Measurement of Pressure–Time Product during Spontaneous Assisted Breathing by Rapid Interrupter Technique

We enrolled eight nonconsecutive patients admitted to our intensive care unit, with acute respiratory failure from various primary etiologies (table 1) and undergoing PSV. The investigational protocol was approved by our institution’s ethical committee (Università degli Studi Milano Bicocca, Ospedale San Gerardo, Monza, Italy); informed consent was obtained according to the committee recommendations.

We used an Evita 4 ventilator (Draeger Medical, Lubeck, Germany) controlled by a personal computer through the serial communication port (MEDIBUS protocol).

Once enrolled in the study, the patients were suctioned if necessary, while their ventilatory settings were kept unmodified from those set by the attending physician (table 1). Throughout the study procedure, the patient’s electrocardiogram, arterial invasive pressure, peripheral oxygen saturation (Spo²), and expired carbon dioxide partial pressure (ETco²) were continuously monitored. An esophageal balloon was placed in the distal third of the esophagus and secured, after correct placement was assessed according to standardized procedures.17The absence of intrinsic positive end-expiratory pressure (PEEP) was carefully checked at enrollment and throughout the study procedure by the simultaneous evaluation of the esophageal and airway pressure tracings. After a stabilization period, in the first 5 min of undisturbed PSV ventilation, we measured the average tidal volume (V^T) of the patient. The V^Twas then divided in 25-ml intervals.

One occlusion maneuver (see below) was then programmed at each of these target volumes, in random order, to encompass the entire V^Trange, during both the inspiratory and expiratory phases. After an occlusion command, when the target volume was reached, the ventilator closed the inspiratory or expiratory valve as rapidly as possible. After 2 s, the occlusion was released, and the ventilator was switched back to the baseline ventilation. A safety mechanism would immediately open the expiratory valve if, at any time, the airway pressure exceeded 40 cm H²O. Twenty to 30 undisturbed breaths were allowed before the next occlusion of the programmed sequence was performed.

In three patients, at the end of the measurement series, we increased the pressure support level by 5 cm H²O and performed a new measurement sequence. We thus obtained a total of 11 measurement sequences.

Airflow was measured by a pneumotachograph (Fleisch No. 2; Lausanne, Switzerland), connected to a differential pressure transducer (Baroscan PC-100; Hoffrichter, Germany). The signal was calibrated at the inspired oxygen fraction in use. The volume was obtained by integration of the flow signal. Airway and esophageal pressures were measured by an F-497 transducer array (Special Instruments, Nordlinger, Germany). A voltage signal, switching from 0 V during the normal ventilation to +5 V during the occlusion, was also recorded, to provide a precise marker in the waveform file for the detection of the occlusions. All of the signals were digitized with a frequency of 125 Hz (NI-DAQ; National Instruments, Houston, TX), displayed on a personal computer monitor in real time, and stored on the hard drive for the off-line data analysis.

The rapid interrupter technique relies on the principle of the pressure equilibration within the respiratory system in the absence of flow. Figure 1shows a representative inspiratory occlusion maneuver. Immediately after the occlusion (vertical dashed line), the flow ceases, and, ideally, the airway pressure equilibrates with alveolar pressure. Further changes in airway pressure are due to the persisting patient’s respiratory effort; finally, the effort ceases and, if the patient relaxes his or her muscles, a plateau develops on the airway pressure tracing (P^{pl, aw}). However, because the occlusion is rapid but not instantaneous, and because the occlusion gives rise to a pressure noise in the signal for a period of 50–100 ms, the value of alveolar pressure at the time of occlusion (P^{ALV, occl}) cannot be read directly and must be back-extrapolated, by linearly fitting the points from the most linear portion of P^awafter the noise ceases (fig. 1, solid line), back to the moment of the interruption.18 

The pressure generated by patient’s respiratory muscles (P^{MUSC, occl}) was calculated as P^{MUSC, occl}= P^{el, rs}− P^{ALV, occl}, where P^{el, rs}represents the elastic recoil of the respiratory system (i.e. , at any time, the alveolar pressure equals the sum of the elastic recoil pressure of the respiratory system and the pressure generated by the respiratory muscles). For each occlusion, on the assumption of a linear compliance throughout the V^Trange, P^{el, rs}= (V^DEL/C^rs) + PEEP, where V^DELrepresents the volume effectively delivered to the patient from the beginning of inspiration to the time of the occlusion, and C^rsrepresents the respiratory system compliance.

The value of C^rsin each patient was obtained averaging the values of V^DEL/(P^{pl, aw}− PEEP) for those occlusions in which a flat plateau (at least 500 ms long) in the airway pressure tracing was seen. During expiration, the same equations were applied.

After the experimental protocol was completed, the tracings of airway pressure, flow, and volume during each occlusion were analyzed with dedicated software. Whereas one marker was automatically placed at the beginning of the occlusion, we manually placed one marker at the beginning of the inspiration (defined as the beginning of the deflection in the P^awtracing), two markers encompassing the P^awtracing portion to be fitted after the occlusion, and two markers encompassing the plateau phase, if present. Provided with these markers, the software performed the aforementioned calculations. The time elapsed between the beginning of the inspiration and the occlusion (t^occl) was also measured.

We then plotted each value of P^{MUSC, occl}over its respective t^occl, obtaining the time course of the inspiratory effort, sampled over different breaths at different times and volumes. The integral of P^{MUSC, occl}over the duration of inspiration is analogous to the PTP and was computed with the trapezoidal rule.

The maximum P^{MUSC, occl}value observed was also recorded (P^{MAX, occl}).

Esophageal pressure measurements were used as reference. From the occlusions showing a flat plateau in the esophageal pressure tracing (P^{pl, pes}), chest wall compliance was calculated as C^cw= V^DEL/(P^{pl, pes}− PEEP^pes). The expected chest wall elastic recoil pressure (P^{el, cw}), for each occlusion, similarly to the above-mentioned calculation, was computed as P^{el, cw}= (V^DEL/C^cw) + PEEP^pes. Therefore, the alveolar pressure at the time of the occlusion can be calculated as P^{ALV, pes}= P^{occl, pes}+ P^{el, l}, where P^{el, l}is the elastic recoil pressure of the lungs (equal to P^{el, rs}− P^{el, cw}), and P^{occl, pes}equals the esophageal pressure at the time of the occlusion. P^{MUSC, pes}was finally calculated as P^{occl, pes}− P^{el, cw}.

The pressure–time product (PTP^pes) was calculated as the area subtended by P^esand the chest wall static recoil pressure, averaging 10 undisturbed V^Ts. From the same V^Ts, the maximum deflection in P^es, another index on respiratory work, was obtained and averaged (P^{MAX, pes}).

With the ventilator connected to a passive model of the respiratory system (compliance, 25 ml/cm H²O; resistance, 12 cm H²O · l⁻¹· s⁻¹), we performed a series of occlusions during volume-controlled mechanical ventilation. We used three different levels of PEEP (5, 10, 15 cm H²O) and three different flow rates (V̇= 0.5, 0.75, 1 l/min) to investigate the influence of the airway pressure and flow rate on the valve behavior. For each combination, we performed interruptions along a V^Tof 500 ml, during both the expiratory and the inspiratory phases. Because the model exhales passively, performing occlusions at different volumes during expiration allowed us to obtain different flow rates at the moment of the occlusions.

Data are expressed as mean ± SD unless otherwise specified. Relations between two series of variables were assessed by linear regression. Ninety-five percent limits of agreement between esophageal- and occlusion-derived variables were calculated as suggested by Bland and Altman.19 

The average time closures for the inspiratory and expiratory valves were 74 ± 10 and 61 ± 13 ms, respectively. Both valves exhibited longer closure times at increasing flow rates (t^cl= 13 ·V̇+ 64, r = 0.48, P < 0.01 and t^cl= 18 ·V̇+ 47, r = 0.46, P < 0.001 for inspiratory and expiratory valves, respectively). The volumes delivered during the closure of the valves averaged 34 ± 17 and 26 ± 16 ml, and they positively correlated with the closure time of the valve (r = 0.48 and r = 0.46, P < 0.01, data now shown). We could not disclose any effect of the airway pressure level on the valve closing time.

A total of approximately 400 occlusions were obtained in the 11 measurement series. Measurement series were well tolerated in all patients without the development of desaturation or hemodynamic instability. A mild increase in the respiratory rate (21 ± 5%, P < 0.05) was observed between the beginning and the end of the measurement sequence.

Figure 2shows the individual relations between V^DELand (P^{pl, aw}− PEEP) for the occlusions where a flat plateau had been identified. For each patient, a linear relation was present between V^DELand (P^{pl, aw}− PEEP), suggesting that muscle relaxation was achieved (fig. 2; average r = 0.95 ± 0.064; range, 0.79–0.99). The estimated compliance averaged 35 ± 9 ml/cm H²O; in the three patients studied at two pressure support levels, the pairs of compliance measurements obtained from the two series (pressure support low–pressure support high) were 22–23, 35–34, and 46–45 ml/cm H²O.

P^{ALV, occl}showed a tight correlation with P^{ALV, pes}(P^{ALV, occl}= 1.11 · P^{ALV, occl}− 2.47, r = 0.88, P < 0.001).

Figure 3shows the comparison of P^{MUSC, occl}with P^{MUSC, pes}. Although the slope is equal to one and the intercept is close to zero, the experimental points are quite scattered. The 95% confidence interval for agreement between the two measures was +5.49 and −5.32, with a bias of 2.7 cm H²O.

PTP^pesand PTP^occlaveraged 6.29 ± 3.07 and 6.47 ± 3.08 cm H²O · s, respectively. PTP^occlwas tightly correlated with PTP^pes: PTP^occl= 0.95 · PTP^pes+ 0.13; the bias between the two techniques was negligible (0.16 cm H²O · s), with a 95% confidence interval between the two measures of 1.94 and −1.61 cm H²O · s (fig. 4). P^0.1averaged 3.15 ± 1.75 cm H²O (range, 1.3–7.9 cm H²O), and the values correlated with PTP^pes(r = 0.76, P < 0.05).

Finally, P^{MAX, occl}was tightly correlated with, but tended to overestimate, P^{MAX, pes}: P^{MAX, occl}= 1.32 · P^{MAX, pes}− 0.81 (r = 0.96, P < 0.001); the confidence interval for the two techniques was between 3.65 and −10.13 cm H²O.

During spontaneous assisted breathing, measuring the patient’s respiratory work can help in setting the mechanical support. In the current study, we reappraised the rapid interrupter technique as proposed by Pesenti et al.  16to estimate the muscular pressure developed by patients during PSV. Particularly, we tested the feasibility of applying the rapid interrupter technique by means of a commercial ventilator, and we obtained reliable measurements of respiratory system compliance and of P^MUSC. When these P^MUSCmeasurements were integrated into an index similar to the PTP, the latter was closely related to the esophageal-derived PTP.

The first result we describe is the measurement of C^rsduring assisted spontaneous breathing, achieved by averaging the C^rsvalues of the occlusion bearing a flat plateau tracing after the occlusion. When trying to measure the compliance value in a patient undergoing assisted spontaneous breathing, adequate muscle relaxation during the plateau phase is mandatory to obtain reliable data. We could not compare the compliance measurements that we obtained during PSV with those obtained during sedation and paralysis. Previous studies have shown, however, that the presence of a satisfactory plateau after the occlusion allows an accurate estimate of C^{pl, rs}when compared with the data obtained in controlled mechanical ventilation.16In the current study, we rely on the intrasubject reproducibility of the measurement and on the tight pressure–volume relation to claim that the compliance measurements we obtained are accurate.

By subtracting P^{ALV, occl}from P^{el, rs}we obtained a measurement of P^MUSC, which was well correlated with the reference esophageal measurement. For the calculation of P^{MUSC, occl}, we did not take into account the plateau pressure of each occlusion; rather, we calculated the expected value of P^{el, rs}. This approach offers two advantages: The first is that the noise possibly present in the single occlusions tends to be canceled when an average value of compliance is calculated. Second, it is possible to obtain estimates of P^MUSCalso from those occlusions in which an adequate muscle relaxation cannot be achieved or if the patient activates his or her expiratory muscles. The population we studied showed relatively low P^0.1values, averaging 3.15 cm H²O, although some subjects showed values up to 8; this might have favored the achievement of muscle relaxation during the occlusions. In this respect, however, it is worth noting that complete muscle relaxation of both the inspiratory and expiratory muscles is not crucial for every occlusion, as far as the compliance values (from which P^{el, rs}is calculated) are adequately computed (as demonstrated by the pressure–volume relations). Moreover, after the compliance value has been adequately estimated on a few occlusions, it could be possible to perform “mini-invasive” occlusions of short duration. These occlusions would only be used to estimate the value of alveolar pressure, whereas the value of the elastic recoil pressure is calculated after the compliance value has been measured. This would reduce the invasiveness of the technique for the patients, particularly if right after the inspiratory occlusions the ventilator could switch back to the preset pressure-support level rather than cycling to exhalation.

Finally, the single values of P^MUSCwere integrated over the contraction time, obtaining a reliable estimate of the PTP. When pooling in one single time course P^MUSCvalues sampled from different V^Ts, it must be kept in mind that each point bears a certain experimental error, in estimation of both P^MUSCand t^occl; moreover, the respiratory efforts of each V^Tare likely to differ from each other. For these reasons, integrating over time the P^MUSCmeasurement in one index (PTP^occl) serves the purpose of providing a “global” measurement of the inspiratory effort, which is more interpretable than the punctual P^MUSCvalues decreasing at the same time the experimental error. Increasing the number of occlusions obtained in each patient might be a path to more reliable results by decreasing the impact of possible outliers. The interrupter technique that we have tested during PSV is conceptually applicable to other forms of assisted spontaneous breathing in which the patients are likely to exploit a greater breath-to-breath variability of the respiratory efforts, such as proportional assist ventilation or airway pressure release ventilation.20Under these conditions, the interrupter technique could still provide an “average” measurement of the metabolic WOB, provided that a sample of occlusions adequately representative of the different respiratory efforts is obtained; yet further studies would be necessary to address this issue. The PTP is usually computed over a minute, taking the effect of the respiratory rate into account.10However, we chose not to perform this operation because multiplying PTP^pesand PTP^occlfor the same number (the respiratory rate) would have introduced a mathematical coupling, spuriously increasing the correlation between the two variables.

The patients tolerated the procedure well, without developing desaturation or hemodynamic instability. The respiratory rate tended to increase compared with baseline. However, we are unable to separate the effect of the occlusions from the effect of the additional dead space because of the presence of the pneumotachograph, which was, unfortunately, quite bulky.

Although the use of a commercial ventilator bears significant advantages in performing the interrupter technique, at least two main limitations of such an approach must be considered. The valve is located inside the ventilator, rather than at the airway opening. In this way, a relatively large volume of gas (the ventilator circuit) is interposed between the valves and the alveoli, representing a compressible volume that can damp the high-frequency oscillations. Even if the valve is closed, the respiratory system volume is not necessarily constant: The pressure swings generated by the patient’s respiratory muscles can displace some gas from the lungs to the circuit and vice versa . Particularly, when the patient relaxes the respiratory muscles, some gas volume can move out of the respiratory system into the tubing. However, the low compliance of the ventilator circuit should minimize the impact of this phenomenon.

Another limitation of using a commercial ventilator, rather than a dedicated valve system, is the closure time of the valves. The valves of a modern commercial ventilator have very high-level performance. The speed of opening and closing is often a compromise between acceptable performance and the wear of the valve itself. The closure times (approximately 70 ms to completely zeroing the gas flow) that we report are greater than those reported in the literature for the rapid interrupter technique (between 5 and 30 ms), and a “long” closure time could affect the measurements. There is a delay between the occlusion and the P^awtracing portion suitable for the slope analysis. During this time, the patient might change his or her effort, making the linear back-extrapolation of the curve unreliable. However, the very fast valves tend to generate oscillations in the pressure signals, making the first period of the occlusion not easily interpretable anyway. Moreover, it has been shown that the reaction time in humans to an airway occlusion is between 200 and 300 ms.21,22Indeed, the P^0.1measurement relies on the assumption that during a 100-ms occlusion, the decrease in the airway pressure is linear.23If the flow interruption is not instantaneous, the volume of the system changes during the occlusion process: The volume in the respiratory system after the occlusion is greater (during the inspiration) or smaller (during the exhalation) than before the start of the occlusion. This translates the entire P^awtracing after the occlusion upward or downward compared with the theoretical one. In our bench study, the volume delivered during the occlusions were relatively small, averaging 34 ± 17 and 26 ± 16 ml during inspiration and expiration, respectively.

Our study population, although with a common diagnosis of acute respiratory failure, was quite heterogeneous for primary diagnosis and severity, and it did not include patients with chronic obstructive pulmonary disease or with a relevant intrinsic PEEP. Indeed, the validity of the technique in patients with long and/or heterogeneous time constants as well as in patients with increased respiratory drive or a very high respiratory rate is questionable and would require specific investigation.

In conclusion, this work shows that the rapid interrupter technique can be performed by means of a commercial ventilator, offering reliable measurement of respiratory system compliance, of P^MUSC, and of the derived parameter PTP. Although further studies are required to verify the robustness and usefulness of the technique in a general intensive care unit population, we suggest that the advantage of performing the interrupter technique without the need for a dedicated apparatus is likely to balance some of its limitations and bears the potential to make the estimate of the metabolic WOB easily obtainable at the bedside.

The authors thank Thomas Handzsuj, M.Eng. (Draeger Medical AG & Co.; Luebeck, Germany), for technical support; and the nursing and medical staff of the Intensive Care Unit of the San Gerardo Hospital (Monza, Italy) for their support and cooperation, without which the study could not have been performed.