With pressure-controlled ventilation (PCV), the ventilator delivers flow in a characteristic way: Starting with high initial rates, flow (V̇) decelerates rapidly and, depending on the set tidal volume (VT), frequency and inspiration-to-expiration (I:E) ratio, reaches zero or near zero level after 30% to 50% of inspiratory time has elapsed. In the case of a partial obstruction of the endotracheal tube (ETT), initial V̇ cannot increase to the same level as with an unobstructed ETT, but full VT continues to be delivered because there is inspiratory no-flow time available that is now used for delivering V̇. To be able to continue delivering the full VT, a ventilator in PCV mode compensates for a partial ETT obstruction by extending the delivery time of V̇, if necessary until the end of the inspiratory phase that, with the unobstructed ETT, had previously been partially a no-flow phase. This compensatory mechanism is effective until inspiratory time becomes too short for the V̇ required for full VT. VT will not decrease until this compensatory mechanism is exhausted, thus there will be no indication of ETT obstruction.
In other words, ventilators are not constructed to monitor the patency of the ETT and the breathing circuit. Monitoring VT implies integrating V̇ during its entire delivery time. In contrast, monitoring for ETT patency implies analyzing how the V̇ profile changes with obstruction, and this analysis is not implemented in modern respirators. This explains why VT monitoring is so poor at detecting ETT obstruction (1). Because partial ETT obstruction causes added work of breathing for a patient on assisted spontaneous breathing (2,3), the importance of reliably monitoring ETT patency is only now becoming more obvious.
Tung and Morgan (1) demonstrated that in PCV, VT does not decrease until the cross-sectional area of the ETT is reduced by almost 70%. In volume-controlled ventilation, Ppeak does not increase until the cross-sectional area of the ETT is reduced by 20%–60%, depending on the inspiratory V̇ rate (4). An alternative to monitoring ETT patency with VT is monitoring the V̇ signal of passive expiration.
In anesthetized piglets receiving PCV, we produced different levels of concentric narrowing along the length of the ETT (simulating partial ETT obstruction) by replacing the initially applied ETT of 9.0 mm inner diameter (ID) with 3 ETTs of smaller IDs (8.0, 7.0, and 6.0 mm). A volume-over-flow curve was plotted, and the expiratory time constant τe (which is the first derivative of that curve or equivalent to resistance × compliance) was calculated at 15% of expiration time (Te). We also calculated the fractional volume expired during the first 15% of Te (Vex fract,15). We hypothesized that VT would not detect ETT narrowing, whereas an increase in τe and a decrease in Vex fract,15 would. Because changes in compliance of the respiratory system (Crs) also affect the expiratory flow signal, we studied the impact of Crs on the expiratory flow rate in a physical model.
The study was conducted in accordance with the Helsinki convention for the use and care of laboratory animals. The local Ethics Committee for Animal Experimentation reviewed and approved the study. We studied 11 healthy male and female piglets of Swedish Landrace breed (mean ± sd body mass, 24.5 kg ± 1.0).
After premedication (tiletamine 3 mg/kg and zolazepam 3 mg/kg; xylazine 2.2 mg/kg; atropin 0.04 mg/kg IM), anesthesia was induced with 100–200 mg ketamine and 1 mg/kg morphine and maintained with ketamine infusion (30 mg · kg−1 · h−1) plus morphine infusion (2 mg · kg−1 · h−1). Neuromuscular block was produced by continuous infusion of pancuronium bromide at a rate of 0.25 mg · kg−1 · h−1. Animals received sodium chloride 4.5 g/L with glucose 25 g/L at 10 mL · kg−1 · h−1 and a 5 mL/kg bolus of dextran-70 (Macrodex 70; Pharmacia Infusion AB, Uppsala, Sweden) to insure normovolemia.
The trachea of the anesthetized piglets was first intubated with an ETT of 9.0 mm ID (Mallinckrodt, Athlone, Ireland). Subsequently, the 9.0 mm ID ETT was removed and replaced by an ETT of 8.0 mm ID, then 7.0 mm ID and, finally, 6.0 mm ID (corresponding to 26%, 43%, and 55% decreases in cross-sectional area compared with an ETT of 9.0 mm ID, respectively). In this way, 3 standardized levels of a pure concentric “obstruction” over the whole length of the ETT were produced stepwise. Pressure difference to ZEEP (zero end-expiratory pressure) was set constant throughout, and VT, as well as the volume expired within the first 15% of Te, was calculated by flow integration. The animals were placed in the supine position and their lungs ventilated by a Servo 300 ventilator (Siemens-Elema, Solna, Sweden) in the PCV mode with a PEEP (positive end-expiratory pressure) of 0 cm H2O, Fio2 of 0.4, and a ventilatory frequency of 18/min. Paco2 was 3.8 kPa ± 0.5 (mean ± sd), and there was an I:E ratio of 1:1.
The expiratory volume-flow (V–V̇) curve was analyzed as follows: the first derivative of the expiratory volume-flow relation (dV/dV̇) was calculated for each of the data points. The resulting curve has the dimension of time, and each data point equals the expiratory time constant τe (compliance × resistance) at that particular lung volume. Any obstruction acts as a brake on expiratory V̇, increasing the time needed to empty the lung and thus increasing τe, for which higher values are read from the expiratory dV/dV̇. The higher the V̇, the more pronounced the “brake” effect of an ETT obstruction, which makes early expiration, with its high flow, the appropriate segment of the dV/dV̇ plot for detecting obstruction.
We decided to identify obstruction (i.e., the different IDs of the ETTs) by reading the value for τe after 250 ms of Te (which, at the frequency and I:E ratio chosen, corresponded to 15% of Te). The choice of 15% of expiratory time or 250 ms was based on the following: 50% of the expiratory VT in ETT with ID 9.0 mm was exhaled in the first 15% of Te. We decided to use this somewhat arbitrary value of 15% of Te as the cut-off point to define obstruction. To eliminate the effects of inertia and of the opening and closing of the ventilator valves on the τe analysis, we excluded the first 5% of the data points (corresponding to 80 ms of Te). The raw flow and volume data were smoothed in an Excel worksheet with a moving average of 20 data points.
Intrinsic PEEP was estimated from the difference between set PEEP and the pressure after a 5-s end-expiratory hold, and end-inspiratory pause pressure was also determined after a 5-s hold. Crs was calculated from VT divided by the end-inspiratory/end-expiratory pressure difference (intrinsic PEEP considered). Resistance of the respiratory system (Rrs), including the ETT, was calculated using the least squares fit technique (5,6).
Airway pressure was measured with a side-port at the site of the pneumotachograph and fed to a pressure transducer. V̇ was measured with a heated Fleisch pneumotachograph (Pneumotach Amplifier Series 1110; Hans Rudolph Inc, Wyandotte, MO) placed between the ETT and the Y-piece. We also collected the flow signal from the ventilator and compared it to the pneumotachograph signal. Signals were sampled at 2000 Hz and fed to a data acquisition system (Acqknowledge, version 3.2.7, BioPac Systems, Inc, Goleta, CA) and then exported to an MS Excel worksheet for offline data analysis.
The impact of Crs on τe was studied in a physical model described in the Appendix.
All values are expressed as mean ± sd. P values for difference in time-related variables were assessed using the paired Student’s t-test. Significance between variables was decided according to the false discovery rate procedure (FDR) (7).
There was a significant increase in τe at 15% of Te for ETT IDs 6.0 (794 ± 166 ms), 7.0 (635 ± 248 ms), and 8.0 (491 ± 125 ms) compared with ID 9.0 (380 ± 87 ms) (Table 1). There was no significant decrease of VT for the 3 steps of ETT narrowing (392 ± 45 mL for ID 6.0, 399 ± 41 mL for ID 7.0, and 398 ± 29 mL for ID 8.0, compared with 392 ± 31 mL for ID 9.0) (Table 1). On the other hand, there was a significant decrease of Vex fract,15 for ID 6.0 (134 ± 9 mL), ID 7.0 (146 ± 10 mL) and ID 8.0 (180 ± 14 mL) compared with ID 9.0(195 ± 17 mL) (Table 1). With ETT ID 9.0, 50% of VT was exhaled in the first 15% of Te, whereas with ID 6.0 it was only 34% of VT. Crs did not change significantly, whereas Rrs increased significantly with each step of ETT narrowing (Table 2).
There was good agreement between the flow signal from the pneumotachograph and the flow signal obtained from the ventilator (Fig. 1).
The main finding of this study was that τe of early expiration increased and Vex fract,15 decreased with each of the 3 steps of ETT obstruction. In contrast, VT did not decrease with decreasing ETT ID, although Rrs increased by 260% for ID 6.0 compared with ID 9.0 (Fig. 2).
Higher degrees of sudden ETT obstruction caused by kinking or blocking of the ETT represent a medical emergency and should be treated immediately. These situations are almost always detected by the traditional warning signs of ETT obstruction (i.e., decreased or zero volume delivered with PCV). In contrast, lower grades of obstruction caused by formation of biofilm or adherence of secretions to the inner surface of the tube may go unrecognized (1,4,8). It is important that simple tools be developed to identify these situations at the bedside so that patients can be treated promptly, thus reducing morbidity (9,10).
Recently, Visaria and Westenskow (11) presented a model-based method for detecting partial ETT obstruction. Conceptually, this method distinguishes between tube obstruction and acute bronchial constriction. The method is appealing because it is noninvasive and works in real time. However, it is first necessary to confirm that the rules for detecting tube obstruction, which are empirically derived from a pattern of parameter changes of the five-element lumped model of Eyles and Pimmel (12), are fulfilled under a broad range of clinical conditions. The latter model has been challenged by other researchers, however (13).
In previous research, we divided the expiratory volume-flow (V–V̇) curve into 5 slices and calculated τe values for each of the 5 slices, detecting partial ETT obstruction by an increase of τe in the early expiration and no or a lesser degree of increase towards the end of expiration (4,14,15). Compared with the “SLICE” approach, the model presented here avoids the assumption that τe is linear for narrow slice segments of the V–V̇ curve. One single value (τe at 15% of Te) instead of τe for 5 slices is indicated. Alternatively, the obstruction-induced deceleration of expiratory flow can be described by the finding that with ETT ID 9.0, 50% of VT was exhaled during the first 250 ms of expiration (15% of Te), whereas with ETT obstruction (ID 6.0) only one third of VT was exhaled in the same amount of time (Fig. 3).
The approach can be simplified even further: because V̇ must be the same throughout the whole system (in the absence of major leakage), good agreement between the ventilator’s and the pneumotachograph’s V̇ signal (Fig. 1) is to be expected. This suggests that the V̇ signal obtained from the ventilator could be as reliable for monitoring ETT patency as the pneumotachograph V̇ signal.
The elastic recoil of the respiratory system, expressed as V/Crs, drives V̇e. Increasing Crs decreases V̇e, to which any decelerating effect of an ETT obstruction would be added. As a consequence, τe (Rtot × Crs) increases, too. Decreasing Crs, by contrast, decreases τe. An obstruction effect could thus either be intensified or masked by changes in Crs, which we studied in more detail using a physical lung model (see Appendix). Indeed, a Crs change (both with Crs being constant during VT and with Crs changing with volume in a nonlinear fashion) obtunded the obstruction effect on V̇e (for details, see the Appendix)). Because secretions will produce a local asymmetrical obstruction rather than a symmetrical obstruction across the entire length of the tube, this difference was also studied. We found that a local obstruction, if it is pronounced enough, has the same impact on the pressure–flow relation as a symmetrical obstruction across the entire length of the ETT. This justifies the use of ETT with different IDs as an obstruction model.
In discussing the limitations of the approach presented, we would like to emphasize that we do not know of any simple breath-by-breath tool for monitoring ETT patency. Analysis of the V̇e signal provides a hint at ETT obstruction, and not even this hint is available at present with the methods falsely assumed to detect ETT narrowing (increase in peak pressure with volume-controlled ventilation or decrease in VT with PCV) (1,4). Second, we would like to emphasize that detection of ETT narrowing is in no way a single snapshot inspection of τe (and V̇e,max) alone. Rather, it is one of many steps required to integrate the τe analysis into the broader framework of continuous, breath-by-breath monitoring of dynamic respiratory mechanics. Narrowing will be detected by comparing a series of subsequent τe values to a starting point, which makes differences detectable by analyzing the trend. Third, we certainly do not claim that ETT narrowing can be detected solely based on information about V̇e. For a complete analysis, all other factors determining the course of V̇e must be known. This includes, most importantly, Crs, but airway resistance and inertance will also have an impact that remains to be studied in detail. However, using the information on changes in elastic recoil (V/Crs) combined with the information on τe, a simple scheme (Table 3) can be set up.
Disregarding simultaneous changes in airway resistance and inertance, 9 combinations of V̇e and elastic recoil changes are possible. With the first combination there is obviously no change. Combinations #2, #4, and #5 strongly suggest ETT narrowing, and combinations #3, #7 and #9 exclude ETT narrowing. In combinations #6 and #8, ETT narrowing is likely but additional information is required.
The simple scheme (Table 3) could serve as a starting point for a pattern recognition program for ETT narrowing. Such a program can be easily incorporated in a ventilator.
We conclude that with the elastic recoil of the respiratory system being appropriately considered, analysis of the expiratory flow signal and Vex fract,15 detects partial ETT obstruction during PCV, whereas VT monitoring does not.
Below, we discuss conditions other than symmetrical ETT obstruction that have an impact on τe: a) Crs change with intratidal Crs assumed to be constant; b) Crs change with intratidal Crs assumed to be volume-dependent in a nonlinear way; c) local asymmetric ETT obstruction; d) local asymmetric ETT obstruction plus nonlinear intratidal Crs change.
a) Crs Change With Intratidal Crs Assumed To Be Constant
A high V̇e rate is equivalent to a small τe. Because fewer steps are required to model the impact of Crs on V̇e compared with the Crs–τe relationship, we here approximate the Crs effect by calculating maximal V̇e (V̇e,max) as follows.
V̇e is decelerated by the resistance of the respiratory system (Rrs), the resistance of the ETT (RETT), and the resistance of the breathing circuit (Re). At onset of expiration, the elastic recoil pressure (VT/Crs) is at its maximal volume charge, thus making the elastic recoil pressure maximal and hence generating the maximal expiratory flow rate (V̇e,max). This can be modeled with the linear RCRETT-model of the respiratory system, on which V̇e,max is calculated according to the equation of motion (equation 1):
Assuming a VT of 250 mL, inserting the resistance values determined in this study for ETTs with ID 9.0 through 6.0 mm in equation 1, and plotting V̇e,max over Crs (Fig. 4), we find that down to a level of around 40 mL · cm H2O−1 Crs has little impact on V̇e,max, whereas with Crs decreasing still further, V̇e,max sharply increases. This implies that at low levels of Crs, its exact determination becomes even more important, as any further decrease in Crs considerably affects V̇e.
b) Crs Change With Intratidal Crs Assumed To Be Nonlinear
We modeled the effect of a nonlinear intratidal change in Crs using a physical lung model consisting of two connected chambers (Fig. 5a). Mechanical insufflation displaces water from chamber 1 to chamber 2 (16). Placing a wedge in chamber 2 (Fig. 5b) simulates a Crs decrease with increasing volume, the course of which was determined by measuring the pressure corresponding to stepwise displacements of 100-mL volume aliquots (up to 1000 mL) from chamber 1 to chamber 2 (Fig. 5c).
The lung model was ventilated in its linear and nonlinear configuration using ETTs with ID from 6.0 mm to 9.0 mm in both configurations, and τe was determined between 10% and 35% expiratory time. Inserting the wedge produces a lower mean compliance, thus with nonlinear volume-dependent compliance τe was smaller compared with linear compliance (Table 4) but significantly increased with decreasing ID, irrespective of the intratidal compliance course (Table 4). Assuming a nonlinear course of Crs being essentially unchanged while at the same time ETT becomes narrower, the obstruction could be inferred from increasing τe.
c) Local Asymmetric ETT Obstruction
To model local asymmetrical ETT narrowing, standard steel balls of different size were placed and kept in a mid-axial position in the ETT by a strong Neodym permanent magnet (Fig. 5d). As expected, effects were more pronounced with the obstruction acting symmetrically and along the entire length of the ETT (i.e., when changing the ID of the ETT) compared to an asymmetrical steel ball-induced obstruction that reduced the patent area (Apat) at a very small segment of the ETT only. We found that a 6.5 mm ball obstructing an ETT of ID 9.0 mm produced a pressure–flow (P–V̇) curve similar to that of an ETT of ID 8.0 mm (Apat 30.4 mm2), a 7.5 mm ball (Apat 19.4 mm2) simulated a 7.0 mm ETT, and an 8.0 mm ball (Apat 13.4 mm2) simulated a 6.0-mm ETT. The 6.5 mm diameter steel ball created a resistive pressure decrease (according to the Rohrer equation: Pres = V̇ · k1 + V̇ˆ2 · k2, equivalent to the airflow resistance at 1 L/s), which is similar to the symmetrical obstruction with ID 8.0 mm (Fig. 6). The 7.5-mm and 8.0-mm diameter steel balls created a higher resistive pressure decrease than their corresponding IDs of 7.0 and 6.0 mm, respectively, which is entirely attributable to the fact that the size of the commercially available steel balls increases in rather big steps, making it impossible to create exactly the same asymmetrical local ID reduction as produced by the symmetrical obstruction. However, these results show that a local obstruction, if pronounced enough, will have an impact on the P–V̇ relation that is identical to a symmetrical obstruction across the entire length of the ETT, justifying the use of different ETT IDs as an obstruction model.
d) Local Asymmetric ETT Obstruction Plus Nonlinear Intratidal Crs
The asymmetric obstruction was combined with a nonlinear volume-dependent Crs and the expiratory time constants were determined. Again, τe was lower compared with linear Crs (Table 4) but significantly increased with decreasing ID, irrespective of the intratidal Crs (Table 4).
1. Tung A, Morgan SE. Modeling the effect of progressive endotracheal tube occlusion on tidal volume in pressure-control mode. Anesth Analg 2002;95:192–7.
2. Rumbak MJ, Walsh FW, Anderson WM, et al. Significant tracheal obstruction causing failure to wean in patients requiring prolonged mechanical ventilation: a forgotten complication of long-term mechanical ventilation. Chest 1999;115:1092–5.
3. Brochard L, Rua F, Lorino H, et al. Inspiratory pressure support compensates for the additional work of breathing caused by the endotracheal tube. Anesthesiology 1991;75:739–45.
4. Kawati R, Lattuada M, Sjöstrand U, et al. Peak airway pressure increase is a late warning sign of partial endotracheal tube obstruction whereas change in expiratory flow is an early warning sign. Anesth Analg 2005;100:889–93.
5. Wald A, Jason D, Murphy TW, Mazzia VD. A computers system for respiratory parameters. Comput Biomed Res 1969;2:411–29.
6. Uhl RR, Lewis FJ. Digital computer calculation of human pulmonary mechanics using a least squares fit technique. Comput Biomed Res 1974;7:489–95.
7. Curran-Everett D. Multiple comparisons: philosophies and illustrations. Am J Physiol Regul Integr Comp Physiol 2000;279:R1–8.
8. Boque MC, Gualis B, Sandiumenge A, Rello J. Endotracheal tube intraluminal diameter narrowing after mechanical ventilation: use of acoustic reflectometry. Intensive Care Med 2004;30:2204–9.
9. Adair CG, Gorman SP, Feron BM, et al. Implications of endotracheal tube biofilm for ventilator-associated pneumonia. Intensive Care Med 1999;25:1072–6.
10. Feldman C, Kassel M, Cantrell J, et al. The presence and sequence of endotracheal tube colonization in patients undergoing mechanical ventilation. Eur Respir J 1999;13:546–51.
11. Visaria RK, Westenskow DR. Model-based detection of partially obstructed endotracheal tube. Crit Care Med 2005;33:149–54.
12. Eyles JG, Pimmel RL. Estimating respiratory mechanical parameters in parallel compartment models. IEEE Trans Biomed Engl 1981;28:313–7.
13. Jackson AC, Lutchen KR. Modeling of respiratory system impedances in dogs. J Appl Physiol 1987;62:414–20.
14. Guttmann J, Eberhard L, Haberthur C, et al. Detection of endotracheal tube obstruction by analysis of the expiratory flow signal. Intensive Care Med 1998;24:1163–72.
15. Guttmann J, Eberhard L, Fabry B, et al. Time constant/volume relationship of passive expiration in mechanically ventilated ARDS patients. Eur Respir J 1995;8:114–20.
© 2006 International Anesthesia Research Society
16. Guttmann J, Eberhard L, Fabry B, et al. Determination of volume-dependent respiratory system mechanics in mechanically ventilated patients using the new SLICE method. Technol Health Care 1994;2:175–91.