the UIP on the respiratory V-P curve was valid. However for some of our children without ARDS and having abdominal problem (numbers 14, 16, and 18), increased intra-abdominal pressure may have altered respiratory mechanics by decreasing chest wall compliance. In these patients, the volume of the UIP may have been underestimated by the respiratory V-P curves (36).

The LFI quasi-static V-P curve was analyzed by a graphic and a mathematical technique. They agreed in all but one child: in child number 3, the graphic analysis determining the UIP did not indicate OD whereas the negative nonlinear c coefficient was indicative of OD. A slight difference in analysis of the curvature of the V-P curve by the two techniques could be responsible for this disagreement. Indeed, the graphic analysis detects an UIP corresponding to the middle of the curvature of the V-P curve, whereas the nonlinear c coefficient becomes negative as soon as a curvature exists. In child number 3, the UIP (11.13 ml · kg¹) was located just above the VT (10.26 ml · kg¹). In this case, the beginning of the curvature could have been detected by the nonlinear c coefficient within the VT while the middle of the curvature taken into account by the graphic analysis was just above the VT.

The pressure corresponding to the UIP of our children with ARDS was similar to that reported for adult patients with ARDS (17, 35). The volume of the UIP of our ARDS children also was similar (17, 35) to that reported for adult patients with ARDS. It was similar or smaller than the volume of the UIP of our children without ARDS. This is consistent with reduced lung volume and earlier flattening of the V-P curve in ARDS (10). Higher volume of the UIP in adult patients with acute respiratory failure (ARF) compared with ARDS patients was also observed by Servillo and coworkers (17).

The pressure corresponding to the UIP of our non-ARDS children was lower than pressure of the UIP of our children with ARDS. The same trend was observed by Servillo and coworkers (17) when using the LFI technique to build static V-P curve. In this pediatric group of patients, the level of pressure of the UIP was, however, lower than that reported in adult patients with ARDS (17, 38) and particularly in our youngest patients (Patients number 6, 17, and 18). These results are related to changes in P-V relations during growth (30). Indeed, the transition between level and steeper parts of the P-V curve observed in healthy children during muscle paralysis, i.e., the UIP, occurs at much lower pressure in infants than in older children.

This study shows that OD can be detected on dynamic V-P curve by the constant flow technique. It also confirms previous observations by Kano and coworkers (20) that the C20/C index is probably an inadequate measure of OD. Indeed, the C20/C ratio using Fisher's criteria (C20/C < 0.80) could detect OD in only one of four cases. Three overdistended children had a C20/ C > 0.80, two > 0.80 but < 1.00, i.e., falling within a range of values on which no conclusion could be drawn from the study of Fisher and coworkers (19). These investigators, analyzing the shape of the V-P curves of mechanically ventilated neonates and defining overdistended curves by visual inspection, did not observe any overlapping of C20/C of the nonoverdistended and the overdistended populations: the former having C20/C > 1.00, the latter < 0.80. Our results suggest that larger ranges of values of C20/C ratio are observed in overdistended V-P curves and that overlapping of the C20/C values of the two groups may be observed. The same conclusion could be drawn from the study of Kano and coworkers (20) in mechanically ventilated animals who used, as we did, quasi-static V-P curves to define the overdistended population. Moreover, Kano and coworkers (20), evaluating different indexes of OD (C20/C and the contribution of the volume-dependent elastance to the total dynamic elastance of the respiratory system [%E₂], had already pointed out that the C20/C ratio is affected by the mode of ventilation and by the pressure required to overcome the pressure drop across the ETT, airways, and tissues of the respiratory system. This resistive pressure can be recognized visually as contributing to the width of the dynamic V-P curve. After subtraction of the resistive pressure drop across ETT from airway opening pressure there is a systematic decrease in C20/C. The difference is most marked for the smaller ETTs. On the contrary, the analysis of the V-P curve during a period of constant flow, i.e., after the immediate step change in airway pressure due to the resistive component of the respiratory system, reflects the compliance of the respiratory system (39). The analysis of this limited part of the V-P curve is not affected by resistive pressure and, therefore, seems to be a more useful tool for estimating OD.

The constant flow technique seems to be an adequate means of assessing OD as qualitative agreement with LFI analysis was substantial and as dynamic nonlinear c coefficients were highly correlated to static nonlinear c coefficient. It confirms the results of Ranieri and coworkers (18) and extends them to a pediatric population. However, the absolute values of our nonlinear coefficients used to describe the shape of the static and dynamic V-P curves were greater than the coefficients observed in adult patients by Ranieri and coworkers (18). This is consistent with a steeper P-V curve and smaller amplitude of the P-V curve (expressed in ml · kg¹) with decreasing age (30).

In two children (numbers 5 and 20) dynamic OD was detected by dynamic C20/C and/or dynamic nonlinear c coefficient but not by any analysis of the quasi-static V-P curve. Dynamic data may also reflect the contribution of heterogeneous time constants to lung OD. At a given respiratory rate, lung units with shorter time constants will receive proportionally more ventilation than lung units with relatively long time constants, possibly contributing to regional overinflation (40). Because changes in airway pressure reflect an average response of the respiratory system to changes in lung volume, the resultant OD of these lung units could be reflected in the dynamic C20/C and dynamic nonlinear coefficient. On the contrary, during quasi-static measurements equalization of pressure between lung units with different time constants is reached. It is therefore possible to observe OD on dynamic measurements while the lungs are being ventilated on the linear portion of their static V-P curve.

Derivation of %E₂ from a dynamic V-P curve (20, 22) allows the analysis of many used flow or pressure patterns but requires a number of mathematical steps. Therefore, it currently remains a research tool. On the contrary, the constant flow technique has the advantage of requiring no complex mathematical tool for the analysis of the V-P curve. Indeed, once the data have been transferred to a spreadsheet program such as Excel, a SOPE can be fitted to the V-P data and the nonlinear coefficient will indicate the existence of OD. Because the nonlinear coefficient is obtained from dynamic V-P data, this technique could be used to monitor lung OD continuously and to optimize volume-controlled ventilation.

In conclusion, quasi-static V-P curves can easily be obtained in children using the LFI technique. The original method that we used does not require any special device, contrary to the protocol of Servillo and coworkers (17), and it limits the time without ventilation. The SOPE offers a good detection of OD on dynamic and LFI V-P curves. This analysis can easily be performed provided that V, P, and data are transferred to a table in a spreadsheet program such as Excel. Further evaluation of this measure of OD must assess whether it reflects changes in ventilator strategy. If this is the case, a more automated means of analysisprovided, for example, by ventilator manufacturerswould increase the clinical applicability of this method.

Provided that a child's lung volume is maintained above the LIP in order to avoid low lung volume trauma, the detection of OD on dynamic V-P (constant flow) by this mathematical analysis could be used routinely to optimize ventilator settings and to avoid OD and barovolutrauma. However one should remember that in the presence of nonhomogeneous distribution of ventilation, dynamic curves might indicate apparent OD while the lungs are being ventilated on the linear portion of their static V-P curve.