A bench-to-bedside study about trigger asynchronies induced by the introduction of external gas into the non-invasive mechanical ventilation circuit

Treatments that require the introduction of external gas into the non-invasive ventilation (NIV) circuit, such as aerosol and oxygen therapy, may influence the performance of the ventilator trigger system. The aim of the study was to determine the presence and type of asynchronies induced by external gas in the NIV circuit in a bench model and in a group of patients undergoing chronic NIV. Bench study: Four ventilators (one with two different trigger design types) and three gas sources (continuous flow at 4 and 9 l/min and pulsatile flow at 9 l/min) were selected in an active simulator model. The sensitivity of the trigger, the gas introduction position, the ventilatory pattern and the level of effort were also modified. The same ventilators and gas conditions were used in patients undergoing chronic NIV. Bench: the introduction of external gas caused asynchronies in 35.9% of cases (autotriggering 73%, ineffective effort 27%). Significant differences (p < 0.01) were detected according to the ventilator model and the gas source. In seven patients, the introduction of external gas induced asynchrony in 20.4% of situations (77% autotriggering). As in the bench study, there were differences in the occurrence of asynchronies depending on the ventilator model and gas source used. The introduction of external gas produces alterations in the ventilator trigger. These alterations are variable, and depend on the ventilator design and gas source. This phenomenon makes it advisable to monitor the patient at the start of treatment.


Scientific Reports
| (2021) 11:23814 | https://doi.org/10.1038/s41598-021-03291-y www.nature.com/scientificreports/ the ventilators used for NIV are not equipped with an oxygen mixer or specific ports for the administration of supplemental oxygen; therefore, supplemental oxygen is usually administered by adding it to the mask or the circuit 11 . This external flow may have an effect on the performance of NIV, especially its trigger function, although this effect has not been studied to date 17 . Thus, our hypothesis was that, when an external flow (of oxygen) was introduced in the circuit, it will impair the trigger function of the ventilator, as it modifies the trigger threshold for flow changes.
The aim of the present study was to evaluate the presence of trigger asynchronies resulting from the introduction of an external gas into the circuit during NIV in two settings: a. In a bench study, under different ventilatory patterns (obstructive or restrictive), sensitivity levels, levels of patient effort 18 and different sources, quantities and locations of the gas introduced (continuous versus pulsatile flow compressor) b. In a short clinical study in patients receiving chronic NIV, in which the bench model previously described is reproduced.

Methods
Design of the bench study. The experiment was performed under simulated conditions in the NIV laboratory. A breathing simulator (series 1101, Hans Rudolph, Inc., Shawnee, Kansas) was used to test the different ventilators, which were connected to the simulator through a single standard 2-m tube with a leak in the expiratory port (Model 5804000, Intersurgical España SL, Madrid) placed at the distal end of the ventilator. The leak port was maintained throughout the experiment to simulate the intentional leak used in clinical practice in NIV with a single-limb configuration. For the introduction of the external gas, a T-piece with a 3.5-mm diameter side port (Model 2713000. Intersurgical España SL, Madrid) was placed proximally to the ventilator first and then distally to the leak.
Three different treatment modalities were added to the NIV set-up: • In the first model, a compressor (Mini Plus, APRES MEDICAL SL, Spain) was connected to a tube in the circuit. The flow of the compressor was monitored prior to placement, and a value of 9 l/min was obtained ( Fig. 1). • The previous model was reproduced in terms of the amount of gas introduced, but a continuous gas source with a pressure reducer and flowmeter (compressed air bottle, Nippon Gases, Madrid) was used. • Finally, in the third model, a continuous gas source was used at 4 l/min. Medical air was used to avoid oxidation of the simulator parts due to the use of pure oxygen.
The simulator was programmed as follows: for lung mechanics variables, compliance was set at 30 ml/cm H2O and resistance was set at 5 cm H 2 O/l/s in a restrictive pattern and at 60 ml/cm H 2 O and 18 cm H 2 O/l/s   Table 1. Figures 1 and 2 explain the settings and protocol of testing.
Signal acquisition system: an external polygraph (16Sp Powerlab, AD Instruments, Australia) equipped with two pressure transducers (model 1050) and two pneumotachographs (S300, instrumental dead space = 70 ml, resistance = 0.0018 cm H 2 O/l/s) was used. The sampling frequency was adjusted to 200 Hz, and the polygraph was connected to a personal computer equipped with Chart 7.0 software for Windows.
The ventilation periods were 1 min for each step (without gas → with gas → without gas). This sequence was performed twice in each group of established conditions. The gas-induced asynchronies were compared by interpreting the graphs of pressure and flow time according to the SOMNONIV algorithm, by three independent observers 19 . Combinations of effort, trigger, and lung mechanics that induced asynchronies without the need of gas introduction were eliminated from the subsequent analysis. admission for acute on chronic respiratory failure, home NIV (single-limb system with intentional leakage) and more than 6 months of use with adequate compliance (5 h/night or more). Patients with underlying psychiatric disease were excluded. The study was conducted during the patient's predischarge phase (the same day or the day before). Patients after a COVID-19 admittance were excluded (all of the included patients had a negative swab test PCR at admission). Due to the epidemiological status of the area, a short period of time between waves was employed. Protocol: The procedure was performed in the patient's room during his or her hospital stay to avoid additional visits. The patient was placed in the supine position, his own home NIV interface was placed in a singlelimb system, along with the commercial ventilator to be studied. A small period for adaptation (5-10 min), if the ventilator tested was different from their home ventilator, was employed, to ensure comfort. The same ventilators tested in the bench model were evaluated at the bedside in random order. The parameters of the ventilator were the same as those that the patients used at home (that is, unlike the bench study, the sensitivity of the trigger was not modified). The monitoring system was essentially the same as previously described, with the incorporation of thoracic and abdominal bands and parasternal electromyography to better evaluate asynchronies, in addition to pulse oximetry control. As the introduction of external gas could cause distortion in the pressure and flow waveforms leading to misclassification of asynchronies, it was helpful the use of monitoring tools to assess patient's ventilatory pattern. Figure 3 reflects an example of a fully monitored patient. The monitoring was prepared before the experiment and maintained until the ventilation was withdrawn. The gas sources were the same as those used in the bench study model. Monitoring of leakage was performed during clinical study through ventilator screen to detect sudden changes in non-intentional leakage that could eventually induce asynchronies. Sequences with changes > 10% in nonintentional leakage were discarded and repeated. All methods were performed in accordance with the relevant guidelines and regulations. The study was approved by the hospital ethics committee of "Comité de ética e investigación con medicamentos (CEIm) del Parc Taulí de Sabadell" (ref. 2019/511). Written informed consent was obtained in the day of the experiment. Trial was registered at Clinicaltrials.gov (NCT04884828).
Statistical analysis. Quantitative data are expressed as means and standard deviation (µ, sd) according to the normality of the variables, which was assessed using the Shapiro Wilk test. Qualitative data are expressed as absolute frequencies. The chi-square test or Fisher's exact test was used in the absence of normality to evaluate the presence or absence of gas-induced asynchrony by comparing the variables controlled in the bench study and the ventilator model, gas source, and interindividual variability in the clinical study. Bonferroni's adjustment for multiple chi-square testing was used. The level of significance was established at p < 0.05. Accepting an alpha risk of 0.05 and a beta risk of 0.2, seven subjects were necessary to recognize a significant difference, consisting in the absence of asynchronies before gas and appearance after gas, considering a drop-out rate of 20%.

Results
Bench study model. A total of 312 simulation sequences were obtained after combining the different variables (gas source, ventilatory pattern, position of entry into the circuit, trigger sensitivity and patient effort). In 67 cases (21.4%), asynchrony under basal conditions was documented; these were therefore excluded from the final analysis, which included 245 combinations. Table 2 shows the differences between ventilators in terms of percentages and types of baseline asynchronies in the traces that were not included in the analysis.
Regarding the factors conditioning asynchrony, statistically significant differences were found in the presence of asynchronies between ventilators, type of gas (pulsatile vs continuous), gas flow (4 vs 9 l in continuous flow), level of effort of the simulator, ventilatory pattern and trigger sensitivity. Table 3 shows the results, differentiated by the asynchronies induced.
After gas closure, asynchronies were induced in 101 situations (96 with autotriggering), and the vast majority (95 in total) were transient. Only in six situations was the autotriggering constant until the end of the phase. Short clinical study. Ten patients were initially selected, one was unable to complete the protocol and two were excluded due to a poor signal quality of the EMG. Seven patients were finally included in the study in the short validation cohort. Their characteristics are detailed in Table 4, which also presents the most relevant ventilator parameters.
After the introduction of external gas, new-onset asynchronies occurred in a total of 22/105 situations (20.4%). The most prevalent asynchrony was autotrigger, in 17/105 tested situations, followed by ineffective effort (5/105). No correction of asynchrony was observed during the period of gas administration. As Table 5 shows, and as in the bench study model, there were significant differences depending on the type of gas and ventilator model used.

Discussion
The main findings of the study are the following. In the bench study, the introduction of external gas to the circuit in NIV was associated with trigger asynchronies in 36% of the situations; autotrigger was the most frequently occurring asynchrony, followed by ineffective effort. In addition, the latter showed a greater degree of spontaneous correction during gas introduction. The occurrence of asynchronies was related to the different ventilators studied, the source and flow of the gas, and lung mechanics. The results of the clinical study also corroborated a significant increase in the number of asynchronies, but in a significantly lower proportion than in the bench study. However, the conditioning factors of the asynchronies (pulsatile flow use and trigger design) were essentially the same as those found in the bench study.
The design of the trigger function of the ventilators used for NIV has undergone significant improvements, always in search of a greater sensitivity (i.e., triggering against decreasing efforts of the patient) and trying to preserve specificity (i.e., avoiding unwanted triggering in the absence of effort). This objective has led to significant differences in design between manufacturers, which may help to explain the findings of this study. Some of the ventilators studied (Astral, Trilogy, PB 560) are equipped with flow trigger systems, which is probably the most frequently used design at present 20 . These systems have shown greater sensitivity than the old pressure trigger models 21 . However, the use of flow thresholds with fixed values for triggering can cause problems in the presence of leaks if the ventilator's leakage compensation flow is higher than its trigger threshold (which might be misinterpreted as patient effort). For this reason, flow trigger systems are often accompanied by an algorithm Table 2. Distribution between ventilators of the simulation sequences in which there was baseline asynchrony (p < 0.01 among ventilators, chi-square test).    www.nature.com/scientificreports/ that can be used to automatically adjust their sensitivity as a function of leakage 22 . These algorithms operate independently of the ventilator's pre-set sensitivity level, but they can lead to paradoxes such as decreased trigger sensitivity in the presence of leaks, which may eventually favour the occurrence of ineffective effort.
To compensate for these drawbacks, alternative (but more complex) designs, such as AutoTrak ® (Philips Respironics), have been devised based on the crossing points that are generated by a delayed virtual waveform and superimposed (using the flow waveform method) on the native waveform 23 or are derived from a calculation based on the so-called trigger energy (Breas) of the first derivative of the flow 24 . In general, these more complex systems tend to be more sensitive but have a greater tendency towards automatic triggering 23 . These differences in design, together with the levels of trigger sensitivity specific to each manufacturer, can explain the differences in the behaviour of the ventilators analysed in this study, even under basal conditions.
Another factor that generated asynchronies was the amount of gas introduced into the system. Aerosol therapy generally requires a minimum gas flow of 8 l/min. This continuous flow contributes to the pressure level that the ventilator reaches during the respiratory cycle and causes a substantial decrease in the flow administered by the ventilator itself 7 . This decrease also automatically modifies the sensitivity of the trigger in the flow models in a manner inverse to the previously described leakage model, and may favour the appearance of asynchronies. The nature of the mechanisms employed to correct trigger thresholds related to leak levels operate in very short timeframes. For example, Resmed uses a moving timeframe of 6 s, that may narrow to 3 s to adjust trigger threshold, marketed as Vsync ®22 whereas other manufacturers employ a limited number of breaths to make any correction. Thus, all these systems operate in a very narrow time frame, under 10 s, allowing for very fast changes in the trigger threshold. This fact may explain why around 30% of initial asynchronies may correct spontaneously in the very next seconds. Finally, the greater number of asynchronies induced by the compressor at equal flow could be attributed to the wide fluctuations in flow generated by the device compared to the continuous flow of gas (see Fig. 1). It should be remembered that the operating mechanism of a compressor is based on the displacement of a piston on a cylinder, which generates a pulsatile flow.
Leak estimation is closely related to the algorithms that may correct trigger threshold in different leak leves. Some authors have evaluated the ability of different ventilators to compensate for leaks 25 and have concluded that certain ventilators cannot guarantee the pre-established tidal volume when involuntary leaks are introduced, a phenomenon that they attribute mainly to the development of asynchronies (autotriggering) rather than to an inaccurate estimation of the tidal volume. Other bench studies have shown that the tidal volume estimation of home ventilators can be substantially modified by the presence of continuous or random leaks 25,26 . Anyway, leak estimation may play a relevant role in the way the ventilator modifies their triggering function in the presence or absence of flow input at the circuit level.
The findings of the present study have implications for clinical practice since they reflect situations (especially the use of compressors with certain ventilator models) that may require ventilator parameters to be modified in some cases.
Some limitations of this study should be considered. First, in both the bench and the clinical studies, the effects were examined for short, limited periods of time. Therefore, we do not know the compensation capacity of the ventilators over longer periods of time, although, based on the recalculation algorithms against leaks at level zero, these usually occur after only a few cycles. Second, the model was applied while the patient was awake; although this is when most similar interventions are performed, the supply of oxygen in the tubing is usually delivered over the long term, and we do not know whether the systematic use of supplemental oxygen is associated with a higher percentage of asynchronies at night.
Third, all of these ventilators are equipped with oxygen inlet ports. Obviously, mixing oxygen with the air previously to be driven by the turbine and sensed by the output flow sensor will not interfere with triggering option. But most of the low-level home care ventilators do not have built-in oxygen ports, and supplemental oxygen needs to be plugged in the circuit, via a t-piece. These homecare ventilators operate with the same algorithms for correcting triggering thresholds as the ones tested here, so results are fully applicable. Even more, to nebulise drugs in the circuit an oxygen input in the circuit is still needed, unless using ICU ventilators with synchronised, built-in flow sources. Finally, to avoid premature oxidation problems with the simulator, pressurized medical gas (21% oxygen accompanied by 78% nitrogen) was used. Under clinical conditions, 100% pure oxygen supplementation would have been used. The different densities of pure oxygen relative to air (relative density to air 1.1052) may modify the calibration, and the ventilator's response to the flow is different from that evaluated in this study. The density (g/l) and viscosity (micropoises) of oxygen (1.429 and 192.6 respectively) are higher than those of air (1.293 and 170.8 respectively).
In conclusion, when an external gas is introduced into the circuit in NIV mode, the ventilators tested show different responses of asynchronies and different degrees of correction. These behaviours differ according to the amount of gas, the trigger level and the external gas source. These findings make it advisable to monitor, and possibly change, the parameters when certain ventilator models and applied gas sources are used.